medical research what's it worth

Browser does not support script.

Go to…

• Visiting faculty
• Evidence Development and Incubation Team
• Experimental Government Team
• Policy Idol
• Student network
• Research impact training
• Research & analysis
• Children & young people
• Cities, growth, innovation
• Communities & opportunity
• Defence & security
• Health systems
• Trust, facts, democracy
• Value of evidence
• News & comment
• Study with us
• Executive education

Medical research: What’s it worth?

Government, charities and the public – through both taxes and donations – invest significant sums of money into medical research each year. Understanding the economic impact of this investment provides accountability, helps secure this investment over the long term and increases our understanding of how research is effectively translated into health improvements. 

Working with RAND Europe and the Health Economics Research Group at Brunel University London, the Policy Institute carried out peer-reviewed research estimating the economic returns generated by public and charitable investment in UK medical research.

This research looks at the returns from research into musculoskeletal conditions, cancer, cardiovascular and mental health research.

The key finding from this work is that every £1 invested in medical research delivers a return equivalent to around 25p every year, for ever.

Read the reports

• Medical Research: What’s it worth? A briefing on the economic benefits of musculoskeletal disease research in the UK
• Medical Research: What’s it worth? Estimating the economic benefits of cancer-related research in the UK
• Quantifying the economic impact of government and charity funding of medical research on private research and development funding in the United Kingdom

Sign up to our newsletter

Latest Research & Publications

• Health Care Systems
• Value, Affordability, and…

Socio-Economic Value of Adult Immunisation Programmes

• Health Technology Assessment…
• Measuring and Valuing Outcomes

Incorporating the Patient Voice in Health Technology Assessment

• Policy, Organisation and…
• Healthcare System Pressure…

Developing a Conceptual Framework of Healthcare System Pressure

The Socioeconomic Burden Of Cervical Cancer in the UK: What are the benefits of achieving the WHO elimination target?

Individual, Health System, and Societal Impacts of Anti-seizure Medicine Use During Pregnancy

• Biosimilars
• Economics of Innovation

The Dynamics of Drug Shortages

2022 OHE Annual Report to the Charity Commission

• Antimicrobial Resistance (AMR)
• Quality of Life

Capturing the Broader Value of Antibiotics

All Publications

• Cell and Gene Therapies
• Chronic Diseases
• Combination Therapies
• COVID-19 Research
• Digital Health
• Drug Development/R&D
• Emerging Markets
• EQ-5D and PROMs
• Health Data and Statistics
• Health Technology Assessment
• Precision Medicine
• Real-World Evidence
• Sustainability
• Use of Medicines
• Value-Based Pricing
• Vaccine Research
• Policy, Organisation and Incentives in Health Systems
• Value, Affordability and Decision Making
• Education Hub
• OHE Graduate School
• EVIA Programme
• The Prize Fund
• 2024 Prize Fund
• 2022 Prize Fund
• OHE Experts
• Drug Development/R&D
• Real World Evidence (RWE)
• Mental Health
• Research Funding

Medical Research: What’s It Worth?

Understanding the nature, extent and processes involved in the return on investment in medical research has been largely neglected as an area of serious scientific study. This report is the outcome of a one-year study commissioned by the UK Academy of Medical Sciences,…

•  Twitter
•  LinkedIn
•  Facebook
• has-icon Email
•  Copy URL

Understanding the nature, extent and processes involved in the return on investment in medical research has been largely neglected as an area of serious scientific study. This report is the outcome of a one-year study commissioned by the UK Academy of Medical Sciences, the UK Medical Research Council and the Wellcome Trust. Its purpose is to compare the economic benefits accruing in the UK from medical research funded in the UK, publicly or by charities only, to the cost of that research. The report includes time series estimates of public/charitable and private pharmaceutical industry expenditure on medical research; explores the concept of knowledge ‘spillovers’ from public and charitably funded medical research; estimates the magnitude of spillovers in the UK; and computes internal rates of return on research investment. Using mental health and cardiovascular disease (CVD) as case studies, the report provides a clear demonstration of an approach that provides an improved theoretical basis for empirically estimating the two main elements of the economic returns from medical research: the value of health gains and the impact on GDP.

Related Publications

Innovation for health system efficiency and improvement.

• Economic Evaluation

Reimagining Prevention for a Healthier, More Prosperous Society

• Value based pricing

Proposals for a Novel UK Antimicrobial Subscription Model: How Will Antibiotic Innovation be Scored?

An error has occurred, please try again later.

Share on Mastodon

• Privacy Overview
• Strictly Necessary Cookies
• 3rd Party Cookies

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.

Strictly Necessary Cookie should be enabled at all times so that we can save your preferences for cookie settings.

If you disable this cookie, we will not be able to save your preferences. This means that every time you visit this website you will need to enable or disable cookies again.

This website uses Google Analytics to collect anonymous information such as the number of visitors to the site, and the most popular pages.

Keeping this cookie enabled helps us to improve our website.

Please enable Strictly Necessary Cookies first so that we can save your preferences!

How valuable is medical research?

This post was written in conjunction with Giving What We Can, and first appeared on their blog . It is part of the Global Priorities Project  series of posts on how we should prioritise research and similar activities.

How valuable is medical research? Finding the solution to a problem in medical research is not only difficult, but it’s difficult to know just how difficult. Given that, how could we know the return we can expect on investing in medical research? Owen Cotton-Barratt is trying to find a framework for tackling precisely these kinds of questions. This post gives a sketch of his current model, and what it might be able to teach us about the value of medical research.

Owen Cotton-Barratt is Giving What We Can’s director of research, and also works on the Global Priorities Project (GPP) [1] within the Centre for Effective Altruism. The GPP works to try to find out how we should compare the different causes that we could be working on. This is closely tied to Giving What We Can’s aim of finding out how individuals can help people the most with their donations.

Owen’s current project is developing a framework to treat problems of unknown difficulty. It’s a work in progress, but he wanted to see whether the kinds of results the model produced would be useful by working through a specific example – the value of medical research. In this post I present a brief sketch of both his model and his results. Note that this work is very preliminary, and my statement of it is rough!

How might you go about finding out how much benefit you can get for an investment in medical research?

You need to build a simplified model that can start to give us a handle on what’s going on, but that gets the major effects approximately correct. This model aims to find a best guess for the average benefit (measured in Quality Adjusted Life Years (QALYs)) produced by an extra unit of resources (measured in dollars) invested in medical research, as well as some idea of how robust this estimate is. It starts by finding the general shape that returns to medical research would have, and then uses rough numbers about the world’s actual research spending and life expectancy increases to find out how many QALYs we might expect to generate per million dollars we put into medical research.

How do the outputs of medical research vary as resources go in?

In research in preparation, Owen argues that we should think the beneficial knowledge from a research area will scale approximately logarithmically, in expectation, with the resources invested. If the research area has many small discoveries rather than a few large discoveries, its development should also exhibit roughly logarithmic behaviour. In the case of medical research, this appears to be a reasonable assumption.

This means that a doubling of resources is equally good regardless of starting point. So money would have diminishing marginal value when put into research, as we might expect. Of course, this model does not hold for arbitrarily large values, but it is a fair approximation over the range we’re interested in.

How many QALYs are produced per $ invested in medical research?

The very rough model above seems to indicate that the knowledge generated by medical research varies logarithmically with money put in. That’s an interesting result, but what we really want to know is how many QALYs we could generate by donating a certain amount to medical research. To assess that, we need to ask how useful extra knowledge is. At first glance, it seems like it generates health benefits in a stream of constant size into the future, so even a small amount of extra research today will produce very large numbers of QALYs.

However, we should assess the value of extra research by the counterfactual where we don’t do this research. Whereas the obvious counterfactual considers the research never happening, the correct counterfactual is generally that the research will happen, but later. Of course the resources that would have been used to undertake this research later will now be freed up, often for further research. But we get diminishing marginal returns from research as the low-hanging fruit are taken, so although there will be a stream of counterfactual benefits going into the future, it will be decaying and its total size may well be finite.

To figure out the actual value, we need some numbers about current medical research. The numbers below are rough, as this is a proof-of-concept calculation. Where they aren’t sourced, they’re Owen’s best guesses. Putting more work into getting good data to use in the model could be a good way to produce more accurate outputs.

Resources put into research

• How much money is spent on medical research in total in a year? Around $270billion, of which around $100billion is publicly funded.[2]
• What about other resources? If we doubled the funding, we wouldn’t double the number of talented people willing to work in the field. And some progress in other areas, such as hardware manufacture, may help medical research. Properly we could consider the elasticity of medical research with respect to these different factors (that is, the responsiveness of the field to change in these factors), with the observation that the elasticities should sum to 1. Alternatively we could estimate the dollar-equivalent value (to progressing medical research) of all of the resources going into the field. We might guess this is between $400 billion and $900 billion, with a median guess of $600 billion.
• How fast is the investment growing? Probably between 2% and 5% per year. [3] The median guess is 3%.

Benefits we get from research:

• How much has life expectancy increased per decade? 2 years. [4]
• Not all of that life expectancy increase is due to medical research. On the other hand, medical research also improves quality of life. Given that, how many QALYs per person does medical progress create each decade? 0.15 to 1.5, median guess 0.5. Hence 0.05 QALYs are added to an expected life per year due to medical progress.
• It’s possible that not everyone benefits from medical research, because its results don’t reach those in the developing world. How many people are benefited by medical research? Somewhere between 1.5 billion and 7 billion. Median guess: 3 billion.
• How much total benefit is produced in one year from the resources we currently put into medical research? The added QALYs are to a person’s total expected QALYs, not the number they get each year. So we model the benefit as just accruing to people dying in a particular year: 1 in 70 people. What we get is 0.05 * 3,000,000,000 / 70, which is approximately 2 million QALYs.

What are the long term benefits and effectiveness of medical research?

You might think that the benefit that medical research produces in a year will continue every year from then on, because once you’ve found a vaccine or cure you can continue to use it. However, we shouldn’t be comparing the case of us finding the vaccine with the case in which it was never found – chances are it would have been found, just later on. On the other hand, the money that would have been used to find that vaccine later on, can now be used to fund other medical research. What we have effectively done is moved medical progress as a whole forward. Therefore, the benefit from funding medical research does last into the future.

In order to find the total value of funding medical research, we need to know how long into the future our funding continues to be a benefit. Since medical funding is increasing each year, the benefit from a particular donation decays over time. How quickly this happens depends on how the benefits from medical research vary with resources put in. Using the result that this is log-linear gives that the benefit of research today decays at a discount rate equal to the rate of increase of medical funding. With a 3% annual increase, the sum of the infinite series generated is equal to 33 times the benefit generated in one year without accounting for this.

The total benefit produced by the $600billion put yearly into medical research is the number of QALYs benefit in a year (around 2 million), times the effective number of years the benefit continues for (33.3). Therefore, the total benefit is around 70 million QALYs. That means per million dollars put in, 110 QALYs are produced. That is $8,500 per QALY. Assuming that the error in our estimates by component are independent, Owen also worked out a rough 90% confidence interval, of between $2,000 and $36,000 per QALY. Accounting for model uncertainty might make it a little wider again.

The result we got – that the return of medical research is around $8,000 per QALY – is very rough, and has a high variance. However, it does give us an idea of its value, which is very useful. There are two particularly interesting things it tells us: This estimate indicates that we are under-investing in medical research. The NHS is typically happy to fund treatments up to around $50,000 per QALY. If medical research on average is more cost-effective than that, rich countries should be investing more in it.

The estimate indicates that medical research as a whole is not typically as effective as the very best direct interventions (like providing people with bed nets). Despite the initial appeal of research as something which will have far ranging ramifications, on average deworming children is more effective.

However, our estimate has been for the average effectiveness of medical research. It is quite plausible that particular research efforts are highly cost-effective, particularly if they are in fields which are likely to be neglected because the beneficiaries would be very poor (for example, developing a vaccine for malaria or schistosomiasis). With further work we could begin to make comparisons of these areas.

• Global Priorities Project
• Wikipedia: Medical Research Funding
• Proposal Exponent: Federal Profiles
• National Institute on Aging

Medical research: what's it worth?

Medical Research:  What’s it worth? Estimating the economic benefits from medical research in the UK

Health Economics Research Group (HERG) Brunel University Office of Health Economics (OHE) RAND Europe For the Medical Research Council, the Wellcome Trust and the Academy of Medical Sciences November 2008

Report produced by

Acknowledgements

Citing this report

This project was commissioned by the UK Evaluation Forum (MRC, Wellcome Trust and Academy of Medical Sciences).

We suggest that you cite this report as follows:

Martin Buxton Steve Hanney Steve Morris Leonie Sundmacher OHE Jorge Mestre-Ferrandiz Martina Garau Jon Sussex Report produced by

RAND Europe HERG Jonathan Grant Martin Buxton

Sharif Ismail

Steve Hanney

Steve Morris

Steve Wooding

Leonie Sundmacher

Institute of Psychiatry OHE

Shitij Kapur

Health Economics Research Group, Office of Health Economics, RAND Europe. Medical Research: What’s it worth? Estimating the economic benefits from medical research in the UK. London: UK Evaluation Forum; 2008.

We would like to thank: members of the Steering Group for their advice and support (Liz Allen, Nick Black, David Cox, Jonathan Haskel, Helen Munn, Briony Rayfield, If you have any queries about this report, Martin Roland, Rhiannon Tudor-Edwards, please contact: Ian Viney and Martin Weale); Thed van Leeuwen (CWTS, Leiden) for bibliometric Professor Martin Buxton Acknowledgements services provided; various colleagues in our Health Economics Research Group respective organisations theirwas helpcommissioned in a RAND Europe This for project by UK Evaluation Brunel UniversityForum (MRC, variety of ways; TonyWellcome Culyer, EviTrust, Hatziandreu and Academy ofUxbridge Medical Sciences). Jonathan Grant and Adrian Towse for reviewing an earlier Middlesex UB8 3PH Sharif Ismail draft; and Nicky Dunne for her like skill to and We would thank: members of the Steering Group for Eddy Nason E [email protected] patience in turning our various their advicedifferent and support (Liz Allen, Nick Black, David Cox, Steve Wooding contributions and frequent changes Jonathan Haskel,into Helen Munn, Briony Rayfield, Martin Roland, a single final document. Rhiannon Tudor-Edwards, Ian Viney and Martin Weale); Thed Institute of Psychiatry

978 1 84129 080provided; 5 van Leeuwen (CWTS, Leiden) forISBN bibliometric services various colleagues in our respective organisations for their help Martina Garau in a variety of ways; Tony Culyer, Evi Hatziandreu and Adrian Jon Sussex for You reviewing draft;personal and Nicky Dunne for her This report is made freely available for non-commercial private study orTowse research. are freean to earlier use or make copies in print or electronic format. skill and patience in turning our various different contributions You may not further reproduce, distribute, or publish any part of the report, except where you have obtained written permission from all the copyright owners. changes a single You must obtain written permission from all the copyright owners if youand wishfrequent to copy the report into or any part of final it for document. a direct or indirect commercial purpose. Jorge Mestre-Ferrandiz

Medical Research: What’s it worth? Estimating the economic benefits from medical research in the UK

Medical Research:   What’s it worth? Estimating the economic benefits from medical research in the UK

Medical Research: What’s it worth? Estimating the economic benefits from medical research in the UK

Executive Summary

Chapter One: Introduction

Chapter Two: Overview of the existing literature and the issues raised

Chapter Three: Expenditure on cardiovascular research

Chapter Four: Analysis of cardiovascular guidelines

Chapter Five: The value of the health gains in cardiovascular disease and their associated health care costs

Chapter Six: R&D spillovers

Chapter Seven: Estimating the rates of economic return from public and charitable CVD research

Chapter Eight: Applying our methodology to mental health

Chapter Nine: Our contribution, our conclusions and main reservations, and a research agenda

Annex to Chapter Two

Annotated bibliography

Annex to Chapter Three Breakdown of expenditure on cardiovascular research (in £m), by year, 1970–2005

Annex to Chapter Four More detailed, qualitative analysis of the importance of UK research in UK guidelines

Annexes to Chapter Five Annex 5A: Studies used to generate QALYs gained for each intervention in each patient group Annex 5B: Estimating the numbers of eligible patients in each patient group Annex 5C: Adjusting for overlapping patient groups Annex 5D: Estimating the uptake rates for specific interventions in each patient group Annex 5E: Adjusting for compliance Annex 5F: Adjusting for polytreatment Annex 5G: Estimates of total QALYs gained by year used to generate Figure 5.1 Annex 5H: Studies used to generate incremental costs for each intervention in each patient group Annex 5I: Estimates of total incremental costs by year used to generate Figure 5.2 Annex 5J: Sensitivity analysis

Annex to Chapter Six Literature review on R&D spillovers

Annex to Chapter Seven Lag structures

Annexes to Chapter Eight Annex 8A: Breakdown of public, non-profit and private expenditure on mental health research (in £m), by year, 1970–2005 Annex 8B: List of mental health guidelines analysed Annex 8C: Studies used to generate QALYs gained for each intervention in each patient group Annex 8D: Estimating the numbers of eligible patients in each patient group, the numbers of initial contacts with community psychiatric nurses and the number of SSRI users Annex 8E: Estimating the uptake rates for specific interventions in each patient group Annex 8F: Studies used to generate incremental costs for each intervention in each patient group Annex 8G: Sensitivity analysis References

The UK Evaluation Forum was first initiated by the Academy of Medical Sciences, the Medical Research Council and the Wellcome Trust in 2004. Drawing together representation from Government, the research councils, medical charities and academia, the broad aim of the Evaluation Forum was to co-ordinate activity in determining the socio-economic benefits of UK medical research. Following an initial mapping exercise of evaluation practices in member organisations, an international symposium was convened in 2005 to discuss evaluation needs and expectations amongst UK research stakeholders and to review what had already been attempted in demonstrating the socio-economic impact of health research in other countries. The outputs of the symposium, and the further deliberations of the Evaluation Forum, were published in the 2006 report Medical Research: Assessing the benefits to society. Chapter 3 of that report summarised previous approaches to assessing the economic and financial impacts of medical research. Particular attention was given to the ‘exceptional returns’ work published in the United States, which suggested that investment in medical research is returned many times over in societal benefits. However, it was emphasised that the US work made a substantial number of important assumptions that may not be applicable to the UK situation. One of the five recommendations of our report was therefore “that research funders should support research to assess the economic impact of UK medical research”.

So, in late 2006, the Academy of Medical Sciences, the Medical Research Council and the Wellcome Trust started a process to commission such research. The overall aim of the work was to compare the macroeconomic benefits accruing from UK medical research with the cost of that research – ultimately to give a quantitative assessment of the benefit of medical research to the UK. It was also expected that the research would critically appraise both the selected approach and previous attempts to estimate the economic returns from research. In this way, the goal was not to obtain a definitive answer about the returns on the investment in UK medical research, but to generate a piece of work that would help to move this young field forward and inform methodologies for future assessments. The work presented in this report, carried out by a consortium involving the Health Economics Research Group at Brunel University, RAND Europe and the Office of Health Economics, certainly fulfils this hope. We are most grateful to the members of this consortium, led by Professor Martin Buxton, for the expertise, care and enthusiasm they have brought to the study. This is an enormously valuable contribution to an important issue for UK medical science and we look forward to working with our partners in the Evaluation Forum and elsewhere to take forward the research agenda presented in this report.

Professor Martin Roland CBE FMedSci Director, National Primary Care Research & Development Centre, Manchester Chair, Evaluation Forum Working Group

Rationale and scope of the study This report is the outcome of a one-year study commissioned by the Academy of Medical Sciences, the Medical Research Council and the Wellcome Trust to compare the economic benefits accruing to the UK from UK publicly and charitably funded medical research with the cost of that research. Understanding the nature, extent and processes involved in the return on investment in medical research has been largely neglected as an area of serious scientific study. Despite a growing international interest in this area there has been relatively little formal analysis of the returns to medical research, particularly in Europe. The study reported here represents an initial step towards rectifying that situation. Past work in the USA and Australia – the ‘exceptional returns’ literature – has attracted publicity. But those analyses contain important flaws. While it is easy to identify the limitations of existing studies, it is less easy – but possible – to reduce these limitations. Our objective is to estimate the returns to UK public/charitable medical research in as transparent a manner as possible, in order to illustrate, to improve on and to explore these limitations and assumptions. This is not intended as a one-off exercise simply to produce a best estimate: rather it is offered as a contribution to an emergent understanding of the issues and as part of a process of establishing a research agenda which should contribute to the production of more robust estimates in future. Economic returns to medical research comprise two, additive, elements: • health gains net of the health care costs of delivering them • GDP gains, that is to say the UK national income that results directly and indirectly from the medical research and the further activity stimulated by it. Both elements are important. Our approach is mainly bottom-up, in contrast to the top-down approach taken in most of the ‘exceptional returns’ literature. Thus we have undertaken detailed analysis of the important research-based changes that have taken place in the treatment of particular disease areas. The returns to medical research as a whole would be the sum of the estimates for each disease area. We initially analysed the returns to public/charitable research from one well-reported therapeutic area, namely cardiovascular disease (CVD), and then tested the same methods in the more problematic area of mental health. We have addressed exclusively the question of what are the economic returns to the UK population and the UK economy from UK medical research. We recognise that UK health research benefits other countries, just as our analysis recognises that the UK benefits from research from the rest of the world. Indeed, some medical research is undertaken in the UK with the expectation that it will predominantly or exclusively benefit health care in other countries (for example most research on tropical

diseases). However, benefits to countries other than the UK are outside the scope of this study.

The contribution our study makes We provide a clear demonstration of an approach which provides an improved theoretical basis for empirically estimating the two main elements of the economic returns from medical research – the value of health gains and the impact on GDP. Our main original contributions are: • A consistent time series of estimates of public/ charitable and private pharmaceutical industry expenditure on medical research in cardiovascular disease and mental health from 1975 to 1992, pieced together from a variety of sources. Given the expected lags between medical research and its impact, this is likely to be the most recent relevant period of research spending to investigate. • A clear conceptual framework to underpin the concept of ‘spillovers’ from public and charitably funded medical research, based on an original broadly scoped literature review. The total social rate of return to an investment comprises the return to the organisation making the investment, the return to other organisations in the same sector (e.g. medical) and the return to all other parts of the economy. The last two are referred to in economic literature as ‘spillovers’, but that is not to imply that they are accidental. On the contrary, ‘spillovers’ are often an explicit objective of investment in research. • Estimates of the magnitude of spillovers in the UK from public and charitable UK medical research, calculated in two different ways: (1) a two-step analysis of the relationship between public/charitable and private R&D and then of the relationship between private pharmaceutical R&D and GDP, and (2) based on the economic literature estimating the social rate of return to public R&D, whether medical or not. • A ‘bottom-up’ approach to estimating the health gain from research. This is a significant improvement on earlier attempts to estimate the economic returns from research in that it is measured in terms of quality adjusted life years (QALYs) and is driven by evidence on the effects and costs of specific research-derived interventions, rather than by macro-level, temporal changes in mortality or morbidity. • A successful test of this approach in two disease areas. The analysis of the gains for cardiovascular disease (CVD) was built up from evidence on 46 different patient indication/treatment combinations, and that for mental health from evidence on six such combinations. • Analyses of UK clinical guidelines in the areas of cardiovascular disease (five guidelines) and mental health (12 guidelines), to provide indicators to inform the important issues of the lag between research

expenditure and health benefits and the attribution of benefits to UK, rather than worldwide, research. • Computation of the internal rate of return (IRR) on past expenditures on research investment in the areas of CVD and mental health, allowing explicitly for the time lags involved and the level of attribution to UK research, with sensitivity analysis around key parameters. The IRR is a convenient way of representing the return to the original research investment, and has the pragmatic advantage that the published empirical literature on the GDP impact of research is expressed in terms of the IRR achieved by that investment. Expressing the return from health gains as an IRR allows it to be added to the IRR for GDP gains to provide an estimate of the total rate of return achieved by medical research. For example an IRR of 10% means that the return to an investment of £1 is equivalent to receiving thereafter an income stream of £0.10 per year in perpetuity. • A comparison of the internal rates of return (IRRs) on research investment from the value of the QALYs gained in these two areas and with the new estimates of the rate of return in terms of GDP effect (which is not specific to individual disease areas).

Expenditure on medical research in the UK Estimates of spending specifically on cardiovascular research are not readily available for either the public or private sectors and we provide here original estimates that have not previously been available. The construction of these estimates relies on data from a number of different sources and various assumptions regarding the split of medical research between different clinical areas. Despite some inevitable uncertainties, we have a time series for total public/charitable research expenditure on cardiovascular disease which represents the research investment that we are studying, and also a series for research expenditure by the pharmaceutical industry. This feeds into our estimates of the indirect impact of public research on GDP. The private figures also make it clear that private sector R&D expenditure greatly exceeds public plus charitable expenditure. We estimated the expenditure on cardiovascular research from the Medical Research Council, Higher Education Funding Councils, Department of Health, British Heart Foundation and Wellcome Trust. Our mid estimates of total annual funding for cardiovascular research in the UK from these public and charitable sources show that it increased from £26 million in 1975 to £88m in 1992 in cash terms, representing an annual percentage increase of circa 7.5%. Expressed in 2005 prices, this equates to a decrease in annual spend from £144m in 1975 to £121m in 1992, with a total expenditure over the period of £2 billion.

Pharmaceutical industry spending on cardiovascular research in the UK grew rapidly in this period: our mid estimate shows a rise to £213m in cash terms in 1992 – 2.4 times the level of public plus charitable expenditure. On a similar basis, total annual public and not-for-profit funding for mental health research in the UK increased from £28m in 1975 to £93m in 1992, representing an annual percentage increase of around 7%. In real terms this equates to a decrease from £155m in 1975 to £129m in 1992. Private pharmaceutical industry expenditure on R&D in mental health was around three times this level in 1992.

The returns to public/charitable medical research Our method To estimate the net value of health gains in the area of CVD we: • reviewed the economic evaluation literature to obtain published figures for the QALYs gained per patient from specific patient group/intervention combinations for cardiovascular disease over the period 1985–2005 • multiplied these figures by estimates of the numbers of users of each intervention, adjusted for compliance rates, to give an estimate of the total QALYs gained from each intervention • monetised the total QALYs gained by multiplying these estimates by published figures on the opportunity cost of a QALY within the current NHS budget – central estimate £25,000 per QALY, i.e. the mid-point of the National Institute for Health and Clinical Excellence (NICE) threshold range of £20,000–£30,000 per QALY • from a review of the economic evaluation literature obtained estimates of the incremental health care costs associated with each intervention and multiplied these by the numbers of users to quantify the incremental health care costs of each intervention. Based on previous studies, we include interventions that are likely to have been important in terms of the health gains they have produced over the period 1985–2005. We used essentially the same approach for mental health. We drew on the extensive economic literature examining the so-called ‘spillovers’ from public/charitable research between organisations and between sectors to estimate the impact of this research on the UK’s GDP. The literature is clear that the spillovers exist, but less clear about the relative importance of different transmission mechanisms. However the literature, especially that looking at the medical and biotechnology sectors, almost without exception takes the view that public research and private R&D are complements, not substitutes. Public research stimulates private, and vice versa. Both kinds of research lead to improved productivity and performance in the economy generally.

Results All our work emphasises to us that our estimates of the rates of return need to be treated with extreme caution. Most aspects of the methods unavoidably involve considerable uncertainties. Therefore all quantitative results are no more than rough approximations. We have generally tried to provide a best/central estimate, and when in doubt have erred on the side of being conservative. In addition we provide high and low estimates around the best/central estimates and undertake sensitivity analyses. The estimated rates of return were very sensitive to the assumed lag between the years when the research expenditure occurs and the years when the ultimate health benefit arises, and to the proportion of the benefit attributable to UK research as opposed to world research as a whole. The issue of lags has often been ignored in the past, but from a policy point of view may be crucial, especially in the context of the current agenda for translational research. Cardiovascular disease

Our best estimate of the total value of the QALYs gained from the specific CVD interventions included in our analysis over the whole period 1985–2005 is £69bn (2005 prices). The upper and lower estimates are £91bn and £55bn respectively. The best estimate of the total incremental health care costs relating to those gains over the same period is £16bn (2005 prices), with upper and lower estimates of £17bn and £11bn respectively. Based on our analysis of citations in UK clinical guidelines in CVD, combined with the findings of previous published studies, we assume that for CVD the proportion of UK health care benefit attributable to UK research lies in the range from 10% to 25% with a central estimate of 17%. Similarly, from our analysis of CVD guidelines and from previous studies, we assume a mean lag between research and impact for CVD treatments of between 10 and 25 years, with a central estimate of 17 years. Our best estimate suggests that for CVD the IRR from the value of UK net health gains alone (ignoring GDP impacts) is just over 9%. Most one-way sensitivity analyses place the IRR within the range of 5–15%. The ‘optimistic scenario’ we examined produced an IRR of over 25%, but in our ‘pessimistic scenario’ the cost of the research investment exceeded the value of the net health gain. We estimated that the GDP gains that result from increased public/charitable medical research deliver an additional rate of return in the range 20–67% (with a best estimate of 30%). These figures are obtained from a small empirical literature, much of it US-centred and only a proportion of it specific to medical research. Hence the application to the UK and to medical research is at best tentative.

Nevertheless combining our estimates, the total health and GDP gains to public/charitable CVD research in the UK 1975–1992 give a total IRR of around 39%. In other words, a £1.00 investment in public/charitable CVD research produced a stream of benefits thereafter that is equivalent in value to earning £0.39 per year in perpetuity. Mental health

For mental health research by the public and charitable sectors in the UK we found the IRR from the net health gains to the UK population for mental health of 7% to be somewhat lower than for CVD (around 9%). Most oneway sensitivity analyses place the rate of return within the range from a situation where the investment exceeded the net benefits to a positive rate of return of just over 11%. Our ‘optimistic scenario’ gave an IRR of over 15%. Available evidence did not permit us to estimate different GDP returns according to the therapeutic area of research. Thus for mental health, as for CVD, our best estimate of the additional rate of return to the public/ charitable research investment from GDP gains is 30%. This gives a total rate of return of 37% for mental health research. These figures cannot be meaningfully compared with the estimates from most other studies, particularly the research from Australia, which not only uses different methods to estimate the returns but uses an unhelpful measure of return on investment. Implications for the future research agenda The limitations of available data, the questions around methodology, and the issues raised by our analysis lead us to identify some key elements for the future research agenda in this area, most prominently: • research funders need to develop and use a standardised (and mapped) way of classifying research funding • further research needs to be undertaken to understand the time lag between research expenditure and health gain • the ‘spillover’ effects of public and charitable research expenditure on the national economy need further, UK-focused, empirical investigation • a deeper understanding of the international flows of knowledge and influence would be valuable • the importance of local research in terms of absorptive capacity: further research to test this would be very valuable, particularly to the health care system and the National Institute for Health Research • we need to understand additionally, what are the global health benefits from UK medical research.

Chapter One

Introduction

KEY POINTS • There is a growing interest in understanding and measuring the return to investment in medical research. • There has been a tension between advocacy and more dispassionate analysis. • Existing studies illuminate the issues but are all flawed. • The aim of this study is to estimate the economic returns to the UK from health research in terms of two main elements: the value of the health gains to the UK population and the GDP gains. • We take two areas of disease: cardiovascular, for which we anticipate there being relatively good data, and mental health, which we anticipate will be more problematic. • In doing this as transparently as possible we hope to contribute to an emergent understanding of the methodological and practical issues. • Additionally this report provides a future agenda for research in this area.

“UK research stakeholders should be more active in demonstrating the benefits that arise from medical research and making the case for continued investment.” (View of the UK Evaluation Forum underpinning its report Medical Research: Assessing the benefits, May 2006)

Background Whether publicly and charitably funded research and the pursuit of new knowledge is seen as an end in itself or as an investment aimed at producing benefits for society, it is inescapable that the resources used to undertake such research could otherwise be put to other uses of benefit to society. In the language of economics: all uses of resources, including research, have opportunity costs. Research needs to be of demonstrable value to justify denying people the other opportunities to benefit that they forgo in order to pay for the research. In the medical field it is relatively easy to find examples of specific research that has led to enormous benefits in terms of a life-saving intervention or to major improvements in the quality of life of patients with a chronic disease. But understanding the nature, extent and processes involved in the return on investment in medical research has been largely neglected as an area of serious scientific study. Despite a growing international interest in this area (UK Evaluation Forum, 2006; European Science Foundation/European Medical Research Councils, 2005; Natural Environment Research Council, 2006), there has been relatively little formal analysis of the returns to medical research, particularly in Europe. Different stakeholders have different reasons for understanding better the returns to medical research. In most cases, there is some tension between assessments to inform future decisions (about how much to spend, on what areas or categories of research, using which funding systems or mechanisms) and to demonstrate the value of past spending. With this tension also comes a narrow line between objective analysis and selective advocacy. In the USA, it was an advocacy initiative of the Mary Woodward Lasker Charitable Trust, entitled Funding First, which commissioned from a group of leading US academic economists a series of papers. These were presented at a conference in December 1999, the summary of which was entitled ‘Exceptional Returns’ (Funding First, 2000). The subsequent book edited by Murphy and Topel (2003) provided a springboard for renewed interest in estimating the returns to health or medical research. In fact the book itself does not provide formal estimates of the return on investment in medical research. The ‘exceptional returns’ are implied by informal comparison rather than calculated in a systematic manner and come from the summary of the conference rather than from within the papers themselves: Using the methods of their earlier calculations, Murphy and Topel estimate that the total economic value to Americans of reductions in mortality from cardiovascular disease averaged $1.5 trillion annually in the 1970–1990 period. So if just one-third of the gain came from medical research, the return on the investment averaged $500 billion a year. That’s on the

Chapter 1: Introduction

order of 20 times as large as average annual spending on medical research – by any benchmark an astonishing return for the investment. (Funding First, 2000, p. 8) Some of the key assumptions and limitations of this ground-breaking work, particularly when viewed from a less insular UK perspective, were highlighted by a study produced in Australia for the Australian Society of Medical Research (Access Economics, 2003). This built on aspects of the US analyses and produced estimates of the annual rates of return to Australian investment in health R&D. This study also concluded that the returns were exceptional, with a base-case estimate of a benefit to costs ratio for health R&D overall of 2.40 (with low and high estimates of 1.08 and 5.04). For ‘cardiovascular’ disease the ratio was 7.88 (3.55 to 16.56), but for research on ‘nervous system and mental’ the returns were much lower at 0.23 (0.10 to 0.48). But this study compared investment to the value of the health gain in the same year. This study has recently been updated (Access Economics, 2008). It uses a substantively changed methodology, comparing past health R&D expenditure with projected future health benefits. The estimated benefit to cost ratio for Australian health R&D as a whole is now estimated as slightly lower at 2.17 (with a range from 1.16 to 3.34). The new report downplays the significance of estimates for different disease areas. It is easy to identify the limitations of these studies and possible, but less easy, to reduce the limitations. Our objective is to estimate returns in as transparent a manner as possible in order to illustrate, to attempt to improve on, and to quantitatively explore these limitations and assumptions. In doing this in an explicitly critical manner, we have undertaken this study, not as a one-off exercise to produce a best estimate but as a contribution to an emergent understanding of the issues and as part of a process of establishing a research agenda which should contribute to the production of more robust estimates in future. The estimation of returns should not be seen as some isolated (or even arcane) technical process. Rather, it brings together and encourages us to be explicit about our understanding of the way in which medical research may lead eventually to valuable outcomes.

Scope of the study For this study we have conceptualised the economic returns to the UK as consisting of two main elements. Firstly, the value of the health gains to the UK population believed to be attributable to the UK research investment, and secondly, the GDP gains. Both are important. Most health researchers and research funders would probably see the objective of their research as being (ultimately) to lead to better health or better health care, and this is important from a national perspective as well, even though national accounts do not measure and value changes in the stock of the population’s health. But governments also increasingly see investment in publicly

funded research as having a major role in contributing to GDP via its effect through the private sector. Given the scale of the commissioned study and our intended approach, it was clear that we could not address all areas of health research. Our approach is mainly bottom-up: it involves detailed analysis of the important evidence-based changes that have taken place in a particular clinical area. The overall returns to medical research would be the sum of the estimates for each clinical area. We therefore chose to focus initially on research in one therapeutic area, namely cardiovascular disease (CVD). This was a focus in the US study; it is an area where clinical developments have had a clear impact on morbidity and mortality and where there have been previous attempts to estimate the mortality impact of specific interventions (eg. Unal et al., 2004). Thus treatment and prevention of CVD provides an area which might be most amenable to rigorous and relatively robust estimation of returns. But this also means that it probably is not typical and any quantitative results may not be generalisable to other areas of medical research. Therefore, in addition, we have taken the methods we develop for the cardiovascular area and have examined how they might be applied to the much more problematic area of mental health. Generally there has been less work on the relevant data for this area and fewer economic evaluations of specific interventions. Whilst the returns from mental health research have been previously considered, they are sensitive to assumptions about changes in the cost of services and rely much more heavily on estimates of improvements in quality of life rather than length of life. As a result studies of returns to research relating to mental health have shown very different results (Access Economics, 2003; Silverstein et al., 1995; Weisbrod, 1983). An intentional but key limitation of our scope is that we have addressed exclusively the question of what are the economic returns to the UK population and the UK economy from UK health research. We recognise that UK health research may benefit other countries, just as our analysis recognises that the UK benefits from research from the rest of the world. Indeed, some medical research is undertaken in the UK with the expectation that it will exclusively, or predominantly benefit health care in other countries (for example most research on tropical diseases).

Structure of the report Our study sets out to: • review the literature and identify key issues for our study (Chapter Two) • estimate time series for annual expenditures on medical research relating to cardiovascular disease from public and charitable sources, and present time series data on private sector research in the same area, both for comparison and to facilitate our calculations of GDP effects (Chapter Three) • estimate what proportion of those gains should be attributed to UK research (Chapter Four) • estimate the time lag between research expenditure and its clinical application (Chapter Four) • estimate a time series for the annual value of health gains to patients in the UK from the most important interventions in CVD (Chapter Five) • estimate the costs of delivering the health care that produced those gains (Chapter Five) • estimate the effects on national income of the public/ charitable research (Chapter Six) • using data from earlier chapters estimate the return on investment in medical research in terms of the net value of health benefits and combine the estimates of health and GDP returns (Chapter Seven) • test the methods used above in the context of mental health research (Chapter Eight) • identify the key limitations in this approach and propose an agenda for future research (Chapter Nine). Each step involves making estimates based on imperfect data. We have aimed to be explicit about these assumptions and the data weaknesses (in broad terms within the chapters themselves and in more detail in the supporting annexes). We have also aimed to err on the conservative side – avoiding underestimating the research investment or overstating the returns. We present ‘best’ estimates within a range where we feel able to do so. Where we simply have a range, and no other basis to judge, we present the ‘central’ estimate and the range.

Chapter Two

Overview of the existing literature and the issues raised

KEY POINTS • A previous review of the diverse but limited literature on the economic returns from specific programmes of health research identified four main approaches: cost savings; benefits from a healthy workforce; the value to society of the health gain; and commercial development. The last two are now seen as the most promising approaches. • Innovative ways of assessing the value to society of the health gain were developed by the Funding First team in the USA and in follow-up studies in Australia. • Increasing attention is being focused on how research impacts on (creates GDP ‘spillovers’ for) the rest of the economy. • A previous study, using bibliometric approaches to assess the contribution made by UK research to UK clinical guidelines, may provide an indication of the relative contribution of UK research to UK clinical practice.

Introduction Several overlapping issues are relevant to assessing the economic benefits of health research. These include the items to be considered benefits and the ways of assessing them. There is no clear consensus about the best approaches to use. In the UK the Evaluation Forum (bringing together the Academy of Medical Sciences, the Medical Research Council and the Wellcome Trust) considered ways of assessing the benefits to society from medical research (UK Evaluation Forum, 2006). They recommended the “improved use of existing evaluation tools, greater sharing of good practice and the development of new approaches where required” (para. 5.3). This Chapter reviews some of the main literature to assess current practice and highlight issues that need to be considered in our empirical analyses. It focuses, in particular, on studies on which our analyses build. A more detailed review, undertaken as part of our study, on the literature relating to the estimation of the contribution of research to GDP, is reported in Chapter Six.

Alternative measures of economic benefit The report from the Evaluation Forum organised its analysis of the areas of economic benefits from health research around the approach developed by the Health Economics Research Group (HERG), Brunel University (Buxton et al., 2004). The HERG review of previous assessments of economic benefits from health research identified an extremely diverse but not very extensive literature and classified studies in terms of four main approaches to measuring economic benefits:

1 valuing direct cost savings to the health system 2 valuing benefits to the economy from a healthy workforce 3 measuring the value to society of the health gain 4 valuing the benefits to the economy from commercial development. Valuing direct cost savings to the health system Studies have shown that health research can lead to new treatments that reduce the overall cost per patient or the number of patients that need to be treated. Some of the clearest examples relate to vaccines or drugs that have resulted in significant reductions or the virtual elimination in some countries of diseases such as TB or polio. Research-based moves towards the control of Chagas disease in the Southern Cone countries of South America have led to considerable cost savings for these countries’ health care systems (Moncayo, 2003). Other areas of research from which cost savings have been identified include health technology assessments (HTAs). Jacob and McGregor (1997) looked at HTAs undertaken in Quebec, Canada, and found that several of these had directly influenced policy and contributed to health care cost savings. Benefits to the economy from a healthy workforce Measuring only health care savings is generally seen as too narrow a focus, and many studies also consider the benefits, or indirect cost savings, in avoiding lost production. Using the human capital approach, which values health gains in terms of the value of the production gained, Mushkin (1979) attempted, despite data problems, to calculate the economic benefits to the USA of all health research. In a series of calculations she estimated the economic value of the total reduction in mortality and morbidity in the USA between 1930 and 1975, and the value of the share caused by biomedical research and, after deducting the cost of the US research, produced a rate of return of 47%. At a smaller scale, a series of case studies by the US National Institutes of Health (1993) analysed particular pieces of research and included estimates of the saving from the lost production that had been avoided as a result of the research results. There are two well-recognised problems in using this human capital approach (Drummond et al., 1992). While it tends to overstate the benefits at times when the lost labour can easily be replaced by unemployed people or through migration, it measures benefits from improved health only to those of working age, and ignores benefits to the rest of the population. Thus, as a measure of the value of any health-related activity it also has uncomfortable equity implications.

The value to society of the health gain More recent studies have attempted to estimate a value of the health gain without resorting to human capital approaches. A major study that has attracted considerable attention is the Funding First report (2000), which was referred to in Chapter One. This piece of advocacy concluded that “the likely returns from medical research are so extraordinarily high that the pay-off from any plausible ‘portfolio’ of investments in research would be enormous”. The basis for this and other such impressive claims lies in a series of highly technical papers, which, while broadly compatible in their approach, differ in terms of some of their detailed analyses and which were subsequently published in the book Measuring the Gains from Medical Research (Murphy and Topel, 2003). A key element of the evidence base underpinning this work is the US research suggesting that individuals’ average willingness to pay for small reductions in the risk of death is equivalent to a value of around US$5m to prevent a fatality or gain a ‘statistical life’. This is then included in calculations of the economic value of the increasing longevity of the US population. The authors then consider what proportion of these gains can reasonably be attributed to medical research. In the area of cardiovascular disease, for example, they cite evidence to suggest that one-third of the decline in cardiovascular disease mortality is due to invasive treatment, one-third to pharmaceuticals and the remaining one-third to behavioural changes. But the complexity of the links between research findings and practice and behavioural changes are also emphasised; one paper, for example, looks at the impact of specific treatments in part of the CVD field and claims to demonstrate that sometimes clinical practice changes prior to formal clinical trials being conducted and at other times there are considerable lags (Heidenreich and McClellan, 2003). The importance of the diverse papers lies in their common use of a willingness to pay value of a statistical life (or life year), which enables the value of the health gain to be estimated. The robustness of the empirical value they use can be questioned, as can many more detailed assumptions they make in the Funding First report, which drew selectively on the series of papers. Nevertheless, this study provides an important influence on, and insights for, our study of the health benefits from research presented here.

analysis of ‘disability adjusted life years’ (DALYs) in Australia which suggested that for each year of life lost due to premature mortality there was a further 86% of a DALY lost due to disability, worth AUS$129,000. They assumed this relationship held for all disease areas. With these estimates they valued the mortality and (estimated) morbidity gains in Australia between 1960 and 1999, and noted that cardiovascular improvements accounted for one-third of these gains. On the other hand, using this approach, there were increases in mortality due to mental illness and hence a negative value to the measure of health ‘gain‘ from mental health. (In the rates of return calculations, as quoted below, this fact was masked by combining the clinical areas of ‘mental health’ and ‘nervous system’ which taken together had a small positive health gain.) Given they were starting from overall changes in mortality/morbidity it was then necessary to estimate what proportion of this should be attributed to health research, rather than general changes in economic wellbeing, education or diet for example. Despite recognising that there was no robust Australian or international economic evidence to separate out these effects and that “the eminent American economists were also not willing to hazard a guess”, they took the view that “health R&D has directly, indirectly or serendipitously accounted for at least half of the gains” (Access Economics, 2003, p. 62). But they recognised that world pool of R&D was the main contributor, and used the proportion of Australia’s contribution to biomedical research outputs (2.5%) as the percentage of the health benefits attributable health research that were attributable to Australian research. A final key assumption relates to their handling of time lags. The overall analysis in the US study effectively ignored lags and compared gains over the same period as the research spend, and the Australian study adopts the same approach: comparing the gains estimated from the year 1999 with the R&D investment in 1998–99. This gave, for their base case, an overall benefit/cost ratio for health research of 2.40 (i.e. they estimated that the overall health gains – the benefit – in 1999 were valued at 240% of the value of the total health R&D – Figure 2.1: Spillovers from private pharmaceutical R&D

The subsequent extensive study in Australia built on the foundations of the US research and attempted to estimate the returns to investment in research in Australia for each main disease area. It used temporal changes in mortality and in morbidity by disease area. Like Murphy and Topel the Australian study used US$5m as the base-case value of a ‘statistical life’ which they calculated was equivalent to a value of a year of life of AUS$150,000. They extended this to allow for improvements in quality of life on the basis of an

Return to other sectors Return to the pharmaceutical sector as a whole Private rate of return to the companies investing in R&D

the cost – in 1998–99). For cardiovascular the base case ratio was 7.88, whilst for nervous system and mental health the value was just 0.23. (They do not state, and it is not possible to estimate from the published figures, what the ‘return’ would be for mental health alone.) This Australian study has recently been ‘updated’, but with substantial changes to the methodology used. The value of a statistical life year has been increased to AUS$266,843 (at 2008 prices) based on a new meta-analysis of studies and the Australian contribution to world R&D outputs has been increased to 3.04% based on more recent bibliometric data on clinical research. Most fundamentally, the analysis now deals with the issue of the lag between R&D spending and the health benefits that are attributed to that spending, by comparing R&D expenditure (in 1993–2005) with ‘projected’ health benefits in 2033–2045 (thus assuming an average lag of 40 years). These ‘projected’ health benefits are based on projections of the future burden of disease from the Australian Institute of Health and Welfare up to 2023, extrapolated forward further to 2033–2045. Thus their current estimate of the return to health R&D investment is not based on an analysis of what has happened but on estimates of what might happen in the future (and logically these estimates themselves should depend on assumptions about the effect of recent/current R&D on future health of Australians). Curiously the headline benefit/cost ratio remains similar (2.17 compared with the earlier 2.40) but this later report does not provide estimates of the return by disease area. One reason for this is that their estimates of ‘projected’ benefits indicate that the return if calculated would be negative for ‘diabetes mellitus, endocrine and metabolic disorders’, for ‘mental disorders, nervous system and sense organ disorders’ and for ‘musculoskeletal diseases’. The report notes that this does not mean that R&D undertaken in these areas was not effective, but that “current R&D spending in these areas is not sufficient to outweigh the expected increase in these disorders”. In which case, the figures for the aggregate of all disease areas also cannot be taken as an indicator of the effectiveness and hence the return on R&D. In a recent study, Johnston et al. (2006) adopted a hybrid approach to assess the costs and benefits of phase III randomised trials funded by the US National Institute of Neurological Disorders and Stroke. Yearly total incremental net benefits of the programmes were calculated by combining trial costs and treatment costs with a monetary value for the QALYs gained from the implementation of the trial findings. The study used a value of a QALY based on US GDP per capita, suggesting that this reflected “the average yearly economic productivity of a US resident, regardless of employment or age”.

Benefits to the economy from commercial development A major UK review identified a range of categories of benefits to the economy but found that none of the studies provided a simple and comprehensive model (Salter and Martin, 2001). In its evidence to the US Congress, the National Institutes of Health cited several studies showing the importance of publicly funded research to the development of significant new drugs. In one study 15 of the 21 drugs identified as having had the most impact on therapeutic practice were developed with input from the public sector (Joint Economic Committee, US Senate, 2000). This study also stressed the complex interaction between publicly and privately funded research but it made no attempt to calculate the social rate of return. For our study we undertook a review of the literature on how public research impacts, or ‘spills over’ to, the rest of the economy. This identified a clear consensus that the relationship between public and private research is complementary, though complex. Incremental public research is associated with increased private research. The published literature finds that the relationship between public and private research sectors works in both directions: public research stimulates private, and private research stimulates public. There is a small literature estimating the size of this complementarity and also a modest number of studies estimating the impact of GDP of public and private research. Figure 2.1 illustrates, for the pharmaceutical sector, the concept of spillovers: R&D investment undertaken in one organisation not only generates benefits to itself (the ‘private’ return – shown in purple) but also could bring about economic benefits to other organisations operating in the same sector (shown in blue) and in other sectors of the economy (shown in green). The main sources of these positive externalities are scientific and technical advances and, more generally, knowledge flows generated and induced by R&D activity. The sum of all three sets of benefits represents the total ‘social’ return to the original investment in R&D. Garau and Sussex (2007) provide an estimate of the spillovers generated by the (private) pharmaceutical R&D carried out by two major companies in the UK. The range of estimates is based on empirical economic literature looking at the rate of return to private R&D.

The use of a willingness to pay estimate of the value of the health gained, as in the US study, has a clear logical basis in cost/benefit analysis, although there are well recognised problems in obtaining good empirical estimates, and available estimates vary considerably (Nordhaus, 2003). However, a value based on willingness to pay (or willingness to accept) is not necessarily the most appropriate for this purpose, particularly if the analysis is aimed to reflect the opportunity of investing in health research rather than directly in health care. In such circumstances, the amount the NHS would normally be prepared, and able given its budget, to pay for health gain might be more appropriate. This line of argument would point us towards the ‘threshold’ used by the National Institute for Health and Clinical Excellence (NICE) for the maximum value it is prepared to pay for a quality adjusted life year, which lies between £20,000 and £30,000 (NICE, 2004). Current research commissioned for NICE may indicate whether the NICE threshold approximates to a UK social willingness to pay value or not. Where the focus is on the value of the health gain itself, some assessment has to be made of the cost of achieving this benefit. The main cost is likely to be to the health care system, but for public health and behavioural interventions the cost may be much more widely spread. It would seem logical to assess these costs and deduct them from the value of the benefits, as a necessary means to implement R&D. This was clearly done in the study by Johnston et al. (2006) but the Australian study appears to make no allowance for these costs of delivering health gain through the health care system. On the other hand studies assessing the return to investment in health care attribute the value of the health gain to the ‘investment’ in the health care system rather than to the research investment behind improved treatments etc. (Luce et al., 2006; Cutler et al., 2006). The HERG review of studies of the economic benefits from health research (Buxton et al., 2004) suggests that there are two issues of concern even before we get to the question of how we value the benefits.

Figure 2.2: Nationality of papers cited in 15 UK clinical guidelines published between 1996 and 1998 and world share of biomedical papers Percentage of cited papers (integer counts)

Several issues emerge from this grouping of benefits. It is clear that both the second and third approaches place a value on health gain, but that these are alternatives: to use both concepts would involve double-counting. Furthermore, if the value of the health improvement can be measured by either of these approaches it will make a more significant contribution than the first area that concentrates on cost savings to the health care system. Moreover valuing the health gain is compatible with recent broad definitions of the economic impacts of research, such as that given in the Warry Report on increasing the economic impact from (all) research councils in the UK. This report suggests that the “effects on the environment, public health and quality of life” should be included (Warry, 2006).

45 Guidelines

Biomedicine (1998)

35 30 25 20 15 10 5 0 UK

Figure 2.3: Age of papers cited on 15 UK clinical guidelines published between 1996 and 1998 200

150 Frequency

Key issues in estimating the return to R&D

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 36 41 48 Age of cited papers (years)

The first concerns the inputs in question. A number of key science policy studies have emphasised the complexity and range of the research that can lie behind advances in health care (Comroe and Dripps, 1976; Raiten and Berman, 1993). It is often unclear precisely which research has contributed to specific health advances. In the studies examined in Buxton et al. (2004), the breadth of the research considered varies considerably. Some studies consider the impact from specific research projects or programmes, while others look at a broad field of research (e.g. cardiovascular research) and a few have attempted an overall assessment of medical research.

The second issue relates to the relationship, or attribution, between the research inputs and health and other outcomes. Overall, it is not known for certain how far research, from whatever source, has contributed to advances in health. For example, McKeown (1979), in an analysis challenged by Mushkin (1979) and more recently by Sussex (2000), suggested that during the 20th century much of the reduction in mortality was due not to medical advances and medical research, but to improvements in general living standards. If temporal change in health is used as a measure of benefit then some of that change should probably be attributed to broader factors other than simply medical research. This latter point is an example of the general issue that we can only observe the events and outcomes that followed past investment in research: we cannot observe what would have happened without that research. Any analysis of returns on that research involves an explicit or implicit assumption of the counterfactual – what would have happened if the research had not been undertaken. As a result there will always be difficulties in establishing unequivocal links between research and its impact. The problem of attributing benefits to specific pieces of research is a particularly acute problem and has led some authors to question the value of attempting to assess the health and economic benefits from research (Royal Netherlands Academy of Arts and Sciences, 2002). There have been criticisms of some series of case studies, including those undertaken for the National Institutes of Health (NIH, 1993; Johnston et al., 2006), because they seem to have made assumptions about the level of implementation of NIH research. It is claimed that one way of attempting to address these problems of attribution in case studies is through use of the payback framework (Buxton and Hanney, 1996; Hanney et al., 2004; Wooding et al., 2005). We have already noted the problem that the research from any one country adds to the international pool of knowledge. Whilst perhaps this problem can be ignored in the USA, which produces a sizable proportion of global health research, elsewhere it must be addressed. The Australian study used its proportion of world health research output as a proxy for the indicator of how much of the benefit could be attributed to Australian research. The analysis of papers cited in national clinical guidelines provides information that could offer an alternative estimate to help link research to practice and inform estimates of the economic return from research. Whilst the UK contributes about 10% of the global scientific research effort (King, 2004), we know from previous work that UK guideline authors disproportionately cite UK research (Grant et al., 2000). In that study the authors examined the bibliographical details of 15 disease management guidelines that covered a range of conditions seen in family practice in the UK. All the cited papers were looked up on the Science Citation Index and additional bibliographic information, including authors’ addresses, was captured. The country of authorship

was derived from an analysis of the addresses showing that 25% of papers cited in a sample of UK guidelines had authors with a UK address, whilst the UK only contributed 10% of the world’s biomedical research papers (Figure 2.2). These data might suggest that between 10% and 25% of health gains could be attributed to UK research. This analysis of guidelines is repeated in Chapter Four and Chapter Eight for cardiovascular and mental health guidelines respectively. A further issue that has been addressed is the proportion of UK research that is funded from public and charitable funding and the proportion funded by industry. Dawson et al. (1998) reported that of biomedical papers with a UK address 18% of cardiology papers and 17% of neuroscience papers acknowledged industry funding, most but not all which came from the pharmaceutical industry. The analysis of guidelines described above calculated the median time lag between guideline date and publication date of cited papers. This provides one estimate of the time lags between when R&D spending occurs and when society benefits from that R&D as a result of the adoption of new interventions or treatment strategies. Grant et al. (2000) estimated that the median ‘knowledge cycle time’ for all 15 guidelines was eight years, with 25% of papers cited being more than ten years old and 4% more than 25 years old (Figure 2.3).

Conclusions Our brief selective reviews of the various literatures suggest that no single existing study adequately addresses the issue of the economic returns to health research. We see the way forward as being to build on, but significantly improve, the methodology of the most promising and relevant approaches, and using them to produce empirical estimates for the UK. Clearly reliable estimates of past UK research investment are required for our two areas of focus: cardiovascular research and mental health research. These are developed in Chapter Three. We then focus on the two main elements of economic benefit which represent the returns on this investment: 1) the net value of the health gain attributable to UK public/charitable research investment and 2) the GDP effect of that investment. These are conceptually quite different and potentially additive in assessing an overall return on research investment. Our chosen methodological approaches to these are set out in Chapters Five and Six. In the final Chapter we summarise the key developments to previous methods that we have made in this study.

Chapter Three

Expenditure on cardiovascular research

• We estimated the expenditure on cardiovascular research from the Medical Research Council, the Higher Education Funding Councils, the Department of Health, the British Heart Foundation and the Wellcome Trust. • Total annual funding for cardiovascular research in the UK from these public and charitable sources increased from £26m in 1975 to £88m in 1992 in cash terms, representing an annual percentage increase of circa 7.5%.

Figure 3.1: Total estimated cardiovascular research spend (public and non-profit), by source of funding, 1975 to 1992, at current prices 100 90 80 70

• In 2005 prices, this equates to a decrease in annual spend from £144m in 1975 to £121m in 1992, with a total expenditure over the period of £2,026m. • Pharmaceutical industry spending on cardiovascular research in the UK grew rapidly in this period and rose to £213m in 1992 – 2.4 times the public/charitable expenditure.

1 Funding Councils

1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992

Figure 3.2: High and low estimates for public and non-profit cardiovascular research expenditure, 1975 to 1992, at current prices 120

Estimates of public/charitable funding Figure 3.1 illustrates the expenditure on cardiovascular research from 1975 to 1992 by six sources of funding from the government and charitable sectors. These graphs are based on a number of overarching assumptions about the nature of R&D funding in the UK. First, we have assumed that the Medical Research Council (MRC), Department of Health (DH) and Higher Education Funding Councils are the sole government funders of cardiovascular research in the UK. This will cover most of total government funding in this area, although it should be noted that there may be small

1 High 2 Mid-point 3 Low

High-low estimates (£m cash)

The focus of our study is the impact of publicly and charitably funded R&D expenditure in the UK. A part of that impact is mediated through the interrelationship between public/charitable medical R&D and private sector medical R&D, and the literature suggests that a proportion of private R&D is likely to have been stimulated by public/charitable R&D. The direct returns from private sector R&D are reflected in the costs associated with interventions that result in cardiovascular health gains and these are therefore netted off from our estimates of monetary value of health gains (see Chapter Five). In addition, public/ charitable research may have a more general effect on GDP through its impact on the productivity of private sector R&D (see Chapter Six). In this Chapter we provide estimates of the magnitude of past spending on cardiovascular research from both the public/charitable and for-profit (pharmaceutical industry) sectors.

pockets of relevant research funded by other Research Councils or by other Government Departments that we have not captured. For the non-profit funders we have assumed that cardiovascular research is supported principally by the Wellcome Trust and the British Heart Foundation (BHF). There are, of course, other non-profit funders of cardiovascular research (such as the Stroke Association) but data available from the Association of Medical Research Charities website indicate that the Wellcome Trust and BHF currently account for 95% of funding by medical research charities between them1, and for this reason we have focused on these two organisations. Finally, we have assumed that the definitions of the cardiovascular field used by the research funders we

include capture the range of the basic research that may have contributed to developments in this area. Our impression from scanning the grant titles included suggests that this is the case, and that even fairly fundamental basic research tended to be classified into clinical fields. This is an important consideration since the difficulty of establishing clear links between basic research and applications means that a proportion of contributing basic research will inevitably be excluded by restrictive field definitions. The data presented in Figure 3.1 derive from a number of different sources and, of necessity, include further assumptions and extrapolations in some cases. We describe these in further detail below; on this basis, we provide a ‘high’ and ‘low’ scenario for total cardiovascular research expenditure as shown in Figure 3.2. • Wellcome Trust (WT). Wellcome Trust expenditure on cardiovascular research is derived from the Wellcome Trust grants database using a combination of keyword searches and classification terms developed and used by the Trust. Historically grants have been classified by Grant Officers using a thesaurus of terms. We selected all grants that had been classified as ‘CARDIOLOGY’, ‘VASCULAR DISORDERS’, ‘CARDIOVASCULAR DISEASE’ and ‘VASCULAR SYSTEM’. In addition a list of search terms developed for a bibliometric search filter for cardiovascular research was applied to the titles of awarded grants (a full list of key words is available from the authors). • Medical Research Council (MRC). Between 1972 and 2005 the MRC has used five different systems for classifying grants. The longest time series is for 1976/77 to 1992/93 and this was the series used for the basis of our estimates. For this period annual spend on funded grants was classified in two ways; the first was based on the primary purpose of the research and provided an ‘exclusive’ measure of spend by a number of headings (in this case ‘CARDIOVASCULAR SYSTEM’). The second was a more ‘inclusive’ measure where spend could be placed against a number of different categories. For the current project we have used the ‘inclusive’ measure as this is the larger value and errs on the side of caution in our ultimate aim of generating a measure of the rate of return on investment. • British Heart Foundation (BHF). The BHF data are taken from their annual reports for the years 1975 to 1985 and 1989 to 1992. We assumed that all BHF grant expenditure was directed at cardiovascular research. Data for the intervening period 1986 to 1988 were interpolated using a linear function. • Department of Health – activity (DH–A). The DH data include the NHS (which was only collected from 1995) and Department expenditure. The NHS data have been ‘backcast’ (using an exponential function)

for 1975–92, using 1995–2005 data published in the online version of the Science Engineering and Technology Indicators (SET) Statistics.2 The DH data for 1986 to 1992 are taken from the online version of the SET Statistics, and for 1981 to 1985 from a published report of the SET Statistics.3 The 1975–80 period is interpolated (using a linear function) from these two series and a 1973 data point estimated in a one-off survey of R&D expenditure in government (Maddock, 1975). This means that we deliberately risked over-estimating expenditure, with the aim of taking a cautious approach to the estimated rate of return. Once the data series for the NHS and DH were derived, we split the NHS expenditure by a ratio of 20% of activity funding and 80% for infrastructure. The activity element of the NHS was combined with cardiovascular-specific DH data, derived by multiplying the DH figures by an estimate of cardiovascular activity. In deriving a figure for cardiovascular activity, we settled on a central estimate of 9% of total research, and we also assumed it to be constant over the time period. This estimate of 9% represents a middle ground between a number of independent data points: • the MRC’s spending on cardiovascular research (using the ‘exclusive’ definition described below) ranged from 6% to 9% • WT cardiovascular funding was more erratic, ranging between 2% and 19% of expenditure being on cardiovascular research • the proportion of peer reviewed research papers in the cardiovascular field as a percentage of all biomedical outputs ranged from 8% to 9% between 1988 and 1995 (Dawson et al., 1998) • the proportion of peer reviewed research papers in the cardiovascular field (as a percentage of all NHS research outputs) was 11.5% between 1990 and 1997 (Wellcome Trust, 2001) • in a short, one-off report, NHS R&D Annual Reporting 2002/03 4, 63 out of 764 programmes (8.2%) were classified under the heading ‘cardiovascular and heart disease’. Given the importance of this estimate of the proportion of research activity that is cardiovascular (where we have no actual breakdown), Figure 3.2 presents the data also using high and low estimates of 6% and 12% respectively, and these are used in subsequent sensitivity analyses. • Department of Health – infrastructure (DH–I). This is calculated as 80% of the total NHS R&D spend, again multiplied by 9% for estimated cardiovascular research activity. The total NHS R&D expenditure is derived from the SET statistics as described above.

Figure 3.3 presents the total public and charitable spending on CVD research, as shown in Figure 3.2 but in constant prices. This emphasises that the growth in cash terms apparent in Figure 3.1 masks the fact that in real terms spending fell from 1975 to 1981 and whilst it rose again to 1991, it had still not got back to the levels of real spending of 1975. In more detail, between 1975 and 1992, total annual cash expenditure by public and charitable funders of cardiovascular research increased three-fold from £26m to £88m. Total estimated cardiovascular expenditure over the period was £900m in cash terms. During the period growth rates for the different components varied: total MRC expenditure increased by an average of 10.5% a year, Wellcome Trust by 23.1%, Gross Expenditure on Research and Development (GERD) by 6.8% per year and Government GERD by 3.5%.

Figure 3.3: Expenditure on cardiovascular research by public and non-profit funders, 1975 to 1992, at constant 2005 prices

Real spend (£m at 2005 prices)

200 180 160 140

Figure 3.4: UK pharmaceutical industry expenditure on cardiovascular research, 1975 to 1992, at current prices 250

Research spend (£m cash)

• Funding Councils (FC). The Higher Education Funding Council for England (HEFCE) directly provided us with funding data for biomedical subjects from 1989/90 to 2007/08. The data for 1989/90 to 1992/93 were for the University Funding Council (UFC) and covered Great Britain. From 1993/94 onwards the data here were for England alone. Biomedical research was defined by the cost centres/unit of assessments used at the time.5 To estimate a time series for 1975 to 1992, we used Funding Council data for the period 1989–2007 (with adjustments to include data for Wales, Scotland and Northern Ireland where otherwise not taken into account) and then back-cast these figures using an exponential function to generate data for the period in question. These data represent the total biomedical spend during our period; they have been multiplied by 9% to give estimated cardiovascular research activity.

200 150 100 50 0

Estimates of ‘for-profit’ sector funding To provide comparative figures for the ‘for profit’ sector, we have assumed that cardiovascular research is principally supported by pharmaceutical companies in the UK. We note that the total non-pharmaceutical health care industries’ expenditure is equivalent to around 4% of that of the pharmaceutical industry. Although we discuss the derivation of this figure in more depth below, we have not included it in our analysis because it cannot be broken down by therapeutic area. Our detailed sources and assumptions for this sector were as follows: • The pharmaceutical industry. The Association of the British Pharmaceutical Industry (ABPI) compiles and publishes data on total R&D expenditure in the UK by the pharmaceutical industry, but does not disaggregate them between different therapeutic target areas. We were unable to find any such disaggregated data for the UK elsewhere. Estimates of worldwide R&D expenditure by broad therapeutic area were provided for the last three years, 2003–05, by CMR International from their survey of approximately 20

multinational pharmaceutical companies.6 These showed cardiovascular R&D accounting for 10% of total worldwide private pharmaceutical R&D expenditure over those three years. Ward and Dranove (1995) present similar estimates, for the years 1966 and 1988, for the therapeutic categories of R&D spend by members of the Pharmaceutical Manufacturers of America association, covering global spend by US companies and US spend by non-US companies. Ward and Dranove quote 14% of pharmaceutical R&D going to cardiovascular in 1966 and 30% in 1988. Thus, we assume overall that cardiovascular may account for anything between 10% and 30% of total pharmaceutical sector private R&D spending. On the basis of the estimates provided above, however, we conclude that an estimate of 15% is more likely to be representative of industry cardiovascular research expenditure than a strict 20% mid-point, and this is the figure we have used in our calculations of the ‘spillover’ effects described in Chapter Six.

Figure 3.5: UK pharmaceutical industry expenditure on cardiovascular research, 1975 to 1992, at 2005 prices

1 High 2 Best estimate 3 Low

500 400 300

• Private, non-pharmaceutical health care sector expenditure. Data on R&D spending by the private non-pharmaceutical health care sector are even scarcer than for the pharmaceutical sector. We were not able to find specific data on UK R&D expenditure. The only available data, which are published in the Department for Innovation, Universities and Skills (DIUS) R&D Scorecard,7 give total annual R&D expenditure figures by the health sector (excluding pharmaceuticals) for the period 1990–2006, on the basis of worldwide spend by UK-based companies plus spend in the UK by non-UK-based companies. The same source presents equivalent data for the pharmaceutical sector which makes clear that health care industry R&D is very small relative to pharmaceutical R&D, being only around 4% of the size of pharmaceutical R&D. Given its small scale and the unavailability of any estimates as to the therapeutic area breakdown of that spending, we have treated non-pharmaceutical health sector private R&D as de minimis for the current exercise.

Conclusions

Figures 3.4 and 3.5 summarise the data. Figure 3.4 shows that our best estimate of private pharmaceutical industry cardiovascular R&D expenditure increased steadily in cash terms from £12m in 1975 to £213m in 1992. When expressed in constant 2005 price terms (Figure 3.5) this becomes an increase in our best estimate from £65m in 1975 to £295m in 1992.

4 www.dh.gov.uk/en/Researchanddevelopment/ Researchanddevelopmentpublications/Informationfromresearch/ DH_4078384 [accessed 12 February 2008].

Estimates of spending specifically on cardiovascular research are not readily available for either sector and we provide here original estimates that have not previously been available. The construction of these estimates relies on data from a number of different sources and various assumptions regarding the split of medical research between different clinical areas. Despite some inevitable uncertainties we have a time series for total public/charitable research expenditure on cardiovascular disease which represents the research investment that we are studying, and also a series for research expenditure by the pharmaceutical industry. This feeds into our estimates of the indirect impact of public research on GDP. The private figures also make it clear that private sector R&D expenditure greatly exceeds public plus charitable expenditure. 1  See: www.amrc.org.uk. 2  www.dti.gov.uk/dius/science/science-funding/set-stats/. 3  Annual Review of Government Funded R&D (Cabinet Office, 1984).

5 These were: Clinical Medicine, Clinical Dentistry, Pre-Clinical Studies, Anatomy and Physiology, Pharmacology, Pharmacy, Nursing, Other Studies Allied to Medicine, Biochemistry, Psychology, Other Biological Sciences for the UFC; and Clinical Laboratory Sciences, Community-based Clinical Subjects, Hospital-based Clinical Subjects, Clinical Dentistry, Pre-Clinical Studies, Anatomy, Physiology, Pharmacology, Pharmacy, Nursing, Other Studies and Professions Allied to Medicine, Biochemistry (Discontinued in RAE2001), Psychology, Biological Sciences for HEFCE. 6 M Ogg, personal communication, 28 November 2007. 7 www.innovation.gov.uk/rd_scoreboard/ [accessed 23 April 2008]. This area was previously under the purview of the Department for Trade and Industry (DTI).

Chapter Four

Analysis of cardiovascular guidelines

KEY POINTS • In this Chapter we look for evidence to suggest what proportion of the impact of all (world) cardiovascular research might reasonably be attributed to UK research. We also estimate a mean time lag between research expenditure and its impact on health. • Analysis of UK cardiovascular clinical guidelines indicates that the proportion of the cited evidence base attributable to UK cardiovascular research is 17%, ranging between 12% and 23% for different guidelines. • Combining this evidence with findings from previous studies and other data points, we assume that the proportion of UK health care benefit attributable to UK research lies in the range of 10% to 25% with a mid-point estimate of 17%. • Analysis of cardiovascular clinical guidelines suggests that the time lag between research spending and citation on a clinical guideline (a proxy for health gain) is around 12.5 years, ranging between nine and 17 years. • Combining this with findings from previous studies and other data points, we assume a mean lag between research and impact of between ten and 25 years, with a mid-point of 17 years.

Introduction As introduced in Chapter Two, to estimate the returns from cardiovascular research we need to know what proportion of health gained can be attributed to UK research and the time lag between when R&D spending takes place and the time that society obtains the health benefits.

At one level, both pieces of information can be generated heuristically. For example, the UK accounts for roughly 10% of the global research (King, 2004), and the time lag between research and impact is often quoted as in the order of ten to 20 years. However, to both validate and improve on these estimates, and to allow for the possibility that these parameters might vary between clinical areas, we undertook an analysis of cardiovascular guidelines based on the approach developed by Grant et al. (2000) and discussed in Chapter Two.

Analysis of UK clinical guidelines relating to cardiovascular disease We identified seven national clinical guidelines which provide a broad representation of current cardiovascular practice (see Table 4.1 below), focusing on UK guidelines only since we are concerned with UK practice. Five of the seven were published by the National Institute for Health and Clinical Excellence (NICE), one by the Royal College of Physicians (RCP) and the other by the Scottish Intercollegiate Guidelines Network (SIGN). The references were extracted from the electronic (PDF) version of each guideline using a bespoke computer program and then matched with the bibliometric database developed and maintained by the Centre for Science and Technology Studies (CWTS) at Leiden University in the Netherlands.1 Of the 2,881 references on the seven guidelines, 2,793 (97%) were extracted by the program, and 2,064 (74% of the extracted papers) were indexed on the CWTS database and were used as the basis of our analysis (Table 4.2). Those references that were not on the database will include non-serial outputs (such as books), journals that are not captured by CWTS in their source databases, or incorrect (and therefore un-matchable) references. In the sections that follow, on country of authorship and knowledge cycle time, our analysis looks only at the numbers of cited papers, and makes no judgements about their relative importance in contributing to the guidelines in question.

Table 4.1: Guidelines analysed Guideline

Publication date

www.sign.ac.uk/pdf/sign96.pdf

Atrial fibrillation

www.nice.org.uk/nicemedia/pdf/cg036fullguideline.pdf

Chronic heart failure

www.nice.org.uk/nicemedia/pdf/Full_HF_Guideline.pdf

Hypertension

www.nice.org.uk/nicemedia/pdf/HypertensionGuide.pdf

Post-myocardial infarction

www.nice.org.uk/nicemedia/pdf/CG48FullGuideline.pdf

Venous thromboembolism

www.nice.org.uk/nicemedia/pdf/VTEFullGuide.pdf

www.rcplondon.ac.uk/pubs/books/stroke/stroke_guidelines_2ed.pdf

Figure 4.1 shows, for the seven clinical guidelines, the distribution of the addresses of the authors cited for G7 countries.2 Most papers were published by authors living in either the USA (35%) or the UK (17%; ranging from 12% for the atrial fibrillation guidelines to 23% for the stroke guideline). The G7 countries accounted for 71% of authors. By comparison, Grant et al. (2000) reported that 36% of papers had an author with a US and 25% an author with a UK address. Both sets of figures suggest a higher than expected level of citation for papers of UK origin – which may reflect either perceived greater relevance of UK evidence to UK context, or simply parochial preference. To get a better understanding of what lay behind these findings, a more qualitative analysis was undertaken of the research cited on one of the guidelines, that on hypertension.3 This detailed analysis highlights that there are varying degrees of UK involvement in the research. A few studies are clearly fully funded by traditional UK funders and conducted solely in the UK. A greater number, however, are international with varying degrees of UK input – and often funded by the pharmaceutical industry. Taken together this analysis suggests that UK-based research probably constitutes well over 10% of the research on key trials cited on the guideline. Discrepancies between the figures derived from the two analyses to some extent reflect problems of attribution inherent in bibliometric analysis of authorship – particularly that addresses of all authors may not be listed. Thus, in the quantitative analysis, both the UK and the USA would be ‘credited’ equally for a paper with six authors, five of whom were at a UK institution and only one of whom was in the USA, although the UK contribution would arguably have been much greater. Combining these various estimates, we assume that 17% of the health gain to the UK population might reasonably be attributed to UK cardiovascular research and this proportion might range between 10% and 25%. Knowledge cycle time The age of the papers cited on a clinical guideline is known as the ‘knowledge cycle time’, and this is illustrated in Figure 4.2. The mean knowledge cycle time for all seven guidelines is around 9.5 years, ranging from just over six years (for the chronic heart failure guideline) to just under 14 years (for venous thromboembolism). A third of references were over ten years old, and 2.5% over 25 years old. It is notable that these figures are similar to those reported by Grant et al. (2000), where the median age was eight years, with 25% over ten years and 4% more than 25 years old. Given a plausible approximate time lag between research expenditure and publication of around three years, we can assume a total time lag of around 12.5 years, ranging between nine and 17 years.

Table 4.2: Number of references, papers and found papers, by guideline Guideline

Number of references

No (%) of extracted papers

No (%) of matched papers

2,793 (97%)

2,064 (74%)

Figure 4.1: Nationality of papers cited in seven cardiovascular clinical guidelines published between 2003 and 2007 Percentage of papers cited on guideline

Country of authorship

Atrial fibrilation

Myocardial infarct

Pulmonary embolism

To complete our understanding of the time lags involved we also looked at the literature on product development – complicated both because it takes time to obtain a tangible research output (e.g. a new medicine) and because maximum uptake of new products is far from immediate. A highly cited estimate of drug development costs and timelines is DiMasi et al. (2003). In this work, the authors look at development costs and timelines for a sample of new drugs first tested in humans between 1983 and 1994. They estimate that the time between the start of clinical testing and submission of a new drug application (NDA) or a biological licence application (BLA)4 with the Food and Drug Administration (FDA) is 72.1 months. On top of this, they estimate that the approval phase lasts on average 18.2 months. Combining these two results, the time from the start of clinical testing to marketing approval in their timeline for a representative drug averages 90.3 months. In an earlier paper (DiMasi et al., 1991), where the authors look

It is important to remember that time lag estimates generated in this way represent proxies for the real values since other factors are often involved. For example, there may be a gap between guideline publication and translation into practice, suggesting that we may be underestimating the time lags. On the other hand, interventions may come into use before guidelines outlining them have been published – suggesting that we may overestimate time lags in some cases. We also recognise that time lags will vary considerably between different types of research. Basic research will precede the health impact that might follow from technologies it underpins by a much longer period than, for example, a health technology assessment confirming in which patients a new technology might be most costeffective. Equally, it may be that research uptake is better in some fields than others. In the absence of better evidence, however, and combining the various estimates we presented above, we assume the lag between when public and charitable R&D spending on cardiovascular research takes place, and the time that it benefits society through measurable health gains, to be 17 years, ranging from ten to 25 years. Aware of the importance of this evidence to our overall estimate of returns, and the paucity of data that it is derived from, in Chapter Eight we undertake an analysis of the sensitivity of time lags on our final results.

Table 4.3: Estimated time lags for pharmaceutical product development Approval phase Time between the start of clinical testing and submission of an NDA/BLA

Time from start of clinical testing to marketing authorisation

DiMasi et al. (2003)

72.1 months

18.2 months

90.3 months

DiMasi et al. (1991)

68.6 months

98.9 months

Figure 4.2: Age of papers cited on seven cardiovascular guidelines published between 2003 and 2007 200 180 160 140

at a sample of new medicines first tested in humans between 1970 and 1982, they estimate that the time from start of clinical testing to marketing authorisation was 98.9 months. Within this earlier time period, the approval phase was estimated to be much longer (at 30.3 months) than in the later period.

120 100 80 60 40 20 0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

Age of cited papers

Figure 4.3: Age of papers cited by review papers on seven cardiovascular guidelines published between 2003 and 2007 700

Analysis of cited reviews

We initially planned to undertake an analysis of the papers cited by the papers cited on clinical guidelines to understand the ‘generational’ effects of the flow of knowledge (as initially done by Grant et al., 2000). However, we began to question the additional utility of this analysis on adding to the estimates derived above, so instead focused our analysis on cited reviews. Within the CWTS all publications are classified as being, for example, original papers, reviews, letters, editorial etc. To our surprise, of the 2,064 references only 122 (or 6%) were reviews, hence we question the value of this additional analysis for the information we are trying to generate. The 122 reviews cited a further 35,774 papers of which 12% were UK (range: 5% for artrial fibrillation to 18% for stroke), and 45% US. An age distribution for these cited papers is presented in Figure 4.3. The mean knowledge cycle time for reviews from these guidelines was found to be around 13 years, varying from just under nine years for those reviews on the hypertension guideline (an outlier), to 14 years for chronic heart failure. Given the approximate time lag between research expenditure and publication of three years stated above, we conclude that total lag time for the review was around 16 years,

400 300 200 100 0

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

Conclusions Although the evidence is limited and our indicators are proxies for the parameters in question, we conclude that for the analysis in this report: • the proportion of the cited evidence base attributable to UK cardiovascular research will be assumed to be 17%, ranging between 12% and 23% • mean lag between research and impact of between ten and 25 years, with a mid-point of 17 years. Postscript: A note on citation impact Although not of direct relevance to the current study, it was notable that the papers cited by the cardiovascular clinical guidelines were themselves extremely highly cited. CWTS considers papers with 20% more citations than the field average as of high relative impact. The group of papers cited by the guidelines had on nine times more citations than would have been expected in their field. As citation patterns differ by field, and subfield, it is important to normalise any citation measure by the mean field citation score (FCSm), i.e. the average number of citations to each paper in the field. In detail, if CPP is number of citations per paper then the ratio of the two measures – CPP/FCSm – gives a relative estimate of impact. If the ratio is 1.0 then the number of citations per paper is equivalent to the expected number for a given field; a figure above one implies a higher relative impact and a figure below one a lower impact. As illustrated in Figure 4.4, the average ratio for papers cited on cardiovascular guidelines was 9.1, ranging from 3.1 (for venous thromboembolism) to 22.0 (for hypertension).5

Figure 4.4: Citation impact compared to world field average 25

ranging between 12 and 17 years. This gives a best estimate similar to the knowledge cycle time for papers cited directly suggesting that citation through a review does not significantly change the speed of uptake of research findings.

ial At r ion a ll t br i

l ic on ous dia nsi ron car tion Ve n l i s m C h ilu r e p e r t e o y c o m far fa y b t t H r m s in Po hea boe m o t hr

Guideline title

1 The Centre for Science and Technology Studies (CWTS) maintains a bibliometric database of all scientific publications (including health and biomedical research) for the period 1981 to 2004. This dataset is based on the journals and serials processed from the CD-ROM versions of the Science Citation Index and associated citation indices (CI): the Science Citation Index (SCI), the Social Science Citation Index (SSCI), and the Arts & Humanities Citation Index (A&HCI), extended with six so-called speciality Citation Indices (Chemistry, Compumath, Materials Science, Biotechnology, Biochemistry & Biophysics, and Neuroscience). Currently, CWTS is changing its database towards the Web of Science version (the internet version) of the Citation Index(es), which covers the period 1981 to 2005, and has a somewhat different journal set coverage. The construction of this database, and the indicators operating on it, are described in various scientific publications (Moed et al., 1995; van Leeuwen et al., 2001; van Leeuwen et al., 2003). 2  Because of the way the databases are constructed authors are not linked directly to addresses. Two separate lists are presented: the authors’ names and a list of addresses; no direct link is made between the two. The countries of authorship are taken as the addresses listed for each paper. 3  Full findings are reported in the Annex to Chapter Four. 4  Once drug developers believe that they have enough evidence of safety and efficacy, they will compile the results of their testing in an application to regulatory authorities for marketing approval. In the USA, manufacturers submit a new drug application (NDA) or a biological licence application (BLA) to the FDA for review and approval. 5  The field definitions used by CWTS are the same as those used by Thompson ISI.

Chapter Five

The value of the health gains in cardiovascular disease and their associated health care costs KEY POINTS • We use a bottom-up approach to estimate the vale of the health gains from specific interventions to treat or prevent cardiovascular disease over the period 1985–2005, and the health care costs incurred in the achievement of these gains. • From published economic evaluation studies, and estimates of usage, we calculate the quality adjusted life years (QALYs) gained from a series of 46 combinations of patient groups and specific interventions. • These QALYs are valued at £25,000 – the mid-point of the ‘threshold range’ used by the National Institute for Health and Clinical Excellence (NICE). This value represents a measure of the opportunity cost if, rather than investing in R&D, the resources had been used directly in the NHS. • Our best, though conservative, estimate of the value of the QALYs gained from the specific interventions included in our analysis over the whole period 1985–2005 is £69 billion. The upper and lower estimates are £91bn and £55bn respectively. • Our best estimate of the total incremental health care costs relating to those gains over the same period is £16bn, with upper and lower estimates of £17bn and £11bn respectively.

Introduction In this Chapter we estimate the monetary value of health gains from specific interventions to treat or prevent cardiovascular disease (CVD) over the period 1985–2005, and we also estimate the health care costs incurred in the achievement of these gains. As discussed in Chapter Two, previous attempts to value health gains from investment in medical research have tended to use a ‘top-down’ approach, monetising the health gains from reduced morbidity and mortality at the macro level and then making assumptions about the proportion of these that can be attributed to medical research. In this Chapter we use a ‘bottom-up’ approach that considers the effects of specific interventions and then sums these to generate an overall estimate. In doing this we use quality adjusted life years (QALYs) as the measure of health gain. Although QALYs have their critics and do not necessarily capture all aspects of the health benefit offered by new interventions in all clinical areas, they are widely used to measure health gain and are the measure favoured by NICE in its appraisals of health technologies. As a result there is a substantial evidence base estimating the QALYs

gained and costs associated with treatment of various diseases, not least for CVD. To estimate the monetary value of health gains from specific interventions to treat or prevent cardiovascular disease we: • review the economic evaluation literature to obtain published figures for the QALYs gained from specific interventions for cardiovascular disease over the period 1985–2005 • multiply these figures by estimates of the numbers of users of each intervention, to give an estimate of the total QALYs gained from each intervention • monetise the total QALYs gained by multiplying these estimates by published figures on the opportunity cost of a QALY within the current NHS budget • from the review of the economic evaluation literature obtain estimates of the incremental health care costs associated with each intervention and multiply these by the numbers of users to quantify the incremental health care costs of each intervention. Based on previous studies, we include interventions that are likely to have been important in terms of the health gains they have produced over the period 1985–2005. In Chapter Seven we deduct the total incremental health care costs estimated in this Chapter from the monetised total QALYs gained. We then combine the result with estimates of the costs of UK medical research, and we produce an estimate of the returns to UK medical research accounting for both the time lag between the research investment and the net returns and for the proportion of health benefits that can be ascribed to UK medical research. As noted in Chapter One, our estimates focus on the returns to UK public and charitable investment in research. The costs of private sector R&D investments are accounted for in our analysis because they are included in the incremental health care costs. These include the prices of interventions manufactured in the private sector, which include a return on its investment in research.

Overview of methods Our analysis is derived from, but ultimately different to, the approach used in the IMPACT Study (Capewell et al., 2007), which inter alia tries to explain the decline in coronary heart disease mortality in England and Wales between 1980 and 2000. We extend the analysis by considering a wider range of interventions for cardiovascular disease, by focusing on the UK, by measuring costs and benefits during the period 1985–2005, by quantifying the cumulative benefits and costs over the whole period rather than the difference between the two end points, and by calculating the QALYs gained rather than the deaths prevented or postponed and life years gained.

To estimate the monetised health gains from specific interventions for CVD over the period 1985–2005 we: 1 identify specific interventions for patients with CVD 2 estimate the QALYs gained for each intervention in each patient group 3 estimate the numbers of patients in each patient group 4 adjust for overlapping patient groups 5 estimate the uptake of each intervention by each patient group in each year 6 compute the numbers of users of each intervention in each patient group 7 compute the numbers of new users each year 8 adjust for compliance with treatment 9 adjust for polytreatment 10 compute the total QALYs gained from each intervention 11 monetise the total QALYs gained 12 undertake a sensitivity analysis. From our review of the economic evaluation literature we also obtain estimates of the incremental health care costs associated with each intervention and multiply these by the numbers of new users who are compliant with treatment to quantify the health care costs associated with each intervention.

Estimating the monetised health gains from specific interventions for CVD over the period 1985–2005 Stage 1: Identify specific interventions for patients with CVD At the first stage we identified specific interventions for patients with CVD to be included in the analysis. The patient groups and treatments were selected according to whether or not they were likely to have been important in terms of the total health gains they have produced in the UK since 1980. At the outset we used patient groups and interventions that were used in the IMPACT study (see Box 5.1), which focused on interventions that were important in terms of the total health gains they have produced since 1980. We excluded from our analysis a number of CVD risk factors that were included in the IMPACT study (obesity, physical activity, diabetes) because the prevalence of these has worsened over time and so the health gains in terms of the QALYs gained from research into these diseases is deemed to have been zero. This is possibly conservative because the increasing prevalence of these conditions may have been even higher without UK medical research. We also did not include deprivation, taking the view that the level of deprivation is unlikely to have been affected by UK medical research.

Box 5.1: The IMPACT study The IMPACT coronary heart disease (CHD) mortality model (Capewell et al., 2007) combines data from various sources on patient numbers, treatment uptake, treatment effectiveness and risk factor trends to estimate the numbers of CHD deaths prevented or postponed over a pre-defined time period due to specific interventions and modification of CHD risk factors. It has been used to estimate the proportion of the decline in CHD mortality that might be attributed to treatments or risk factor changes and has been applied to data from Europe, New Zealand and China (Capewell et al., 1999, 2004; Unal et al., 2004, 2005; Laatikainen et al. 2005). The model includes a range of patient groups and interventions including: treatment of acute MI (cardiopulmonary resuscitation, thrombolysis, aspirin, primary angioplasty); secondary prevention of CHD post-MI (aspirin, beta blockers, ACE inhibitors, statins, warfarin, rehabilitation); revascularisation (CABG surgery, angioplasty); treatment of unstable angina (aspirin, glycoprotein IIB/IIIA antagonists); treatment of chronic stable angina (aspirin, statins); and treatment of heart failure (ACE inhibitors, beta blockers, diuretics, aspirin, statins). It also includes treatment of hypertension and hypercholesterolaemia for primary prevention of CHD and the modification of a number of CHD risk factors (smoking, deprivation, obesity, diabetes, physical activity). The model has been extended to include stroke (Zhechun et al., 2006), with the following patient groups and interventions: treatment of acute stroke (aspirin, anticoagulants, tr-PA, stroke unit, early diagnosis and treatment); secondary prevention of stroke (rehabilitation, aspirin, statins, warfarin, antihypertensive drugs). It also includes use of warfarin and aspirin plus treatment of hypertension and hypercholesterolaemia for primary prevention of stroke, and population diabetes control.

We added interventions for treatment of arrhythmia (implantable cardioverter defibrillators [ICDs]), treatment of heart failure (cardiac resynchronisation therapy [CRT] devices [CRT-P], CRT devices plus ICDs [CRT-D]) and heart transplants. We also added clopidogrel, which is used to treat CVD. The interventions we added were not included in the IMPACT study but are likely to have generated health gains over the period 1985–2005. The patient groups and interventions considered are in Table 5.1 column 1. Note that some treatments are considered more than once because they are used in different patient groups.

We have not included all CVD interventions provided by the NHS in the UK – it would be impractical to do so. We assume that the monetised total QALYs gained minus the incremental health care costs of the interventions not included directly in our analysis is equal to zero. The effect is that these interventions are assumed to be neutral in terms of their impact on the calculation of the net returns to research. This effectively assumes that the net benefits to society from the provision of these interventions (value of health gains minus costs) are zero. Stage 2: Estimate the QALYs gained for each intervention in each patient group We undertook a systematic review of the economic evaluation literature to obtain figures for the QALYs gained for each intervention in each patient group. We systematically searched the economic evaluation literature using three databases: • published reports from the NIHR Health Technology Assessment Programme (www.ncchta.org/project/ htapubs.asp) • the NHS Economic Evaluation Database (NHS EED) at the Centre for Reviews and Dissemination at the University of York (www.york.ac.uk/inst/crd/ crddatabases.htm) • PubMed (www.ncbi.nlm.nih.gov/sites/ entrez?db=pubmed). We also searched the reference sections of retrieved papers from these databases to identify additional papers. We selected studies according to whether or not they focused on the specific intervention in the particular patient group, and whether or not they focused on UK patients. If we were unable to find studies that focused on interventions in the particular patient group then we considered studies that investigated the same intervention in other similar patient groups. If we were unable to find studies that focused on UK patients then we considered studies that used non-UK patients. We preferred studies that focused on interventions in the particular patient group over UK studies. If we found more than one study that focused on the specific intervention in the particular patient group and that used UK patients then we used the most recently published study. Where a range of baseline estimates of QALY gains were reported, we took a conservative view and always used the lowest value. The estimates of the QALYs gained from each intervention in each patient group are in Table 5.1 column 3. See Annex 5A for further details of the studies used to generate these estimates. Because we focus on new users, all the studies would ideally consider a lifetime time horizon. Not all did, and the studies varied in terms of the time horizons they considered and the discount rates they used.

Stage 3: Estimate the numbers of patients in each patient group To calculate the total QALYs gained we multiplied the QALYs gained for each specific intervention by the numbers of new users of that intervention in each year. For some interventions we have data on the numbers of new users directly, but for the majority of interventions we do not have this information and we model it as a function of the numbers of patients in each patient group and the uptake rate for each specific intervention. We therefore require estimates of the numbers of patients in each patient group in each year, which we call the numbers of eligible patients. These data were obtained from systematic searches of the literature. In some cases we were able to find data on numbers of eligible patients for England and we applied these figures to UK population data. In the majority of cases we were unable to find figures for the entire period 1985–2005 in which case we modelled the numbers of eligible patients in other years: we extrapolated the percentage of the UK population in each patient group to UK population figures in these time periods with missing data. See Annex 5B for further details on how these estimates were generated. Stage 4: Adjust for overlapping patient groups The strategy described above for quantifying the numbers of eligible patients is likely to overestimate the total because some patients will feature in more than one group (e.g. some patients with chronic stable angina will also have a revascularisation procedure). We adjusted for this overlap by reducing the numbers of eligible patients in some patient groups according to the criteria used in the IMPACT study. See Annex 5C for details on these criteria. Stage 5: Estimate the uptake of each intervention by each patient group in each year We obtained figures for the uptake of specific interventions in each patient group based on systematic searches of the literature. As with the data for eligible patients, we were unable to find complete data for every intervention. In all cases we were able to identify when the intervention was first used and so we knew the time period before which the uptake rate was zero. We were also able to identify non-zero figures for one or more later time points. In the absence of additional data we generally linearly interpolated uptake rates where there was a gap between two time points and we assumed a constant uptake rate at later time points. Where it was used the latter is probably a conservative assumption in that generally uptake increases over time and so it will underestimate the numbers of users of each intervention. See Annex 5D for further details on how these estimates were generated.

Table 5.1: Summary of results by patient group/intervention, 1985–2005 1

QALYs gained Patient groups/interventions

Compliant new users (000s)

Per new user

Treatment of acute MI:

Total (000s)

Incremental costs Total monetised (£ million) Per new user (£)

Total (£ million) 58.7

Community resuscitation

Hospital resuscitation

Thrombolysis

Clopidogrel

Primary angioplasty

Secondary prevention of CHD post-MI: Aspirin Clopidogrel Beta blockers ACE inhibitors Statins Warfarin

2.7 3,178.2

-501.7 67.6

CABG surgery

Angioplasty

Rehabilitation Revascularisation:

Treatment of unstable angina: Aspirin

Glycoprotein IIB/IIIA antagonists

Treatment of chronic stable angina:

Statins Treatment of arrhythmia: ICD Treatment of heart failure:

1,675.8 -283.3

1,363.5 68,805

1363.5 375.2

ACE inhibitors

Beta blockers

Heart transplant: Heart transplant

Treatment of acute stroke:

127.9 -201.8

Anticoagulants

Stroke unit

Early diagnosis and treatment

Antihypertensive drugs

Secondary prevention of stroke: Rehabilitation therapy Aspirin

Primary prevention of CVD: Treatment of hypertension Treatment of hypercholesterolaemia Primary prevention of stroke: Warfarin

Smoking cessation: Quitting smoking Total

265.4 21,780.2

57.5 -42.7 248.5 55

248.5 15,627.4

In column 1 the patients groups are highlighted in bold. CHD = coronary heart disease; MI = myocardial infarction; ICD = implantable cardioverter defibrillator; CRT-P = cardiac resynchronisation therapy device; CRT-D = CRT device plus ICD; CVD = cardiovascular disease.

Stage 6: Compute the numbers of users of each intervention in each patient group We computed the numbers of users of each specific intervention in each patient group by multiplying the numbers of non-overlapping eligible patients in each time period by the uptake rate for each intervention in that time period. Stage 7: Compute the numbers of new users each year The estimates of the QALYs gained for each intervention obtained from the systematic review usually adopt a time horizon that is greater than one year. Since the numbers of users of a given intervention in each year will include a proportion who started using the intervention in a previous year, to avoid double counting we focus only on the numbers of new users. We computed the numbers of new users each year by subtracting from the numbers of users in each year the numbers of users in the previous year, also accounting for the numbers of deaths from all causes among users in the previous year. The numbers of deaths were obtained from life tables produced by the Government Actuary’s Department (www.gad.gov.uk/Demography_Data/Life_ Tables/). For simplicity, we assumed a constant annual mortality rate of 1%, which is the rate for males aged 60 years and females aged 65 years in the UK. Males and females older than these ages have a higher mortality rate and so the numbers of deaths, and therefore the numbers of new users, will be underestimated in these groups. In addition, these annual mortality rates are for the general population; among patients with CHD the annual mortality rate is likely to be higher. Hence, our mortality rate is probably conservative, which means that we underestimate the numbers of new users each year. Stage 8: Adjust for compliance with treatment Not all new users will benefit from treatment because a proportion will be non-compliant. We searched the economic evaluation studies used to obtain the QALYs gained estimates for each specific intervention. In some of these non-compliance was accounted for in the analysis. Where it was not accounted for we applied compliance rates used in the IMPACT study. See Annex 5E for details on the values used. Accounting for compliance gives an estimate of the numbers of compliant new users of each specific intervention in each year. These are reported for each specific intervention in Table 5.1 column 2. Stage 9: Adjust for polytreatment A polytreatment adjustment is considered because patients in a particular patient group may receive more than one intervention. Ideally we would have data on the numbers of patients receiving every combination of interventions, but invariably we do not have these data. Hence, we need to account for the use of multiple interventions in some other way.

We adjusted for polytreatment in three ways: 1

we assumed there was no polytreatment

we assumed there was maximum polytreatment and that the QALYs gained from each specific intervention are additive

we assumed there was maximum polytreatment and that the QALYs gained from each specific intervention are not at all additive.

Options 1 and 2 yield the same results and are used to derive our best estimates of the total QALYs gained. Option 3 is used in the sensitivity analysis (see below) to generate a lower estimate. (See Annex 5F for a more detailed description of the methods used.) Stage 10: Compute the total QALYs gained from each intervention We compute the total QALYs gained from each intervention in each year by multiplying the numbers of new users in that year by the QALYs gained from each intervention. We then summed the result across the period 1985–2005 to calculate the total QALYs gained. These are reported in Table 5.1 column 4. Our best estimate of the total QALYs gained over the whole period 1985–2005 is 2,755,223. Stage 11: Monetise the total QALYs gained We monetised the total QALYs gained by multiplying the estimated total QALYs gained by published figures for society’s willingness/ability to pay for a QALY within the current NHS budget. According to the Guide to the Methods of Technology Appraisal published by NICE: Below a most plausible ICER [incremental costeffectiveness ratio] of £20,000/QALY, judgements about the acceptability of a technology as an effective use of NHS resources are based primarily on the cost-effectiveness estimate. Above a most plausible ICER of £20,000/QALY, judgements about the acceptability of the technology as an effective use of NHS resources are more likely to make more explicit reference to factors including: • the degree of uncertainty surrounding the calculation of ICERs • the innovative nature of the technology • the particular features of the condition and population receiving the technology • where appropriate, the wider societal costs and benefits. Above an ICER of £30,000/QALY, the case for supporting the technology on these factors has to be increasingly strong. (NICE, 2004, p. 33)

Figure 5.1 plots the estimated total QALYs gained for each year. See Annex 5G for the estimates used to generate the figure. The trend is broadly upwards from left to right. The variation in the trend from year to year is due to the numbers who quit smoking, which according to our calculations are very erratic, and also due to the conservative assumptions we have made regarding the numbers of new users each year (in particular the assumption that uptake rates are constant where no data for these rates is available). The figures for smoking cessation are based fundamentally on the prevalence rate of smoking measured in the General Household Survey. The trend in the prevalence rate is generally downward, but in some years it increased and in others it remained constant. In these years we set the numbers of quitters to zero. This is probably conservative because even in years when the prevalence increased there may have been people who quit smoking due to medical research, but there were fewer of these than the number of people who started smoking (who did not start due to medical research).

Figure 5.1: Total QALYs gained by year 250

QALYs gained (000s)

This evidence indicates that NICE is prepared to recommend interventions that cost up to £20,000–£30,000 per QALY gained, and we therefore interpret these figures as estimates of the willingness/ ability of the NHS to pay for a QALY given the current NHS budget. They therefore reflect the opportunity cost of a QALY if the investment in research were instead to be directly invested in health services. We use the middle of the range, £25,000, as our central estimate. Hence, we multiply the total QALYs gained by £25,000 to generate a monetised value of the total QALYs gained. The results are reported in Table 5.1 column 5. The total monetised QALYs gained over the whole period 1985–2005 are estimated to be £68.9bn.

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

• Lower estimate: we calculated the lower estimate of the total QALYs gained by assuming there was maximum polytreatment and that the QALYs gained from each specific intervention are not at all additive (see Annex 5F for a more detailed description of the method). As noted above, our best estimate of the total QALYs gained over the period 1985–2005 is estimated to be 2,755,223 QALYs and the total monetised QALYs gained are £68,880.6m. Using the approaches described above the upper estimates over the whole period are 3,648,331 QALYs and £91,208.3m and the lower estimates are 2,195,732 QALYs and £54,893.3m (see Annex 5J for further details).

Stage 12: Undertake a sensitivity analysis At the final stage we undertook a sensitivity analysis to generate upper and lower estimates of the total QALYs gained around our best estimate. These were calculated as follows: • Upper estimate: smoking cessation is an important component of the total QALYs gained. We assumed in our best estimate that 22–25% of people who quit smoking did so as a result of medical research, based on the numbers of quitters who were told by medical services to quit smoking for medical reasons. This may underestimate the contribution of medical research because some people may quit smoking for medical reasons without being told to do so by the medical services. Hence, to generate an upper estimate of the total QALYs gained, we assumed that 100% of people who quit smoking did so as a result of medical research.

From our review of the economic evaluation literature, we also obtained estimates of the incremental costs associated with each intervention. Where UK papers were used to generate the QALY gains estimates we used the same papers to obtain the incremental cost estimates. Where non-UK studies were used to generate the QALYs gained we attempted to find cost estimates from UK cost studies or UK economic evaluations with outcomes other than QALYs. If we were unable to find any UK studies then we used the non-UK study that was used to generate the QALYs gained estimates and converted the costs to UK£. Where a range of baseline incremental cost and QALYs gained estimates were reported we used the cost value that was commensurate with the QALYs gained estimate used in our analysis. All costs were converted to 2005 UK£ using NHS pay and prices indices (Curtis, 2007) and where appropriate GDP purchasing power parities published by the OECD (OECD, 2008). The estimates of the incremental cost of each intervention in each patient group are in Table 5.1 column 6. Note that in some cases the incremental costs are negative because the intervention is cheaper than the alternative. See Annex 5H for further details on the studies used to generate these estimates. We then multiplied the cost estimates by the numbers of new users to quantify the health care costs associated with each intervention. We used estimates of the numbers of users who are compliant with treatment on the basis that non-compliers are unlikely to continue to incur treatment costs over a prolonged period. We computed the total incremental costs of each intervention in each year by multiplying the numbers of new users in that year by the incremental costs of each intervention. We adjusted for polytreatment using the same methods as for the total QALYs gained. We then summed the result across the period 1985–2005 to calculate the total incremental costs. These are reported in Table 5.1 column 7. Note that in some cases there are cost savings because the incremental costs per new user are negative. The total incremental health care costs over the whole period 1985–2005 are estimated to be £15.6bn. Figure 5.2 plots the total incremental costs for each year (see Annex 5I for the estimates used to generate the figure), which generally increase over time. We also computed upper and lower estimates around our best estimate using the same approach that was used for the total QALYs gained. The upper estimate over the whole period was estimated to be £16.5bn and the lower estimate £10.8bn (see Annex 5J for further details).

Figure 5.2: Total incremental costs by year 1600

Incremental costs (£ million)

Estimating the health care costs from specific interventions for CVD

1400 1200 1000 800 600 400 200 0

Example One: statins for secondary prevention of CHD post-MI Ward et al. (2007) undertook an economic evaluation of statins for the primary and secondary prevention of coronary events in the UK. On the basis of their modelling work, and using a lifetime time horizon, they report lifetime QALY gains associated with statin therapy in secondary prevention of coronary events in the range 103–493 QALYs gained per 1,000 patients, depending on the age and gender of the patient. We used the lowest figure (which applies to males aged 85 years) to give a conservative estimate of 0.103 QALYs gained per patient with statins over their lifetime. The numbers of patients eligible for statin therapy was based on estimates of the proportion of the population who had ever had a heart attack (3.8% of males and 1.7% of females) published in the Health Survey for England in 2003 (Department of Health, 2004a). We only have estimates for a single year, and we therefore assume that the same rate applies across the whole period. This rate, which is for the English population, is applied to UK population estimates published by the Office for National Statistics (2007). Uptake rates for statins among patients with CHD for 1994–98 were taken from Key Health Statistics from General Practice 1998 (National Statistics, 2000). The population weighted average estimates of the figures for males and females were 3.6% in 1994, 6.4% in 1995, 10.8% in 1996, 17.0% in 1997 and 23.8% in 1998. The Department of Health (2007) shows that statin use was zero prior to 1990, and that from 1998 onwards it has increased exponentially. We assumed a constant rate of increase from 0% in 1989 to 3.6% in 1994. We then assumed conservatively that the annual increase from 1997 to 1998 (6.8% in absolute terms) applied to each year from 1999 to 2005. These are figures for England and Wales, which were applied to UK population estimates.

We multiplied the numbers of eligible patients by the uptake rates to estimate the numbers of users each year. These range from zero for 1985–89 and then increase up to a maximum of 1.1m in 2005. The numbers of new users each year is the difference between the number of users in that year minus the number of users in the previous year multiplied by a factor 0.99 to account for the assumed 1% mortality rate among users. The estimated total number of new users across the period 1985–2005 was 1,225,451. Ward et al. (2007) accounted for compliance in their estimates of the QALYs gained from statin therapy and therefore we do not make any additional adjustments for this; the estimated number of compliant new users of statin therapy was 1,225,451. On average, each compliant new user of statin therapy in secondary prevention of CHD achieves a QALY gain of (at least) 0.103 QALYs. Hence the total health gain from statin therapy in the secondary prevention of CHD post-MI is (at least) 126,221 QALYs. Multiplying this figure by the presumed value of a QALY (£25,000) gives a monetised health gain of £3,155.5m. Ward et al. (2007) also estimated the incremental costs per patient of statin therapy in the selected patient group to be £1,615 in 2004 UK£. As noted above, these costs include private sector returns on research investments. Converting these costs to 2005 UK£ using the NHS pay and prices index gives an incremental cost of £1,680. Multiplying this figure by the number of new users (1,225,451) gives a total incremental cost of £2,058m. In Chapter Seven, the total incremental costs are then deducted from the monetised health gain to calculate net benefits. Example Two: smoking cessation Smoking cessation accounted for a relatively large proportion of the monetised total QALYs gained, and because of the nature of the data used it has been treated differently to the other interventions included in the analysis, as explained below. Wang et al. (2008) undertook an economic evaluation of nicotine replacement therapies in smoking cessation. They adopted an NHS/Personal Social Services perspective and they used a lifetime time horizon. Based on previous studies they report QALY gains by smokers who quit at different ages. These range from 0.99 to 2.58 QALYs depending on the age at which the person quits. We used the lowest figure (which applies to patients aged 55–64 years) to give a conservative estimate of 0.99 QALYs gained per person who quits smoking. To estimate the numbers of quitters we first obtained data for 1985–2005 from the General Household Survey (Office for National Statistics, 2008) on the proportion of the population who smoke and applied this to UK population estimates (Office for National Statistics, 2007)

to generate figures for the total numbers of smokers each year. We then computed the numbers of quitters each year by subtracting the number of smokers in the previous year from the number in the current year and then from this figure subtracting the numbers of people who died from smoking each year obtained from Peto et al. (2006). Since the numbers of smokers generally declined over time we interpret this as the number of quitters each year. People quit smoking for a variety of reasons, some of which are a result of medical research and some of which are not. We wish to estimate the numbers who quit as a result of medical research. As a crude indicator of this we used data from the Health Survey for England on the proportion of smokers who had contact with medical services who told respondents to quit smoking for medical reasons (Department of Health, 2004a). These data were available for 1993–99 and 2003–05 (the range of values reported was 0.22–0.25). We applied the 1993 value to 1985–92 and the 2003 value to 2000–02. We assumed that the proportion of quitters who received advice to stop smoking was the same as the proportion who received this advice who did not quit, and therefore multiplying these proportions by the numbers of quitters provides an estimate of the numbers who quit following the receipt of advice by the medical services to stop smoking. We treat these figures as a crude approximation of the proportion who quit due to medical research. This is probably a conservative assumption because some people may quit smoking of their own accord (rather than due to contacts with medical services) due to the negative impacts of smoking on health (that are known due to medical research). We investigate the impact of this in the sensitivity analysis. We then accounted for the specific role of cardiovascular disease-related medical research on the decision to quit smoking, as opposed to medical research in other disease areas, most notably cancer. We did this by multiplying our estimates of the proportion who quit smoking due to medical research by published estimates of the proportion of deaths due to cardiovascular disease published by Peto et al. (2006). These estimates were available for the year 2000 and were estimated to be 0.27. We applied this proportion to every year of our data. This method assumes that the contribution of cardiovascular disease-related medical research on the decision to quit smoking is proportional to the numbers of smoking deaths due to cardiovascular disease. Using these assumptions, the estimated total number of people who quit smoking due to CVD medical research over the period 1985–2005 was 265,376. Each quitter is assumed to achieve a QALY gain of 0.99 QALYs. Hence the total QALYs gained from smoking cessation were 262,722. Multiplying this figure by the presumed value of society’s willingness to pay for a QALY (£25,000) gives a monetised health gain of £6,568.1m. Smoking

cessation accounts for around 10% of monetised total QALYs gained, and for the reasons outlined above, this is probably an underestimate. Wang et al. (2008) also estimated the incremental costs of various strategies to quit smoking and their success rates over one year. We used the most costly intervention per success at one year, which had a cost of £54.88 and a success rate at one year of 5.86%. Assuming these represent the actual costs of trying to quit smoking and the actual success rate then the 265,376 people who quit represent 5.86% of those who attempted to quit, giving a total number of 4,528,601 people who tried to quit over the period 1985–2005. Assuming that each of these people incurred a cost of £54.88, then the total incremental cost was £248.5m. Smoking cessation accounts for around 2% of the total incremental costs. This is probably an overestimate because we assumed the cost of quitting was equal to the cost of the most expensive intervention and in many cases it will be less than this, possibly even zero. Also, we have not included the cost savings from the reduction in the use of health services after quitting.

Challenges This bottom-up approach has a clear advantage over a top-down approach starting with overall changes in mortality and morbidity. It is more transparent and considers only gains for which there is clear evidence of their resulting from changes in the use of specific interventions or in specific risk factors. However, the major problem is that this approach is data-hungry, and the overriding difficulty in implementing it, even for cardiovascular disease, is the patchiness of relevant data. Specific problems include, but are not limited to, the following: • We were unable to find estimates of the QALYs gained for every intervention in each patient group. In some cases we were required to use non-UK studies and in others we used data on the specific intervention in other patient groups. The studies used varied in terms of the time horizons they considered and the discount rates they used. • We were unable to find data to construct complete time series of the numbers of eligible patients. We modelled the numbers of eligible patients in years with missing data by extrapolating the percentage of the UK population in each patient group in years where data were available. • We were unable to find comprehensive data on the numbers of overlapping patients between groups. We used the best estimates that were available, which were taken from the IMPACT study. • We were unable to find data on uptake rates for specific interventions in every patient group for the whole time period. We extrapolated and interpolated

estimates where we had missing data, making conservative assumptions where necessary. • We were unable to find good data on the compliance rates associated with each intervention in every patient group. We used the best estimates that were available, which were taken from the IMPACT study. • In the absence of data on the use of different combinations of treatment we adjusted for polytreatment in our best estimate by assuming zero polytreatment/maximum polytreatment with additive effects. In the sensitivity analysis we assumed maximum polytreatment with non-additive effects. • We were unable to find complete estimates of the incremental costs associated with each specific intervention in each patient group. In these cases we were forced to use non-UK studies or different UK studies from those used to obtain the QALYs gained estimates. In the light of the uncertainty surrounding the analysis we have attempted to be conservative in the estimates we have produced. This is evidenced by the following: • We assume that the monetised total QALYs gained minus the incremental health care costs of the interventions not included directly in our analysis are equal to zero. This assumes that the net benefits to society from the provision of these interventions are zero. • When selecting the incremental QALYs where a range of baseline estimates was given we always selected the lowest value. • Because we focus on new users, all the studies would ideally consider a lifetime time horizon. The studies used varied in terms of the time horizons they considered and not all took a lifetime time horizon. In these cases the QALYs gained are likely to have been underestimated. • We have probably underestimated the numbers of eligible patients due to the assumptions we made when extrapolating the estimates to account for missing data. • We have probably underestimated the uptake rates of specific interventions by assuming these are constant in later years when they are likely to have increased. • We have probably underestimated the all cause mortality rate when computing the numbers of new users each year. • Smoking cessation accounts for a relatively large proportion of the monetised total QALYs gained and we have probably underestimated the total QALYs gained from smoking and overestimated the incremental costs associated with these.

Chapter Six

R&D spillovers

KEY POINTS • In addition to health gains, UK publicly and charitably funded medical research generates additional national income. • Published literature indicates that additional public medical research leads to additional private R&D spending. Both contribute to increasing UK gross domestic product (GDP). • Relevant empirical literature is sparse but implies that the total social rate of return to public and charitable medical research is in the range 20–67%, with a best estimate of 30%. • Thus every pound spent on public/charitable medical research yields additional GDP for the UK that is equivalent to a net return of 30p per year in perpetuity (range 20–67p per year).

Conceptual framework It is extremely important to consider wider economic gains as well as health gains when assessing the economic benefits of medical research. These economic gains are additional to the health gains; they refer to the income for UK residents that is generated by public medical research investment. This is distinct from the monetary value of health (QALY) gains that were estimated in Chapter Five. Investment in medical research by one organisation, public or private, may benefit not only that organisation but also other organisations in the medical sector, in other sectors, and also in other countries; i.e. there are what the economic literature refers to as ‘spillovers’. Spillovers should not be viewed as accidental: they can be, and are, a deliberate policy objective of spending on public research. Our interest lies in spillovers generated by public R&D, as measured in Chapter Three above. Given the link between R&D and knowledge, we do not distinguish here between R&D spillovers and knowledge spillovers but refer to both taken together. For illustrative purposes, we take a simplified view of the inputs and outputs of the R&D process. Public research refers to R&D carried out or funded by public and charitable organisations, and includes university research. Private R&D is carried out by privately owned enterprises. These inputs, either in isolation or in combination, lead to some ‘output’: new products, new patents or better performance (measured in a number of ways) by firms. These outputs generate additional GDP, that is to say additional income for the residents of the UK. As a preliminary step to quantify the spillovers generated by the R&D analysed in this work, we conducted a literature review on the topic of spillovers. This review

was originally intended to cover only public medical research. The literature was relatively scarce, so we extended the search to cover public and private R&D in general, without focusing on any industrial sector in particular. The Annex to Chapter Six contains all the details of the methodology used for this review and summarises the findings. Hence, the conceptual framework presented in this chapter applies to R&D in general, as well as to medical/pharmaceutical R&D. A. Complementarity between public and private R&D Within the biomedical sector, public research plays an important role in the discovery of new drugs and other health technologies, but the reality of the interaction between the public and private sectors is much more complex than the conventional picture of public research providing a straightforward input of basic knowledge to downstream, applied private research. Thus, the relationship between the public and private sectors in the pharmaceutical industry is not a ‘cascade model’. Figure 6.1 presents our (simplified) overall conceptual framework of how public research generates GDP and economic rent for UK residents. In overview: UK public and charitable research, at the left-hand side of Figure 6.1, leads to increased UK GDP at the right-hand side. Some or all of the effect is mediated via the UK private sector increasing its research (shown in the middle of the diagram). The arrows A1 and A2 illustrate the twoway relationship between public and private R&D. Arrow A1 illustrates the fact that some private R&D, labelled ‘new’ R&D, takes place in the UK thanks to public R&D in the UK. Some private R&D would take place even if all public R&D activities were to be eliminated, and this is labelled ‘existing’ R&D. Moreover, as the literature suggests, some public sector R&D is stimulated by the existence of private R&D – arrow A2, which stems from both ‘new’ and ‘existing’ private R&D. The empirical literature, particularly in the medical and biotechnology sectors, has focused primarily on quantifying how much private R&D is generated by publicly-funded R&D. Recent studies by Toole (2000, 2007), based on US data, show that basic medical and applied clinical research funded by public agencies, mainly undertaken in university and not-for-profit laboratories, stimulates and supports private investment on R&D in the pharmaceutical and biotechnology sector. Similar arguments have been presented by the US Congressional Budget Office: It is seldom possible to identify particular cases in which the private sector would have performed research if the government had not. Thus, most of the available empirical evidence is based on aggregate studies. On balance, that evidence suggests a positive relationship between public and private pharmaceutical R&D. (CBO, 2006, p. 31)

Figure 6.1 Sources of spillovers: public and charitable R&D and its interaction with private R&D

R&D inputs

Outputs (GDP)

(a) New start-ups (b) Existing firms

‘New’ R&D

Innovation (patents/new products)

‘Existing’ R&D

Firms’ performance: • Stock market valuations • Growth rates

A2 Caveats:

Public & charitable R&D

An earlier paper by Ward and Dranove (1995) also quantified this complementarity. Another part of the literature looking at the public– private research relationship explores how new firm start-ups and the location decisions of new firms are affected by university and other public research. Several authors argue that the decision by new firm start-ups of where to locate, including in the medical and biotechnology sectors, is influenced by traditional regional characteristics (such as size of cities, population) as well as by the opportunity to access knowledge generated by universities and other public laboratories. We have found no quantitative evidence on the magnitude of arrow A2 in Figure 6.1, although, as argued by Cockburn and Henderson (1996, 1998, 2000), the private pharmaceutical sector does invest heavily in basic research, viewing it as fundamental to the maintenance of a productive research effort. For instance, their work shows that there is extensive co-authoring of research papers between the public and private sectors. They conclude that the private sector results can have importance for public research.

Private R&D

• The spillovers within ‘private’ R&D are not shown. • The Figure does not include the geographic dimension. • The Figure does not distinguish between different sectors, e.g. health and non-health (for the ‘outputs’ in particular).

B. Public research as a source of spillovers There is wide agreement about the importance of universities and other public laboratories in generating economic growth. They are recognised as generators and repositories of public knowledge. Also, centres of commercial innovation and entrepreneurship are linked to proximity to universities. The literature has identified two potential effects of university research spillovers on: (1) innovation (patents/new product innovations); (2) performance/growth of firms. Referring back to Figure 6.1, arrows B1 and B2A/B correspond to effects (1) and (2) respectively. The link represented in Figure 6.1 by the arrow B1 is included in the literature on the (modified) ‘knowledge production function’. The knowledge production function explores the relationship between knowledge inputs, such as public and/or private R&D, and innovative outputs, such as patents and new products. This strand of the economic literature provides some evidence of the importance of geographically mediated commercial spillovers from public research. Indeed, public research has been found to have a positive and significant effect on both patents and new product innovations.

Arrow B2 in Figure 6.1 is included in the literature that argues that public research has a positive impact on private firms’ performance and growth. A number of papers suggest that geographic proximity and university/public laboratory spillovers are complementary determinants of firms’ performance. A combination of both factors results in significantly higher stock market performance and productivity for firms. As illustrated in Figure 6.1, arrow B2 shows potentially (at least) two channels for this university spillover. One channel implies that the firm’s performance improves because public R&D improves (somehow) the productivity of the existing R&D carried out by existing firms, which in turn leads to better performance (arrow B2A). The other channel by which public research can improve firms’ performance is a more direct impact on their productivity, other than via their own private ‘existing’ R&D (arrow B2B). Unfortunately, the literature leaves relatively undefined the exact mechanism by which this occurs, which is why we show arrow B2 with two ‘branches’. C. Spillovers from private R&D The literature has identified three types of spillovers generated by private R&D: (1) improving the productivity of other firms’ R&D; (2) encouraging entry of potential competitors; (3) reduction of production costs. For the first effect, the evidence presented in a number of articles suggests that the productivity of a firm’s R&D is dependent not only on its own internal R&D, but also on the R&D of other firms. This strand of the literature has explored this type of spillover in the pharmaceutical market in particular. For instance, Cockburn and Henderson (1994) find that competitors’ research appears to be a complementary activity to a firm’s own R&D: rivals’ R&D results are positively correlated with own research productivity. The authors interpret this as evidence of significant spillovers of knowledge across firms. Existing private R&D has also been shown to affect the entry decision of new firms (effect 2 above). For instance, Aharonson et al. (2006) argue that new entrants are influenced systematically by factors promoting the benefits of co-location, and seek locations that would allow them to benefit positively from knowledge spillovers. A number of papers have also explored how knowledge spillovers generated by existing firms shape the locational dynamics of the new entrants to the biotech sector, especially in the USA.

Arrow C in Figure 6.1 represents these effects taken together. Note the arrow comes out from the private R&D ‘bubble’, without distinguishing between ‘new’ and ‘existing’ R&D, because spillover effects could arise from either or both ‘types’ of private R&D. D. Mechanisms transmitting spillovers The literature highlights a number of mechanisms facilitating the transmission of spillovers. First, there are the mechanisms facilitated by universities, which include their pool of talented graduates, the ideas generated by faculty, the high-quality libraries and other facilities of research universities, and their publications. Second, networking and social interactions are also deemed to be important mechanisms, both formal and informal interactions. Formal ways of interaction include technology transfer programmes, such as licensing from universities to firms. Both means of interaction seem to be relatively important for the pharmaceutical and biotech market. It is important to highlight that part of the literature looking at networks concludes that geography might not be a sufficient condition for accessing a local pool of knowledge but rather that it also requires active participation in a network of knowledge exchanges. The third transmission mechanism discussed in the literature relates to ‘absorptive capacity’, i.e. the ability of economic agents to recognise, assimilate and apply new scientific knowledge – and thereby to appropriate some of the returns accruing to investments in new knowledge made by others (Cohen and Levinthal, 1989). These authors argue that, whereas the conventional wisdom was that R&D generated only one product, namely new information, R&D also enhances the firm’s ability to assimilate and exploit existing information – the firm’s ‘learning’ or ‘absorptive’ capacity. This second ‘face’ of R&D is very important, as it represents an important element of a firm’s ability to create new knowledge. Fourth, entrepreneurship has also been identified as an important mechanism to transmit spillovers, in that knowledge spillovers are the source of knowledge creating the entrepreneurial opportunities for new firms. International trade is usually deemed to be one of the most important mechanisms by which spillovers are transmitted across countries, but these are outside the scope of this work.

The third type of spillover effect generated by private R&D is reduction in production costs. For instance, it has been estimated that a 1% increase in the R&D spillover can decrease average costs between 0.05% and 0.2% (Levin and Reiss, 1984; Bernstein and Nadiri, 1989).

E. Measuring spillovers: the geographical dimension Figure 6.1 abstracts from the geographical dimension. For instance, it does not take into consideration where the public and private R&D actually takes place and how close you have to be to the source of the spillover to access it. This is nevertheless an important dimension, as the evidence gathered in our literature review clearly shows. The empirical evidence suggests that the link between knowledge inputs and innovative outputs becomes stronger when the unit of observation becomes increasingly aggregated (e.g. when the unit of observation is the country or the industry), and that location and proximity clearly matter in exploiting knowledge spillovers. Geographic concentrations of knowledge are likely to create higher levels of innovation than would otherwise be achieved. Not only do product innovations exhibit a pronounced tendency to cluster in regions which contain concentrations of innovative inputs, but also innovative activity tends to cluster more in industries where knowledge spillovers play a decisive role. The propensity for innovative activity to cluster is attributable to the role of knowledge spillovers and not merely to the geographic concentration of production. The literature also suggests that the role of knowledge spillovers is geographically bounded: innovative activity is more likely to occur within close geographic proximity to the source of that knowledge – ‘localised knowledge spillovers’.

Approaches to the quantification of spillovers Three main methodological approaches have been used to assess the value and benefits from R&D: econometric studies, surveys and case studies. The first method relies on the analysis of large databases. Surveys have been conducted of R&D managers drawn from the private sector, while case studies attempt to trace all the antecedents to an innovation. Each method has its own advantages and disadvantages. In this paper we draw upon the evidence generated by all three methods. The economics literature distinguishes between two types of return to investment: • ‘private’ or direct return to R&D investment, meaning the economic benefits generated by a specific R&D programme and accrued by the organisation (whether in the public or private sector) undertaking that research, through royalties and/or sales of a new product or process

The aim in this section of the report is to estimate the social return to the UK that is generated, in addition to any health gains, by public and charitable sector cardiovascular medical research in the UK, i.e. assessing the contribution of innovative efforts funded by government and charitable institutions to the national economy, excluding the health gains discussed in Chapter Five. The literature does not enable us to distinguish different strengths of effect according to which therapeutic area is invested in initially by the public and charitable sector. We therefore apply average estimates of the impact of medical research in general to the quantity of cardiovascular research spending, in order to estimate the impact of that particular category of research. The empirical literature provides quantified estimates of the extra GDP created as a result of extra public research expenditure, but not specifically for medical research. Indeed the clearest evidence is from the agricultural sector. However there are empirical estimates of the amount of private medical R&D stimulated by public medical research, and literature quantifying the extra GDP that results from extra private R&D. Hence, we have used two approaches to quantify the social return to public medical research: 1 two-stage approach: a) estimating the private R&D stimulated by public research (which is represented by arrow A1 in Figure 6.1) b) estimating the social rate of return to the private R&D so stimulated (which is represented by arrow C of Figure 6.1) 2 one-stage approach: a direct estimate of the social rate of return generated, by whatever transmission mechanisms, by public medical research (which is represented by arrows B1, B2A and B2B in Figure 6.1). The two-stage approach excludes any social benefit that might be achieved by public medical research independently of private R&D and so should in principle be expected to be less than or equal to the estimates produced by the one-stage approach. But because empirical estimates to enable the one-stage approach to be taken come only from non-medical sectors (mainly agriculture), we considered it desirable to provide a check by undertaking the two-stage estimates as well, as described in the following paragraphs.

• ‘social’ or total return to investment, which incorporates not only the benefits captured by the organisation undertaking the R&D but also the benefits spilling over for third parties to exploit, e.g. new knowledge and economic conditions that stimulate and enhance innovation and technical progress.

1a) Private R&D attracted by public/charitable R&D There are few studies measuring the positive pay-off generated by publicly funded medical research. We found two particularly relevant published empirical studies exploring the relationship between publicly funded and privately funded R&D in the pharmaceutical industry: Ward and Dranove (1995) and Toole (2007). Both studies refer to publicly funded medical research in the USA, funded by the National Institutes of Health (NIH), and to the impact of that on the total of R&D expenditure in the USA by all companies plus R&D expenditure by US-based companies worldwide. Toole (2007) provides separate estimates of the long-term elasticity of private R&D with respect to publicly funded basic research and publicly funded clinical research (Table 6.1). A 1% increase in NIH expenditure on basic research leads to a 1.69% increase in pharmaceutical industry R&D, after a lag of eight years. In the USA, private pharmaceutical R&D investment spending was 4.96 times the level of public spending on basic biomedical research. Thus an increment of $1 in US public basic research spending is estimated to stimulate $8.38 of private pharmaceutical R&D (1.69 x 4.96 = 8.38). According to Toole, a 1% increase in public clinical research expenditure leads to a somewhat smaller, but still significant, increase in private pharmaceutical industry spending. The transmission mechanism for this is not made clear but we suppose it might include public–private joint funding of clinical research projects. Toole finds that a 1% increase in NIH expenditure on clinical research leads to a 0.40% increase in private pharmaceutical industry R&D. This is fully achieved after a lag of three years, compared to the eight years it takes for the full extent of increased private R&D to be seen following an increment of public spending on basic research.

We do not have data on the split of public and charitable medical research in the UK between basic and clinical. But the split within the Medical Research Council appears to be around 60:40 and may be similar in the charitable sector. NHS-funded research is likely to be weighted towards clinical rather than basic. Thus we assume as a rough overall approximation that public and charitable medical research in the UK is in total split 50:50 between basic and clinical research. Thus, based on Toole’s US estimates, we assume that a 1% increase in public/ charitable research produces a ((1.69 + 0.40) / 2 =) 1.05% increase in private pharmaceutical R&D in the UK. Ward and Dranove (1995) do not distinguish between basic and clinical research NIH funding but they do distinguish between the impact of public medical research spend on private R&D spend within the same therapeutic category and the impact on private R&D spend in different therapeutic categories. They estimate that a 1% increase in publicly (NIH) funded basic research expenditure in a particular therapeutic category would, after a lag of seven years, cause a 0.76% increase in private industry R&D spend in that same therapeutic category and a 1.71% increase in private industry R&D spend in other therapeutic categories. Thus a 1% increase in NIH spend across all therapeutic areas leads to a 2.5% increase in the total of private pharma R&D spend in the USA plus worldwide R&D by US-based biopharmaceutical companies, taking seven years to have the full effect. Ward and Dranove do not report their data on the ratio of industry R&D to public R&D. We estimate the marginal impact relevant to the UK combining the elasticities estimated by Ward and Dranove (1995) and by Toole (2007), and the ratio of pharmaceutical industry to public R&D spend in the UK based on the data we collected for this study, which are presented in Chapter Three. For the latter ratio, in line with Toole’s method, we employ the private R&D figure

Table 6.1: Toole’s estimates of the impact of NIH-funded R&D on pharmaceutical industry R&D, USA Type of public research

Basic research

Clinical research

Long-run elasticity of industry R&D relative to publicly funded research

Ratio (industry total R&D/publicly funded research of that type)

Marginal impact

Source: Toole (2007)

Table 6.2: Estimated marginal impact based on Toole (2007) and Ward and Dranove (1995) elasticities Toole

Ward and Dranove

Ratio (UK industry R&D/public R&D)

Marginal impact (estimated)

Sources: Ward and Dranove (1995) and Toole (2007)

in the last year of the considered period (2005) relative to the average public research figure for the period 1997–2004 and the period 1998–2004, depending on the reported lag (eight years in Toole and seven years in Ward and Dranove). If we assume that the estimates in Table 6.2 represent the upper and lower ends of the marginal impact of extra public medical research spending in the UK, then a £1 increase leads to an increase in private pharmaceutical industry R&D spending in the range of £2.2 and £5.1.

industry and inter-industry level generated by R&D efforts, including cost function approaches (Bernstein and Nadiri, 1988, 1991), total factor productivity (TFP) (Griffith et al., 2004a) and production functions using patents as a measure of firms’ output (Jaffe, 1986; Scherer, 1982, 1984). Each study considers different sets of industries and data relating to US companies or OECD countries. Some of the references included in Table 6.3 are reviews (Garau and Sussex, 2007; Nadiri, 1993; PICTF, 2001).

A great deal of caution must be exercised when considering the relevance of these studies to the question of the impact of UK public and charitable research on UK private sector R&D because:

We conclude that the total (i.e. social) rate of return to private sector R&D tends to be around 50% but could be significantly higher.

• the scale of both publicly funded medical research and private sector R&D in the pharmaceutical industry are several times greater in the USA than in the UK; there is thus more scope for public/charitable R&D to have a within-country impact in the USA than in the UK • estimates of rates of return (the marginal impact) are sensitive to the estimated levels of R&D spend. It should also be noted that the two source papers do not consider the geographical dimension, that is the extent to which private firms’ propensity to raise R&D expenditure (for example opening a new R&D facility or improving an existing one) is linked to the location of key sources of (scientific) knowledge such as universities and publicly funded research institutes. The literature shows that geographic concentrations of knowledge generated by public research are likely to attract R&D investment of private firms and therefore give rise to regional clusters of innovative activities. Evidence provided by Furman et al. (2006) shows that this is the case in the biopharmaceutical sector.

Combining 1a) and 1b) As illustrated in Table 6.4, £1 extra spent on public medical R&D in the UK leads to an estimated increase of £2.2–£5.1 in R&D by the private pharmaceutical industry in the UK, which in turn yields a 50% rate of return to the national economy as a whole. Overall, for every extra £1 spent in public R&D plus the extra £2.2–£5.1 consequently spent by the private sector, the national economy earns a return equivalent to an extra £1.1–£2.5 per annum, respectively, of GDP thereafter. This represents a social rate of return to the total sum of public and private R&D investment (i.e. £3.2–£6.1 in our example) that is equivalent to 26% using Toole’s findings and 34% using Ward and Dranove’s findings. Thus the total social rate of return to the marginal ‘investment project’ that commences with £1 extra of UK public medical research spending is estimated to lie in the range 26–34%, i.e. of the order of 30%. This rate of return is assumed to apply equally to public cardiovascular research and to any other therapeutic target area of medical research.

1b) Social rate of return to private investment on R&D The second step of the two-stage calculation is to estimate the impact on the UK economy of the private R&D that is stimulated by the public R&D (arrow C of Figure 6.1). Table 6.3 summarises the findings of the empirical literature on the total economic returns – i.e. the ‘social’ returns – to private R&D spending. The rate of total (i.e. social) return to all parts of the economy is typically around 50% and greatly exceeds the rate of ‘private’ return captured by the firm doing the initial R&D (typically around 20%) in every case. The difference between the social and private returns is the return captured by firms, organisations or individuals other than the firm that made the original investment. A social rate of return of 50%, for example, means that for every pound invested now the economy earns a return that is equivalent to 50 pence per year for every year thereafter, indefinitely. The empirical studies listed in Table 6.3 use various approaches to estimate the productivity growth at the

Table 6.3: Social return to private R&D Study

Private rate of return

Social rate of return

Bernstein & Nadiri (1988)

Bernstein & Nadiri (1991)

Goto & Suzuki (1989)

Griffith et al. (2004a)

Griliches & Lichtenberg (1984)

Jaffe (1986)

Mansfield et al. (1977)

Nadiri (1993)

Approx. 50%

PICTF (2001), Garau & Sussex (2007)

Scherer (1982, 1984)

Sveikauskas (1981)

Note: In this table the ‘private’ return is that accruing solely to the firm making the R&D investment. The ‘social’ return is the total return to all organisations and individuals.

These calculations show that for £1 additional public medical research undertaken this year, the future gain to the UK economy would be equivalent to a stream of around £0.30 per year thereafter. But if the resources tied up in this public research, and the consequent private research it stimulates, had been put to other uses instead, then it can reasonably be assumed that these other uses would also have led to some increase in GDP. Nevertheless, there is published evidence (e.g. PICTF, 2001; Garau and Sussex, 2007) to show that medical research yields more extra GDP than would be gained if the same resources were invested in their next best alternative uses: that is to say medical research yields ‘economic rent’. Based on Garau and Sussex (2007) that extra amount is equivalent to about one fifth of the extra GDP generated: i.e. four fifths of the extra GDP you could equally well get by investing the same resources in something other than medical research, but one fifth is only obtainable if they are invested in medical research and nothing else.

2) One-stage approach: social rate of return generated by publicly funded biomedical R&D The second approach we explore relies on empirical evidence aimed at measuring the total social rate of return (excluding health gains) to R&D investment by the public and charitable sector, i.e. including the economic benefits spilling over from the party performing R&D to other organisations and bringing about output increase in the form of patents and new products (as represented by arrow B1 of Figure 6.1) and improvement of firms’ performance (as represented by arrows B2A/B). Empirical literature provides estimates of the social return to public investment in agricultural research but not medical research. Table 6.5 summarises the main findings.

Most of the works estimate the social rate of return to publicly funded research (in the agricultural sector) to be in the range of 20–67%. The implied 26–34% social rate of return to the total of public and private R&D, which we estimated in the two-stage approach above, lies within and towards the bottom end of this overall range, which is consistent with our prior expectation that the two-stage method would yield estimates equal to or less than the one-stage estimate. As noted by Buxton et al. (2004), and based on the review by Salter and Martin (2001), none of the studies measuring the benefits to an economy from publicly funded research provides a simple and comprehensive model. The main limitations of these studies are that they may be biased towards those government R&D programmes that proved successful and that they mainly employ US data which may not translate directly to the situation of the (smaller) national economies of Europe. Nevertheless, they are indicative of the existence of a large and positive economic contribution by public research to the national economy as whole, with a social rate of return of at least 20% and probably significantly higher than that. More recent contributions on the issue of measuring the rate of return of publicly funded research include the studies conducted by Mansfield, which focus on the contribution of research conducted by academic centres to innovation delivered by commercial sectors (arrow B1). Mansfield (1991) concludes that over 10% of new products and processes marketed by surveyed firms could not have been developed (without substantial delay) in the absence of academic research, and that this proportion is as high as 27–31% in the pharmaceutical sector specifically. Mansfield also provides an estimate of the social rate of return on academic research of 28%. The results of Mansfield’s later (1998) study are similar with respect to the impact of academic research on commercial innovation but provide no rate of return estimate.

Table 6.4: Social rate of return generated by increased public medical research Extra spend in public R&D

Additional private R&D induced

Social rate of return of private R&D

Overall social return due to a £1 increase in public R&D spending

Toole (2007), UK data (see Chapter Three)

Nadiri (1993), PICTF (2001), Garau & Sussex (2007), Griffith et al. (2004)

Ward & Dranove (1995), UK data (see Chapter Three)

Summary Taken together, our two estimation approaches imply that every £1 of extra public/charitable research spending in the cardiovascular (or any other) therapeutic area would yield a total social rate of return of at least 20% and perhaps as much as 67%, based on the, fairly sparse, empirical literature available. The more conservative of our two estimation methods implies a total social rate of return of around 30%. Taking this as our ‘best estimate’ of the GDP impact of medical research, implies that for an extra £1 invested in cardiovascular research this year, the UK’s GDP will be £0.30 higher next year and every year thereafter, than it otherwise would have been. This represents a healthy return on the investment and that is before any account is taken of the value of the health gains produced, as estimated in Chapter Five above.

Table 6.5: Estimates of social rate of return to public R&D in the agricultural sector Study

Griliches (1958)

Griliches (1964)

Huffman & Evenson (1993)

Knutson & Tweeten (1979)

Peterson (1967)

Schmitz & Seckler (1970)

Sources: Griliches (1991), Salter et al. (2000)

If all of the £122m (in 2005 price terms) of public and charitable cardiovascular R&D that was invested in 1992 were to yield a 30% rate of return, that would be equivalent to £37m of GDP every year thereafter. A 20% rate of return, the bottom of our range of estimates, would be equivalent to £24m of GDP every year thereafter; and a 67% rate of return, the upper end of our range, would be equivalent to £82m of GDP every year.

Chapter Seven

Estimating the rates of economic return from public and charitable CVD research KEY POINTS • In this Chapter we bring together the various elements necessary to calculate the return to the estimated investment in public and charitable cardiovascular research (from Chapter Three). • We consider the health return to be the value of the QALYs gained net of the health service costs to generate them (from Chapter Six). • We allow for our estimated proportion of this net benefit attributable to UK research (17%) and a time lag between research and health gains of 17 years (Chapter Four). • We define the internal rate of return (IRR) and use this as our preferred measure. • Combining our best estimates gives an IRR from net health benefits of 9.2% for CVD research. • Our one-way sensitivity analyses indicate a range of values for the IRR from 5% to 15%. In our ‘pessimistic’ scenario the research investment exceeds the value of the net health benefit; in our ‘optimistic’ scenario the net health benefit gives an IRR of over 25%. • In addition, we add our best estimate of the return in terms of GDP of 30%. • Our overall best estimate of the health and GDP gains combined from CVD research is an IRR of 39%.

This Chapter brings together the various estimates and pieces of evidence set out in previous Chapters to provide estimates of the rate of return on the investment in public and charitable funding of cardiovascular related health research.

The value of the research investment The first main component is our estimates of the research investment. In Chapter Three, drawing on multiple original sources, we produced estimates of the total expenditure on UK publicly and charitably funded research related to CVD for the years 1975–92. Once inflated to constant 2005 price levels, the central estimate showed that expenditure on cardiovascular research in 1975 was £144 million. Spending fell in real terms to £93m in 1981 and then rose fairly steadily again to £121m in 1992. Our alternative estimates, as used in the sensitivity analyses, were: a series of lower estimates of £103m (1975) to £97m (1992), and a series of higher estimates of £186m (1975) to £145m (1992).

The QALYs gained that can be attributed to CVD research In Chapter Five, drawing on a large number of health technology assessments and economic evaluations, we estimated the QALYS gained (and their value) for all the interventions leading to the main health gains in cardiovascular disease. We calculated the patientlifetime discounted incremental QALYs from 46 patient category/treatment combinations for interventions instigated in the years 1986–2005. An important element in the estimates of total QALYs gained is the gains from quitting smoking and for the reasons explained in Chapter Five these estimates show substantial variation between years. Our best estimate considers only those quitters whom we estimate did so with some specific intervention or contact with the NHS. This assumption produces estimates of the total QALYs gained per year in the period 1986–2005 ranging from 52,000 to 167,000 QALYs, with an annual average of 138,000. A more conservative methodology provides a series of somewhat lower estimates with an average of 112,000 QALYs per annum over the same period. Alternatively, taking a view that all those who quit smoking might reasonably be argued to have done so directly, or indirectly, as a result of medical research, gives a considerably higher series of estimates for the QALY benefit with an average of 182,000 QALYs per annum. We use these lower and higher alternatives in our sensitivity analyses. These are valued using the ‘opportunity cost’ threshold range for the maximum amount NICE is willing/able to pay to obtain a QALY from health care (given the current NHS budget) of £20,000 to £30,000, with a central estimate of £25,000. Each of these estimates only attributes to CVD research the benefits from a proportion of those who quit smoking equal to the proportion of total deaths avoided from smoking that are related to CVD. (This implies that if we were also estimating the return on R&D in cancer the health benefits from the remaining deaths avoided by quitting smoking would be attributed to cancer research.)

The health service costs of generating these QALYs In addition in Chapter Five, we estimated the discounted (net) patient lifetime NHS costs of delivering these treatments by the year in which they were instigated (at 2005 prices). These cost estimates ‘marry’ to the use of interventions that generate the QALYs. These also show annual variability (being influenced by the estimates of the number of interventions instigated in each year). Our best estimate series of these NHS costs per annum in the period 1986–2005 ranges from £415m to £1,355m, with an average of £781m per annum. The average of the lower range is £540m and that of the upper range £824m per annum.

We have already noted that the literature differs in terms of the expenditures to which the health gain is seen as a return. Some studies consider health gains as a return to health care, ignoring the role of R&D (e.g. Luce et al., 2006). Others attribute all the returns to R&D, effectively ignoring the health care costs (as in the Australian studies: Access Economics, 2003, 2008). We take the perspective of an initial investment (or stream of investments) in R&D resulting in a subsequent stream of net monetised health benefits; i.e. from the value of the health benefit that we ascribe as attributable to R&D, we deduct the cost of the health care required to generate that benefit. We are of course including the expenditure on all CVD research, not just that which might in principle be related to the interventions we consider, because the return on investment has to allow for not just the major R&D successes but also the research that was less successful, or unsuccessful, in relation to clinically valuable interventions. Effectively our approach assumes that the value of all other CVD interventions just equalled the cost of providing them (that is to say there was no ‘net benefit’ from these interventions at the prevailing value of a QALY). We believe this is probably a neutral assumption, assuming a degree of economic rationality in the use of NHS resources. The reality is not known: it might be that the net benefit (value of the health gain minus the cost of achieving it) for all other CVD interventions is positive or negative.

The time lag between the research expenditure and health benefits Combining evidence from our new analysis of citations in UK cardiovascular guidelines with evidence from other sources, we concluded in Chapter Four that our best estimate of the time lag is 17 years. We also consider in our sensitivity analyses the effects of lags of ten and 25 years. Given the time periods for which we have relevant data (which were determined in part by anticipation of likely lags we would use but also heavily by the availability of comparable data), the implication of these lags is that in our best estimate (17 years) we relate the investment in CVD research in the years 1975–88 to health benefits and their costs in 1992–2005. In the alternative ten-year lag estimates we relate investment in 1976–92 to net benefits in 1986–2002, and in the 25-year lag estimate we compare just five years’ investment in the period 1975–80 to benefits in 2001–05. (These time lag structures, and those used subsequently for Mental Health in Chapter Eight, are set out in the Annex to this Chapter.)

The proportion attributable to UK research CVD research is an international endeavour and much research is available rapidly as a public good to potential users anywhere in the world, so not all the UK net benefits can be attributed to UK research. (The corollary of course is that not all the net benefits from UK research

accrue in the UK, but as emphasised in Chapter One we are specifically ignoring benefits from UK research that accrue elsewhere). Our analysis of CVD guidelines, combined with existing evidence, led us in Chapter Four to conclude that our best estimate of the proportion of UK net benefits attributable to UK research is 17%, with low and high estimates of 10% and 25%.

Estimating the rate of return on investment With these various components and assumptions we are able to estimate the internal rate of return (IRR) on the research investment in terms of the net value of the health gains generated. The IRR is a convenient way of representing the return to the original research investment, and has the pragmatic advantage that it is the method used in the published empirical literature on the GDP impact of research discussed in Chapter Six. Expressing health gains as an IRR allows them to be added to the IRR for GDP gains to provide an estimate of the total rate of return achieved by medical research expenditure. For example an IRR of 10% means that the return to an investment of £1 is equivalent to receiving thereafter an income stream of £0.10 per year in perpetuity. Calculating an IRR avoids the need to apply a specific discount rate: the IRR is effectively the interest rate which would yield a zero net present value. However we have additionally presented net present values (NPVs) for these streams of research investments followed by net health benefits, calculated at the current Treasury approved discount rate of 3.5%. We have generally avoided using the benefit/cost ratio (the sum of the discounted benefits divided by the sum of the discounted costs) or the so-called ‘return on investment’ (the net sum of the discounted health benefits minus the discounted costs, divided by the discounted costs) because these ratios are susceptible to arbitrary definitions of what is included in the costs and benefits. This problem is particularly important in our analysis where we allow for the health service costs of delivering the health care that yields the benefits attributable to R&D.1 This problem does not occur in our IRR analysis because the health service costs are subtracted from the monetised health benefits each year. Table 7.1 summarise our estimated IRRs (and NPVs) for our best estimate and our main tests of sensitivity to alternative key parameter values and assumptions. Our sensitivity analyses consider the separate effect of changing the value of key parameters (the estimate of CVD research expenditure, the estimate of the net health benefit, the value of a QALY, the time lag, and the proportion of benefits attributable to UK research) using our alternative high and low estimates. In addition we present ‘pessimistic’ and ‘optimistic’ scenarios in which the values of a combination of parameters are changed together.

The best estimate IRR of 9.2% is a substantial return, given the general conservatism of the methods used. By comparison, an intervention that NICE ‘just accepted’ as generating QALYs at a cost per incremental QALY of around £25,000 would effectively be yielding a 3.5% return (the required discount rate). The sensitivity analyses suggest rates of return acceptable to the public sector have been achieved under all but the ‘pessimistic’ scenario. The sensitivity analyses also appear to behave predictably. They emphasise that the uncertainty around the estimates of research expenditure have a greater effect than uncertainty around the magnitude of the net benefits (the QALYs gained minus the NHS costs involved in producing them), at least within the ranges we have considered, but the value placed on a QALY is important. The results suggest that even more important may be: 1) the assumption concerning the appropriate lag; and 2) the percentage of total health benefits attributable to health research that are attributable to UK research (again within the limits of the ranges we have considered). For both of these assumptions the evidence is relatively weak. Our analysis in Chapter Six suggests that in addition to these returns in terms of the value of net health benefits, the social return in terms of the impact on GDP is even larger with a best estimate of a rate of return of 30% (with a range from 20% to 67%). Whilst technically our health return represents an average for the total cardiovascular research in the UK, and the social return estimates are for a marginal change in (any) healthrelated public research, it is not unreasonable to see them as additive: thus spending on past cardiovascular research may well have been delivering a total return of almost 40% (9.2% plus 30%). This effectively means that on average £1.00 invested in research will yield approximately £0.40 worth of benefits every year over an extended period of time.

Table 7.1: Estimated IRRs (and NPVs) for the health gain from CVD research Assumptions

NPV (@ 3.5% discount rate)

Best estimate (central/best estimates as explained in text)

Low estimate of research expenditure

High estimate of research expenditure

Low estimate of net health benefit

High estimate of net health benefit

QALY value of £20K

QALY value of £30K

25-year time lag

10-year time lag

10% of benefits attributable to UK research

25% of benefits attributable to UK research

‘Pessimistic scenario’: High research investment; low net benefit; QALY = £20K; 25-year lag; 10% attributable to UK research

Inv > net benefit

‘Optimistic scenario’: Low research investment; high net benefit; QALY = £30K; 10-year lag; 25% attributable to UK research

1 In our CVD base case, the discounted health benefits are £7,165m, the discounted health care costs are £1,646m and the discounted research costs are £1,228m. If the health care costs are netted off the benefits (in the numerator) the B/C ratio is 7.2; if they are added to the R&D costs in the denominator the B/C ratio is 2.5.

As will be discussed in the final Chapter, this does not tell us whether the return will be the same in future, or whether the return on a greater level of expenditure would or would not rise proportionately. It does however confirm that even using rather cautious methods and assumptions the return on CVD research is substantial. The final Chapter also attempts to throw light on the comparison of our data with that of the ‘exceptional returns’ claimed of the US and Australian studies.

Chapter Eight

Applying our methodology to mental health

KEY POINTS • Total annual public and not-for-profit funding for mental health research in the UK increased from £28m in 1975 to £93m in 1992, representing an annual percentage increase of around 7%. In real terms (2005 prices) this equates to a decrease from £155m in 1975 to £129m in 1992. • Based principally on the evidence of citations in 12 UK mental health guidelines, we estimate that the proportion of health care benefit that can be attributed to UK mental health research expenditure is 28%, and we test sensitivity to a range from 10% to 60%. • Based principally on evidence from analysis of the same guidelines, we estimate the time lag between mental health research spending and the health care benefit is around 12 years, ranging from nine to 14 years. • We estimate the net benefit from six patient group/intervention combinations that arose from research broadly in the relevant period. • The total value of the QALY gain from these interventions over the whole period of 1985–2005 was £31,435m. • The total incremental health care cost over the same period was £2,263m. • Our best estimate of the internal rate of return on UK public and charitable investment in mental health research is 7%. Our sensitivity analyses indicate that under plausible assumptions there is considerable uncertainty in the rate: it might range from around 11% to a situation where the benefits might not have exceeded the costs of research. • Overall this application suggests that our general methodological approach is equally applicable to mental health but that the greater data problems and uncertainties around the effect of interventions in mental health (as compared to CVD) mean that we can have less confidence in the results.

Earlier chapters have demonstrated that it is possible, though not without difficulty, to estimate the returns in terms of the value of incremental health gains that might reasonably be attributed to UK research in the area of cardiovascular disease. In defining this project we anticipated that to do the same for mental health would be more problematic. In part this reflected the findings in the Australian ‘Exceptional Returns’ study. The methods used there – building principally on temporal changes in disease specific mortality and morbidity – produced rather implausible estimates that the health gain from

research on mental health was negative. (This result was in part masked in their conclusions by combining research – and health gains – on ‘nervous system’ and ‘mental health’.) In this Chapter, we essentially follow the same steps presented in Chapters Three to Seven in relation to mental health to demonstrate to what extent the same approach as we adopted for CVD research can be applied to mental health research.

Estimated expenditure (public and non-profit) on mental health research This section summarises our estimation of the UK expenditure on mental health research in the period 1975–92. These estimates are based on similar assumptions to those outlined in Chapter Three, with some important differences: • For the non-profit sector, we have assumed that mental health research is supported principally by the Wellcome Trust, and identified relevant research using the grants database and the search term ‘mental health’. Because of the variations in classification scheme over the period 1975–92, all neuroscience panel grants were also included. • As for cardiovascular research, the MRC has used a number of systems for classifying grants within the broader field of mental health research over time. To allow for cross-comparison with cardiovascular research, we retained our focus on the years between 1975 and 1992, but the MRC used at least three different classification systems during this period. From 1972 to 1975, all research grants were classified under an ‘inclusive’ category entitled ‘mental disorders’. From 1976 to 1992, the MRC distinguished between research on ‘mental handicap and psychiatric disorders’ and ‘addiction’, each of which had ‘exclusive’ and ‘inclusive’ definitions (understood in the same way as for cardiovascular research). Since 1992, all grants in this area have been classified as ‘neuroscience and mental health’. Our figures for the MRC are based on the ‘mental disorders’ category for 1972–75, and then totals for inclusive definitions of ‘mental handicap and psychiatric disorders’ and ‘addiction’ from 1976–92. • Based on figures for MRC and Wellcome Trust expenditure, as well as those cited in a one-off report (Dawson et al., 1998), we have assumed mental health research activity to be 10% of total public and charitable health research activity in the UK (and looked at a range from 5% to 20%), where no breakdown by clinical area is available.1 • For the for-profit sector, Ward and Dranove (1995) do not show data specifically for a mental health therapeutic category, but rather for ‘neuro’, which will include neurological conditions outside of mental health. According to their data – for global R&D spend

Figure 8.1 presents our estimates of expenditure on mental health research between 1975 and 1992 by five major sources of funding, from the government and non-profit sectors (see Annex 8A for the data used to construct the figure). We find that, between 1975 and 1992, expenditure on mental health research increased in current prices from £28m to £93m in 1992, representing an annual average increase of around 7%. Total expenditure over the period was £929m. Figure 8.2 provides comparative estimates of research expenditure by the UK pharmaceutical industry on mental health research and indicates that at the end of the period industry spending in this area was approximately three times that of the public/charitable sector. Figure 8.3 provides high and low estimates for the public/charitable expenditure. Figure 8.4 shows the public/charitable spending at constant (2005) prices and shows that in real terms expenditure fell substantially in the first decade and then increased again but did not regain its original value.

Figure 8.1: Total estimated expenditure (public and non-profit) on mental health research, by source of funding, 1975–92, at current prices

Total MH Research spend (£m cash)

Figure 8.2: Total estimated expenditure by the pharmaceutical industry on mental health research, 1975–92, at current prices 300

Spending (£m cash)

by US pharmaceutical companies and US R&D spend by non-US pharmaceutical companies – ‘neuro’ accounted for 22% of total R&D expenditure in 1966 and 17% in 1988. Similarly, CMR International refer to the therapeutic category ‘nervous system’ which includes, but goes wider than, mental health.2 From their sample of around 20 multinational pharmaceutical companies, they found that nervous system R&D represented 19% of total global R&D expenditure in 2003, 16% in 2004 and 17% in 2005. Combining these two sources of information, we assume that ‘neuro’ or ‘nervous system’ R&D is around 19%, ranging between 16% and 22% for our high and low estimates, of total private sector pharmaceutical R&D spending in the UK. We are unable to estimate within that how much went to mental health specifically rather than neurological conditions.

Analysis of UK clinical guidelines in mental health We reviewed 12 mental health guidelines published by the National Institute for Health and Clinical Excellence plus the Mental Health National Service Framework (NSF) published by the Department of Health (see Annex 8B for a list of the mental health guidelines that were analysed).3 A total of 3,423 papers from the 13 guidelines that were matched to the CWTS database were analysed. Key results from the bibliometric analysis were: • 28% of cited papers were published by UK authors (ranging from 11% for obsessive–compulsive disorder to 58% for the Mental Health NSF). (There is a suggestion that the contribution to healthsystems-related guidelines from UK authors may be higher than their contribution to more biological and biomedical guidelines, their contribution to the latter being closer to the general UK contribution to

0 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992

combined international research in clinical medicine.) We use 28% as our best estimate and look at the sensitivity of the results to a range from 10% to 60%. • The mean knowledge cycle time (the average time between the date of cited publications and the date of publication of the guideline) for all 13 guidelines was around nine years, ranging from nearly six years for the Mental Health NSF to just over 11 years for obsessive–compulsive disorder (OCD). Adding three years as a plausible estimate of the period between funding and publication gives a total time lag of 12 years, ranging between nine and 14 years.

Estimating the number and value of the QALYs gained We applied essentially the same approach as for CVD to estimate the monetary value of health gains from specific interventions to treat or prevent mental disorders adopted over the period 1985–2005, and estimate the health care costs incurred in delivering these gains. To do this we:

Figure 8.3: High and low estimates of total mental health research spend by public and non-profit funders, 1975–92, at current prices 160

1 identified specific interventions for patients with mental disorders

2 estimated the QALYs gained for each intervention in each patient group

4 adjusted for compliance with treatment 5 computed the total QALYs gained from each intervention 6 valued the total QALYs gained. From our review of the economic evaluation literature, we also obtained estimates of the incremental costs associated with each intervention and multiplied these by the numbers of users who are compliant with treatment to quantify the health care costs associated with each intervention. Chapter Five sets out the principles of our approach in more detail. Here we simply summarise the steps, highlighting methodological decisions which were specific to the estimation of health gains and costs of interventions to treat mental disorders. This general account of the methods is followed by an example (in Box 8.1) explaining how these principles were applied in the specific case of SSRIs. Stage 1: Identifying specific interventions for patients with mental disorders In the absence of a starting point equivalent to the IMPACT study (Capewell et al., 2007), we selected the patient groups and treatments based on treatment recommendations in the National Service Framework (National Health Service, 1999; Department of Health, 2004b). We focused on a small set of interventions in depressed and schizophrenic patients thought to be most relevant in terms of the additional health gains they have produced since 1985. In summary, for the treatment of depression, we included: • Selective serotonin reuptake inhibitors (SSRIs) – on the assumption that patients receiving these would

Figure 8.4: Expenditure on mental health research by public and non-profit funders, 1975–92, at 2005 prices Total MH research spend (£m real at 2005 prices)

3 computed the numbers of users of each intervention in each patient group

We note, in passing, that the papers cited on mental health clinical guidelines were being cited three times as often as would be expected in the field, consistent with the expectation that the very best (or at least the most cited) papers in the field inform guidelines.

1 High 2 Mid-point 3 Low 300 250

not otherwise have received drug treatments because SSRIs removed the safety concern (particularly relating to drug overdoses) of the existing treatments. We make no assumption that on a per-patient basis SSRIs provide additional QALYs over the tricyclic antidepressants: the QALY gain from SSRIs comes from treatment of patients who would otherwise have remained untreated. • Cognitive behavioural therapy (CBT) – on the assumption that these patients would otherwise have received usual care in general practice. For the treatment of schizophrenia, we included: • Atypical antipsychotics – atypicals are widely perceived as safer, are easier to prescribe than earlier antipsychotics and have been more widely promoted, which has led to much wider use of antipsychotic medications allowing treatment of patients who would otherwise have remained untreated.

years 1985–87, we applied the 1988 figure. For 2004–05, we applied the 2003 figure. See Annex 8D for further details on how new contacts with community psychiatric nurses were calculated.

For the treatment of depression, schizophrenia and other mental disorders: • Community psychiatric nurses (CPNs) – on the assumption that these patients would otherwise have received usual GP care. The patients groups and interventions are in Table 8.1 column 1. Stage 2: Estimating the QALYs gained for each intervention in each patient group Based on a review of the economic evaluation literature, we estimated the QALYs gained from each intervention in each patient group (Table 8.1 column 3). All the estimates of the QALYs gained we obtained from the literature adopted a time horizon between one and 1.5 years. To provide comparable estimates across interventions, we calculated the number of QALYs gained in each of the first 12 months of treatment and then summed the monthly gains up to obtain a QALY estimate for one calendar year. When we did not have data on the quality of life score in each month, we assumed a linear increase in quality of life between baseline and 12-month measurement. In all cases we assumed no continuing benefit (or effect on costs) beyond the period of treatment. See Annex 8C for further details on the studies used to generate these estimates. Stage 3: The numbers of users of each intervention in each patient group To calculate the total QALYs gained, we multiplied the QALYs gained for each specific intervention by the numbers of users of that intervention in each year. For SSRIs we could approximate the number of users (strictly ‘user years’) based on the number of SSRI prescriptions in the UK in each year. A detailed description of how the SSRI users were computed can be found in the example later in this chapter and in Annex 8D. For the community psychiatric nurses we had data on the numbers of initial contacts in each year between 1988 and 2003. For the

For all other interventions we did not have this information. Analogous to the CVD model, we therefore modelled the numbers of users as a function of the numbers of patients in each patient group and the uptake rate for each specific intervention. The approach is explained in Chapter Five. When we were not able to find complete data on the number of eligible patients or the uptake rates, we modelled the numbers based on the available empirical evidence. Annex 8D gives details how the estimates for the each eligible patient group were generated and Annex 8E gives details on how the estimates for the uptake rates were generated. Stage 4: Adjusting for compliance with treatment We searched the economic evaluation studies used to obtain the QALYs estimates for each specific intervention. None of the studies included noncompliance rates in their analysis. We therefore applied a compliance rate of 55% for schizophrenic patients taking atypical antipsychotics, and a compliance rate of 65% for depressed patients taking SSRIs. The figures are based on research by Velligan et al. (2003) and Cramer and Rosenheck (1998). We did not adjust for compliance with community psychiatric nurses and cognitive behavioural therapy because we used data on direct contacts for these interventions in our analysis. Accounting for compliance gives an estimate of the numbers of compliant users of each specific intervention in each year. These are reported for each intervention in Table 8.1 column 2. Stage 5: Estimating total QALYs gained from each intervention We computed the total QALYs gained from each intervention in each year by multiplying the numbers of users in that year by the QALYs gained from each

Table 8.1: Summary of results by patient group/intervention, 1985–2005 1

QALYs gained Patient groups/intervention

Compliant user years (000s)

Incremental costs

Per user year

Total monetised (£ million)

Per user year (£)

Total (£ million) 931

Depression: 9,160

Community psychiatric nurses

Behaviour or cognitive therapy

Atypical antipsychotics

Schizophrenia:

Other mental disorders: Community psychiatric nurses Total

SSRI = selective serotonin reuptake inhibitor

intervention. This approach is different from the approach taken to calculate the monetised health gain from CHD research. The estimates of QALYs gained from CHD interventions reflect the health gain over a lifetime so that we had to multiply the QALY gain in each year by the number of new users. The QALY estimates for the interventions to treat mental disorders reflect a 12-month gain and could therefore be multiplied by the number of users in each year. Focusing on user years rather than new users is on the one hand likely to be conservative because it assumes that the QALYs gained in the first year of use, when there may be a delay from the start of therapy to the time at which an effect is achieved, apply to subsequent years. On the other hand it may overestimate the QALYs gained because it assumes that the treatment effect does not diminish over time. We then summed the result for each intervention over the period 1985–2005 to calculate the total QALYs gained. These are reported in Table 8.1 column 4. Stage 6: Valuing total QALYs gained We placed a value on the total QALYs gained by multiplying the total QALYs gained by £25,000, which is the central point of the UK’s National Institute for Health and Clinical Excellence’s threshold range of £20,000 to £30,000 per QALY, giving a value of the total QALYs gained of £31,435m over the period 1985–2005 (Table 8.1 column 5). Estimating the health care costs from the specific mental health interventions From our review of the economic evaluation literature, we also obtained estimates of the costs associated with each intervention. Where UK papers were used to generate the QALY gains estimates we used the same papers to obtain the cost estimates. This was the case except for cognitive behavioural therapy for depressed patients, which was from the Netherlands, and the QALY estimate for atypical antipsychotics for schizophrenic patients, which was from the USA. For cognitive therapy in depressed patients, we used a cost estimate from a UK economic evaluation of cognitive behavioural therapy in patients with depression with an outcome other than QALYs (Scott et al., 2003) and for atypical antipsychotics, we used a UK economic evaluation of atypical antipsychotics in schizophrenic patients who are responding poorly to previous therapy (Davies et al., 2007). Annex 8F gives further details on the studies used to generate the estimates for costs for each intervention to treat mental disorders. A limitation of our cost estimates is that the major gains in treatment came not just from SSRIs, but from their widespread use and adoption. We have used cost data for prescribing SSRIs but not for any associated medical care such as GP visits. This is consistent with a counterfactual that these patients would have received GP care but would not have received medication for depression. But this will underestimate the costs if without SSRIs these patients

would not have received as much GP or other care. A similar issue may also apply to our costs of atypical antipsychotics. The costs of each intervention per user year are reported in Table 8.1 column 6 and the total cost for users in the period 1985–2005 are reported in Table 8.1 column 7. The total cost over the period was £2,263m.

Upper and lower estimates of QALYs gained and associated health care costs The overwhelming proportion of the QALY gain is estimated to come from the use of SSRIs. In general, SSRIs (the new medications introduced circa 1987) have largely replaced the older medications (tricyclics) and have been prescribed to a much larger share of the population. The main advance of the SSRIs was their ease of prescription as well as their general safety. This led to much wider use of SSRIs by non-psychiatric specialists and for a much wider range of anxiety-affective conditions than just major depression. A similar concern has been expressed about atypical antipsychotics, as the greater increase in use of these medications has been outside the core indication of schizophrenia (Domino and Swartz, 2008). Unfortunately, we do not have data on the breakdown by the specific conditions for which these drugs are increasingly used in the clinic, or the QALY gains in those precise situations. These are the key factors that we have reflected in our upper and lower estimates of the net health benefit. On the other hand we have been conservative in our assessment of the QALYs gained for a new patient receiving atypical antipsychotics. Our assessment of the change in utility is based on change from baseline in a recent and robust study (Rosenbeck et al., 2006). However, that study obtained QALY estimates in patients switched from a current treatment to an atypical antipsychotic. But many of the patients in that study were already on treatment so their baseline QALY level might have been higher, and hence utility gain from the new medication might have been lower, than for patients who were receiving no treatment and were started on atypical antipsychotics. An older, shorter (eight-week) study in patients not receiving treatment, but using a less appropriate basis for calculating QALYs, indicated a considerably greater utility gain of 0.146 (Chouinard, 1997). Our lower estimate of the net health gain made three adjustments compared with our best estimate: • It adjusted downward the quality of life gain from SSRIs, reflecting the suggestion that the QALY gain compared to placebo may be more modest as they are increasingly used in general practice and less severe patients (Kirsch et al., 2008). Our lower estimate assumed that each SSRI user gained only half the QALY gain compared to that assumed in the best estimate.

• Rather than simply allowing for the prescription cost of SSRI treatment and assuming that prescribing the drug had no effect on patient contacts with the health care system, our low net benefit case used an estimate of the full cost of patient care, reflecting the alternative assumption that these patients would otherwise not be receiving active health service care (Kendrick et al., 2006). • It used a more conservative estimate of the uptake rate for atypical antipsychotics between 1985 and 2005. Our upper estimate made two adjustments: • Our best estimate calculated the QALYs gained per year as a steady increase to observed improvements at one year. This assumption may have underestimated the QALY gain for those patients who are on medication for more than one year. The upper estimate assumed that half the years of use were for new users and the other half for patients continuing on SSRIs and enjoying the full benefit throughout the year. • Our best estimate of QALYs gained for patients receiving atypical antipsychotics reflected a study in which patients switched from a current treatment to an atypical antipsychotic (Rosenbeck et al., 2006). The gain for these patients may have been less than for patients who were not previously treated. For this higher net benefit estimate, we used a higher estimate based on a small earlier study of previously untreated patients (Chouinard, 1997). Using the adjustments described above, the lower estimate assumptions resulted in total QALYs gained of 656,000 over the whole period, the total monetary value of QALYs gained of £16,408m and total health service costs of £10,585m (see Annex 8G). The upper estimate assumptions resulted in total QALYs gained of 1,500,000 and total value of QALYs gained of £37,490m. The estimated health care costs for the upper estimate were unchanged compared to our best estimate.

Challenges to the estimation of QALYs and costs The use of this methodology in mental health produced few, if any, distinctive problems but emphasised the need for a strong evidence base relating to the use of interventions and the health gain (in QALYs) and the costs associated with their use. Generally the appropriate evidence was much weaker for mental health, in part reflecting less use, and arguably less applicability, of QALYs as a measure of effect in this disease area. Thus even for the few interventions we considered we were forced to use some non-UK studies and were not able to use any meta-analyses. A further aspect of the weaker data was that there were few studies that considered long-term cost and benefits and most of our analysis is based on one-year studies.

Box 8.1 Example of SSRIs Kendrick et al. (2006) undertook an economic evaluation of SSRIs and tricyclic antidepressants (TCAs) for the primary treatment of depressed male and female patients in the UK. Based on the findings of a randomised controlled trial, they reported the EQ-5D score at baseline and 12 months for the SSRI and TCA group respectively (the SSRI group improved from 0.61 at baseline to 0.78 at 12 months). We calculated the utility gain in each of the 12 months for patients on SSRI as compared to baseline, summed it up and arrived at an estimate of 0.13 QALYs gained per depressed patient per year. This is probably a conservative estimate as we allow for a gradual increase in quality of life for patients receiving SSRI treatment. The number of patients using SSRIs was based on antidepressant prescription data for the UK between 1975 and 1998 (Middleton et al., 2001). For the years following 1998 we linearly extrapolated the prescription data based on the increase in prescriptions between 1997 and 1998. We then translated the number of prescriptions in the number of users for each year (or strictly the number of user years). In the study by Kendrick et al. (2006) 931 SSRI prescriptions were given in 81.6 person years of treatment. Consistent with this, we divided the number of prescriptions in each year by a factor of 11.4 to estimate the total number of SSRI user years. The estimated total number of user years for the period 1985 to 2005 was 14,092,497. Following Cramer and Rosenheck (1998) we assumed that only 65% of all patients will comply with SSRI therapy. The estimated number of compliant user years of SSRI therapy was 9,160,123. The prescription data show that the use of SSRIs is exponentially increasing after their introduction in 1987, while the number of prescriptions for other antidepressants is effectively constant between 1975 and 1998. We assumed that the steep increase in the total number of SSRI prescriptions between 1987 and 1997 is fully attributable to SSRI users who would not otherwise have received antidepressants medication. Our assumption reflects the general belief that SSRIs are thought to be safer for treating depressions because the drugs themselves cannot be easily used for suicide in contrast to other classes of antidepressants. We consequently attribute the full QALY gain (between baseline and 12 months) of 0.13 to each of the compliant SSRI user years. This gives a total gain from SSRI therapy in depressed patients of 1,196,541 QALYs. (Our analysis is not based on an assumption of superior efficacy of SSRIs over TCAs.) Multiplying this figure by the opportunity cost of a QALY (£25,000) gives a monetary value of the health gain of £29,914m. Kendrick et al. (2006) estimated the yearly costs per patient on SSRI therapy to be £87. Converting these to 2005 UK£ using the NHS pay and prices index gives an annual cost of £101.7 for SSRI treatment. Multiplying this figure by the number of users gives a total cost of £931m.

Further, we are likely to have been conservative in the inclusion of only six patient indication/intervention combinations. For example, we were unable to find appropriate evidence to include an estimate of the health gain from developments in case management practice. We have however included our estimate of all mental health research as the research investment and some of this may have led to beneficial interventions and to the study of conditions that we were not able to include. As for CVD, the assumption, in the absence of information on other interventions, is that on average the net benefits (in terms of health gains minus costs) from the provision of these non-analysed interventions are zero.

Estimates of the return to mental health research Applying exactly the same methodology as for CVD we calculated the internal rate of return (IRR) on investment in mental health research. We used the annual estimates of research expenditure shown in Figure 8.4, and the annual estimates of health gains and costs summarised in Table 8.1. Specific estimates for mental health of the time lag between research and health effect of 12 years (range 9–14 years) and of the proportion of health gain attributable to UK research of 28% (range 10–60%) were applied. This gave a best estimate of the IRR for mental health research of 7.0% and from a series of one-way sensitivity analyses the IRR ranged from a return that was less valuable than the research investment, up to returns of around 11.0%. A ‘pessimistic scenario’ (inevitably given the one-way sensitivity analyses) showed an even greater excess of the cost of the investment over the value of the net health benefit, whilst an ‘optimistic scenario’ showed an IRR of over 15%. These results (see Table 8.2) do not show the same relationship to lag structure as was shown in CVD. This is explained by the fact that the net benefits we have measured were very low at the beginning of the period (1985 and subsequent years) but increased dramatically in the later years (2001–05). The longer time lags omit the earlier years (with small health gains) but include the much larger gains in the later years (see Annex to Chapter Seven). (This problem would be lessened if longer time series were available.) Otherwise the relationships and conclusions are similar to those for CVD. Here again, the results emphasise the importance of the assumptions regarding the proportion of the health benefit attributable to UK research. Our proxy indicator of the proportion of UK guideline citations from UK addresses showed greater variability in mental health and the sensitivity analysis using that range (10–60%) showed considerable variability. These rates are better than the implicit rates for mental health in the earlier Australian study (Access Economics, 2003) but lower than those for CVD in this study. For

Table 8.2: Estimated IRRs (and net present values) for the health gain from mental health research Assumptions

INV > net benefit

14-year lag

60% of benefits attributable to UK research

‘Pessimistic scenario’: High research investment; low net benefit; QALY = £20K; 9-year lag; 10% attributable to UK research

‘Optimistic scenario’: Low research investment; high net benefit; QALY = £30K; 14-year lag; 60% attributable to UK research

mental health as for CVD, a best estimate GDP return of 30% (range between 20% and 67%) should be added to the net health benefit return calculated above to give an overall return to the public/charitable investment in mental health.

So are the methods applicable to clinical areas other than CVD? This test, applying our methods to mental health, emphasises that our bottom-up approach is data hungry and requires definition of a set of appropriate interventions and detailed evidence relating to the use of, costs of and benefits from them as compared to a counterfactual assumption. Each of these steps was more difficult than for CVD, and because of the less clear evidence base the results may well be more contentious, but the process was possible and the attempt was informative. We see no reason why the approach could not be applied in other areas too, but confidence in the results will depend on the quality of the data and other evidence available. 1 See www.dh.gov.uk/en/Researchanddevelopment/ Researchanddevelopmentpublications/informationfromresearch/ DH_4078384. 2 M Ogg, personal communication, 28 November 2007. 3 The NICE guidelines are available at www.nice.org.uk/guidance/index.jsp ?action=byTopic&o=7281&set=true. 12 of the 14 guidelines that NICE has published in the field of mental health were chosen. The two not selected were Anxiety and Eating disorders as it is was felt these subject areas were too broad to provide useful data in the current context. The mental health NSF is available at www.dh.gov.uk/en/Publicationsandstatistics/ Publications/PublicationsPolicyAndGuidance/DH_4009598.

Chapter Nine

Our contribution, our conclusions and main reservations, and a research agenda KEY POINTS • The main contribution of this report is to have demonstrated a theoretically sound basis for empirically estimating the two main elements of the economic returns to medical research. • The more specific original contributions in terms of both data and analysis are summarised. • We stress that these estimates need to be treated with caution and explain the main caveats. • We emphasise that our measure of return – the internal rate of return or IRR – is quite different to the benefit/cost ratios used in the Australian studies. • We reiterate the main characteristics of our methodology (compared with other studies including the Australian studies), namely that it: • is based on detailed bottom-up estimates, rather than calculated top-down from temporal differences in mortality and morbidity • uses only observed historic data – not projections of what may happen in 20 or more years’ time • nets off the health care cost necessary to deliver the health gains • uses a value of a QALY that reflects the opportunity cost in the NHS rather than a theoretical social willingness to pay. • Finally we set out a research agenda to improve our understanding of key issues and to enable us in future to make better estimates of the returns to medical research.

The methodological contribution of our report Overall we have provided a clear demonstration of an approach which provides an improved theoretical basis for empirically estimating the two main elements of the economic returns to medical research – firstly the returns in terms of the value of the health gained and secondly the considerable returns in terms of impact on GDP. More specifically our main original contributions are to have provided: • A consistent time series of estimates of UK public/ charitable and private pharmaceutical industry expenditure on medical research in cardiovascular disease and mental health from 1975 to 1992, pieced together from a variety of sources.

• A clear conceptual framework to underpin the concept of ‘spillovers’ from publicly and charitably funded medical research, based on an original broadly scoped literature review. • Estimates of the magnitude of spillovers in the UK from public and charitable medical research, based on existing literature, but calculated in two different ways: a two-step analysis of the relationship between public (and charitable) and private R&D and then of the relationship between private pharmaceutical R&D and GDP; and based on the economic literature estimating the social rate of return to public R&D, whether medical or not. • A ‘bottom-up’ approach to estimating the health gain from research, developed for this study. This represents a significant improvement on earlier attempts to estimate the economic returns from research in that it is measured in terms of QALYs and it is driven by relatively hard evidence on the effects and costs of specific research-derived interventions, not from macro-level, temporal changes in mortality or morbidity. • A ‘successful’ test of this approach in two disease areas: the analysis of the gains for cardiovascular disease was built up from evidence on 46 different patient indication/treatment combinations: that for mental health on six such combinations. These two areas had been chosen for the following reasons: cardiovascular disease on the prior assumption that there was more appropriate data and evidence, and mental health as a particularly challenging area to investigate, which was potentially likely to expose methodological and theoretical difficulties of the analysis. • Analyses of UK clinical guidelines in the areas of cardiovascular disease (five guidelines) and mental health (12 guidelines), to provide indicators to inform the important issues of the lag between research expenditure and health benefits, and the attribution of benefits to UK (rather than worldwide) research. • Computation of the internal rate of return (IRR) on past expenditures on research investment in the areas of CVD and mental health, allowing explicitly for the estimated time lags involved and the level of attribution to UK research, with sensitivity analysis around key parameters. • A comparison of the IRRs on research investment from the value of the QALYs gained in these two areas, and with the new estimates of the IRR in terms of the GDP effect (which is not specific to individual disease areas).

Our conclusions on rates of return and our main reservations All our work emphasises to us that our ‘bottom-line’ estimates of the rates of return need to be treated with extreme caution. Nevertheless, our ‘bottom-line’ best estimate suggests that for CVD the IRR from the value of net health gains alone (ignoring GDP impacts) is around 9%. Most oneway sensitivity analyses place the IRR within the range of 6–14%, but in the ‘pessimistic scenario’ the value of the net benefit was less than the cost of the investment whilst in the ‘optimistic scenario’ the IRR was over 25%. For mental health, our best estimate of the IRR from net health gains was around 7%. Most one-way sensitivity analyses place the IRR within the range from a situation where the investment exceeded the net benefits to a positive rate of return of just over 11%. In the ‘pessimistic scenario’ again the research investment exceeded the net benefits, but in the ‘optimistic scenario’ we estimated a return of around 16%. These rates of return are very sensitive to the lags between the research and the health gain. The relationship in these two particular cases is made more complex by the data limitations and artefacts of the differing time periods we were able to consider with different lag structures. However it is clear that in reality, other things being equal, the sooner the benefits can be realised the greater their value relative to the research investment. From a policy point of view this is crucial: a ‘guaranteed’ way to improve the return is to speed up each or any step in the various stages from basic research to the health gains that it can lead to, and so to reduce the time lag. It is inevitable that these rates will be compared with the ‘exceptional returns’ that have featured in the international literature. As previously stated, the ‘exceptional returns’ for the USA were largely implicit rather than formally estimated as rates of return on a specific past investment, so direct comparisons are not possible (Funding First, 2000; Murphy and Topel, 2003). However, the two analyses from Australia are much more explicit (Access Economics, 2003, 2008). However, not only are the Access Economics methods of estimation different in the two studies and different again from our own, but crucially their method of measuring the return is peculiar, in that they express it as a ratio of the value of health benefits to cost (B/C ratio) or as the ratio of health benefits minus research costs to research costs (so called return on investment). We principally estimate the more meaningful internal rate of return: see Chapter 7 for a fuller explanation of this measure. Using IRR as our standard measure enables us to combine the value of the health gain and GDP returns and it can also be compared, if desired, with returns to other public and private investments.

Therefore, the first point to make is that our estimate of an IRR of 9% for CVD cannot in any way be compared with the estimate of an ‘ROI’ of 788% for CVD in the 2003 Australian study or of 117% for health R&D overall in the 2008 Australian study. If we take our own (discounted) base-case estimate of CVD benefits and of CVD research investment, the ROI, broadly as measured in the Australian study, would be 484%. But we do not accept that that is a meaningful figure. So what are the key substantive differences in our estimates? The key differences between our methods and those of the Australian studies are: • The lag structure: the 2003 Australian study compares expenditure and health gain in the same year; the 2008 study uses a 40-year lag. Our CVD best estimate uses 17 years; our mental health best estimate uses 12 years. • Neither Australian study appears to net off the health service costs involved in achieving the health gain from improved services. • Both Australian studies use high values for a life year: the first, based on an ‘eclectic’ calculation underpinned by the US estimates of individuals’ ‘willingness to pay’ for a statistical life used AUS$150,000 per life year gained, and the second, based on a meta-analysis of studies, used AUS$267,000. We valued a QALY at £25,000 (approximately one-fifth of the most recent Australian value), a value that reflects the opportunity cost of the spending on R&D if it were instead to be spent directly on health care. • Both Australian studies use a ‘top-down’ approach, starting with overall or disease-specific estimates of national changes in mortality and morbidity over time, and attributed 50% of these changes to health research. Moreover, to cope with the 40-year lag, these were projections of what morbidity/mortality might be in 2033 and later. Our study is ‘bottom-up’, aggregating health gains from observed changes in the use of specific evidence-based technologies. We believe that our approach is both more logical and more firmly based in the reality of UK clinical practice and health policy. Nevertheless, we stress that the results should be treated cautiously. Our reservations and cautions relate both to what we have done and to what we have not done: 1 In terms of our analysis, our estimates of the investment and of the health gain both involve a series of assumptions and estimates, about which we have been explicit. Further research and data collation, more systematic reviewing of the evidence, and greater input from clinical specialists could no doubt improve some of these assumptions, and refine these estimates. But it is not self-evident from the lack of

sensitivity of the return to differences in the estimates of the QALY gain, that this effort would be worthwhile at this stage. 2 There are broader concerns that are more conceptual or require new or different lines of research. One example of a key conceptual issue relates to the net benefits of the continued and expanding use of interventions whose ‘invention’ and initial use clearly predate the period of research investment under consideration. Should these be excluded (as we have aimed to do) from the returns to this investment? Or might it be argued that subsequent research which fails to create, or demonstrate the value of, replacement interventions for these patients is effectively research confirming the continued value and appropriateness of the interventions produced initially by earlier research funding? 3 This leads into the whole issue of time lags and time structure, which needs further consideration and research. Many previous studies have avoided this issue. We have related research over a period of years to health benefits for treatments in (or initiated in) the same number of future years with a specific lag. It is arguable that if we are trying to estimate the impact (net benefits) of research funded during a particular period (for example 1975–85) as compared to a ‘counterfactual’ that the research funding would not have happened, it may be more appropriate to measure the gains over the lifetime of the use of the interventions that were ‘invented’ during the period of funding, and these might continue to be used for a period much longer than the period of the research investment. But equally one could argue that the research underpinning those interventions may go back many years or decades to key items of past facilitating research, as might be consistent with Comroe and Dripps (1976). Once we consider that it is simply the UK’s contribution to research that we are interested in, then one might presume that most of the research would be undertaken by the rest of the world at some point, if it were not done in the UK, and that it is the timing of interventions and of their adoption that is affected by the UK’s contribution not the existence of the interventions. Such alternative conceptualisations of the counterfactual will probably have much more impact than changes to specific data or assumptions within a particular conceptualisation. We have estimated rates of return for cardiovascular disease and mental health that are of similar broad magnitudes, unlike the earlier Australian study for example. Is this error, chance, or can we assume that this is the order of magnitude in other areas? We do not believe it is error, but can certainly see how different assumptions in our analyses in one or both clinical areas could have produced different (and differing) results. But it may be chance that they are similar and we would

need to replicate the approach in a number of areas to see whether the results are more broadly applicable: the default assumption should be that they are not. It is also very important to stress what we have not done. We have estimated average returns in these two areas: we have not estimated marginal returns. This health gain analysis does not tell us what would have been the effect of spending a little more or less in the area. We may have been in the region of increasing or declining returns to marginal investment. Moreover, as the small print at the bottom of the financial investment advertisements reminds us, past performance is not an indication of future performance. Using an estimate of past returns, however accurate, does not, and cannot, provide a reliable indication of future returns. Thus research funders need to use the estimates provided in this volume with considerable care in justifying future returns from research expenditure. Finally, we have only estimated the benefit, not examined how the benefits of research could be maximised in the future. To do this requires a very different approach, using studies focusing on much more specific research or on specific positive outcomes to research, to understand how research is translated and developed and where the scope for improving this process lies.

A future research and action agenda Some of the caveats highlighted above are potentially amenable to future research. Starting from those, and our initial thoughts regarding other questions related to the value of UK public and charitable medical research, we propose that a future research agenda should include the following key areas: Research funders need to use a standardised (and mapped) way of classifying research funding. Compiling estimates of public and charitable research expenditure for the two disease areas proved to be problematic and consequently involved the use of numerous assumptions. Such estimates are unlikely to be improved by further analysis or research. In the future, if research funders wish to continue with this type of analysis, then it will be important to use a standard system of research classification and – just as importantly – to ensure that subsequent generations of systems can be mapped onto one another. The development of the UKCRC Health Research Classification System (HRCS) provides this opportunity. It would be similarly desirable to initiate collection, e.g. via surveys of companies, of data on private sector medical R&D in the UK using the same classification system.

Further research needs to be undertaken to understand the time lag between research expenditure and health gain. It is clear from the analysis in Chapter 7 that the internal rate of return to medical research is sensitive to assumptions regarding the time lag before any benefits are realised. The IRR ranges from 9% to 19% for CVD research as the time lag is changed from ten to 25 years. We used a number of indicators to estimate what an appropriate time lag may be, but stress that these are only indicators and they have a number of potentially significant flaws. Given the importance of time lag in the overall IRR calculations, we would recommend that further research efforts are focused on this area. The ‘spillover’ effects of public and charitable research expenditure on the national economy are substantial and need further, UK-focused, investigation. As demonstrated in Chapter 6, public and charitable medical research expenditure does not only lead to health gain. Such investments have a knock-on effect on the investment decisions of the private pharmaceutical and health technology sector which in turn has a positive impact on other industries and the national economy. Much of the evidence base supporting this analysis is either old (from the 1960s and 1970s) or relates specifically to agriculture research. More recent analysis for medical research is largely based on US data. The size of the USA makes it an uncertain comparator for the much smaller UK economy and research community. Future research should be commissioned to understand and provide empirical estimates of the effects of medical research for the UK economy – ideally, at a disease specific level. This has two strands: (1) direct estimation of the magnitude of the relationship in the UK between public/charitable medical research and changes to national income; and (2) estimation of the strength of the stimulus to private R&D that public/charitable research provides in the UK, and of the effect of geographic proximity/distance on the strength of this relationship. There is a need to improve our understanding of how basic research relates to clinical practice. We have seen that research funders have used different definitions to map their expenditure at different times, making comparison difficult. An additional problem arising from definitions commonly used, however, is that a great deal of more basic research work can only be linked rather loosely or arbitrarily to specific disease areas, so there may be considerable benefit from basic research to disease areas other than that to which it has been ‘allocated’. At present, the relationship between general, basic research work and clinical application is poorly defined, and future studies may benefit from a more nuanced understanding in this area. This might be explored in terms of specific case studies which could also throw light on the time lags involved.

Further research needs to be done to understand the flows of research between fields. We have assumed that the flows of knowledge and influence are the same into and out of each subject area of research, and from each area of research into the cognate treatment area. There may be research areas that contribute more than they gain from other areas – for example cardiovascular research may contribute more to cancer care than cancer research contributes to cardiovascular care. We have currently examined the overall rate of return for spending on health research in terms of health gain, not the marginal return. Any changes in health research spending are likely to take place at the margin, so a deeper understanding of the payback for health research at the margins would be valuable from a policy perspective. What scale and range of UK medical research would be most efficient? In order to help public and charitable funders of medical research maximise the benefit per pound spent, research could usefully address the question of the presence and magnitudes of (dis)economies of scale or scope in medical research. Is research better concentrated on a few targets in a few research centres, or spread across a portfolio of either or both? A deeper understanding of the international flows of knowledge and influence would be valuable. We have estimated the extent of UK research influence on UK practice using guideline citation rates and UK share of publication production, and this figure is used in the estimate of overall rate of return. However, publication-based measures are only one measure of flows of knowledge and a more nuanced understanding of these knowledge flows could improve the accuracy of our estimate. It is also true that UK research is likely to have significant health benefits beyond the UK, not least in the developing world, so our current figure underestimates the global value of UK R&D. Further research needs to be done to identify the contribution of UK public and charitable research in informing UK practice. Currently we have looked only at the proportion of UK publications cited on clinical guidelines. But we need to examine a wider range of policy documents and we need to disentangle the contribution made by publications from public and charitable funders from that made by publications funded by the private sector.

Other disease areas could be examined. In this study we selected our two examples of clinical areas for pragmatic and design reasons, to demonstrate and test our approach. Further studies could lead to further refinement and codification of the methods, and would indicate whether higher or lower returns prevailed in other areas. If the approach were to be rolled out it will be important to select other disease areas carefully. One approach would be simply to cover the major clinical areas; an alternative would be to sample clinical areas using different criteria perhaps to test hypotheses about relative rates of return. Where future research is to cover all areas, it would have the advantage that any arbitrariness in allocating research spending to clinical areas, or the impacts of research in one clinical area on treatment of patients in another disease area, would be balanced out. Research on the relationship between national clinical guidelines and national research investment. In this study we have placed quite a lot of importance on guidelines as a link between research and practice. It would be interesting and useful to look more deeply in an international study on the extent to which guidelines (and their quality, relevance, acceptance by practitioners and impact) are related to the existence of locally generated research or whether good (local) guidelines simply require access to a good world literature. The importance of local research in terms of absorptive capacity. We noted earlier the importance of absorptive capacity in relation to firms undertaking research and in our work we have for some time argued that one of the benefits of having a research capacity in the UK health care system is that it helps the system to absorb health research undertaken anywhere. Apart from one study mentioned in the annotated bibliography this remains an area primarily of speculation and further research to test this would be very valuable, particularly to the health care system and National Institute for Health Research. What are the global health benefits from UK medical research? In our study we did not attempt to measure the health gains, net of health care costs, achieved in the rest of the world as a result of UK medical research. These are potentially many times the scale of the net health gains achieved for the UK population alone. A study referring to health care practice in other major population centres and its links to UK medical research would address this.

Annotated bibliography Introduction This selective list of key literature focuses on publications that have key methodological points relevant for the assessment of the economic returns from medical research. It is not intended to be a comprehensive review of previous studies of the economic returns from health research as this was contained in the review conducted for the WHO (Buxton et al., 2004: see first publication on the list below). The bullets below outline some of the main issues covered in this current list, especially ones covered by more than one publication: • the ways in which we should value the benefits from health research • the direction of analysis when trying to establish the links between (UK) research and health gain, i.e. should it be backwards from health gain to the research, or forwards from research to health gain (and other possible benefits) • overlapping with this is the question of how far our aggregate totals should be based on building up from more disaggregated data.

Publications on the issues in general 1 Buxton M, Hanney S, Jones T. Estimating the economic value to societies of the impact of health research: a critical review. Bull WHO 2004;82:733–9. This paper contains a review of previous literature. It highlights the need to consider three steps: the inputs in terms of resources spent on research; accurately ascribing the impact of the research; and appropriately valuing the impact. Four possible ways of valuing the research are identified: direct cost savings to the health service; the human capital approach of valuing the indirect cost savings that arise when better health leads to the avoidance of lost production; assessing the intrinsic value of health gain by placing a monetary value on a life; and the gains to the economy in terms of product development and consequent employment and sales. The paper forms a starting point for the approach set out in our proposal for this project, but the proposal had some specific emphases. First, the suggested ways of valuing the benefits from health research concentrated just on the third category (the intrinsic value of the health gain) and the final category (product development etc). Second, greater importance was given in the proposal to issues such identifying the costs of the health service.

www.who.int/bulletin/volumes/82/10/733arabic.pdf

2 Sussex J (ed.). Improving Population Health in Industrialised Nations. London: OHE; 2000. Chapter 4 by Mackenbach provides a concise presentation of three key contributions (by Thomas McKeown in 1976, Johan Mackenbach and John Bunker) that use very different methods (and different directions of analysis) to assess how much of health improvement over the long term can be attributed to health care in total, as opposed to improved hygiene, nutrition, etc. It provides references to the key sources by the original authors and describes how they all come to a figure of less than 20% for the historical contribution made by health care. The same book contains a chapter (1) by Sussex and Yuen, updating the McKeown analysis. This highlights the crucial point about the significant fall in mortality from CVD since McKeown was writing. 3 Sussex J. Estimating Pharmaceutical Companies’ Value to the National Economy: Case study of the British pharma group. London: OHE; 2007. This OHE Briefing describes the economic rent method for valuing the contribution of R&D based pharmaceutical companies. 4 Mushkin S. Biomedical Research: Costs and benefits. Cambridge, Mass: Ballinger Publishing Company; 1979. Mushkin presented a contemporary challenge to McKeown’s analysis claiming his method was “deficient” and she suggested that research-based advances in technology were responsible for about 30% of the improvements in mortality from 1900 to 1975. Using the human capital approach of valuing health gains in terms of the value of production that is no longer lost due to morbidity and premature mortality, Mushkin attempted to calculate the value to the USA of all its health research. She estimated the economic value of the total reductions in mortality and morbidity between 1930 and 1975, estimated the value of the share caused by biomedical research, and after taking away the cost of the research produced a rate of return of 47%. 5 Funding First. Exceptional Returns: The economic value of America’s investment in medical research. New York: Lasker Foundation; 2000. The Exceptional Returns report summarises various contributions to show how by producing estimates for the value of a life based on willingness to pay (WTP) experiments in labour economics a very large figure can be produced for the value of the reductions in mortality in the US population, especially in the CVD field ($1.5 trillion annually from 1970 to 1990). The authors suggest that about one third of the total gain is the result of medical research that led to new drugs and treatment protocols and some of the remainder was attributed to changes in public policy and individual behaviour that depended on research. The report draws on various papers

which were subsequently written up as the book below (6) edited by Murphy and Topel.

www.laskerfoundation.org/advocacy/pdf/ exceptional.pdf

6 Murphy K, Topel R (eds). Measuring the Gains from Medical Research: An economic approach. Chicago: University of Chicago Press; 2003. The main book describing the detailed work first presented in the Funding First report, Exceptional Returns (see above). Whilst the analysis in each chapter is very rich, it is perhaps unfortunate that the introduction is not very detailed and there is no proper conclusion to pull the various contributions together. It is also noteworthy that the bibliography makes very little, if any, reference to key analyses from the past such as that by Selma Mushkin. 7 Lichtenberg F. Pharmaceutical innovation, mortality reduction and economic growth. In Murphy and Topel (2003; 6 above). pp. 74–109.

now deals with the issue of the lag between R&D spending and the health benefits. Unlike the previous study it does not provide figures for the returns for individual fields.

This chapter by Lichtenberg estimates the social rate of return on pharmaceutical investment in research (in terms of the value of additional life years generated) is around 67%. Whilst this chapter fits well into Lichtenberg’s wider stream of research on related topics it is not particularly well integrated into the rest of Murphy and Topel’s book.

8 Access Economics. Exceptional Returns: The value of investing in health R&D in Australia. Canberra; 2003. This Australian replication of the Funding First study uses the same value for a life to estimate the return on Australian health R&D. It is based on improvements in lifespan which combine reductions in specific mortality and morbidity rates for a range of illnesses. The base-case assumption is that R&D is responsible for 50% of the improvements in healthy lifespan, and that Australian R&D contributes 2.5% of the total gains, this being the percentage of global R&D undertaken in Australia. The study leaves unresolved difficulties. It uses disability-adjusted life years (DALYs) to allow for gains in mortality and morbidity; this leads to the suggestion that because there is a decline in DALYs in the mental health field, there is negative value to the measure of health ‘gains’ from mental health research. 9 Access Economics. Exceptional Returns: The value of investing in health R&D in Australia II. Canberra; 2008. This report updates the one above using new figures and a somewhat revised methodology. The value of a statistical life year has been increased to AUS$266,843 (at 2008 prices) based on a new meta-analysis of studies and the Australian contribution to world R&D outputs increased to 3.04% based on more recent bibliometric data on clinical research. Most fundamentally, the analysis

www.accesseconomics.com.au/publicationsreports/ search.php?searchfor=exceptional+returns&from=0 &search=Go

10 McGuire A, Raikou M. Inferring the Value of Medical Research to the UK. LSE Health Working Paper 5/2007. London: LSE; 2007. Broadly replicates the methodology described in Murphy and Topel to the UK. Produces a large figure of £2.6 trillion as the net gain for 1970–2000, but attributes all the improved life expectancy to medical R&D and acknowledges but does not seem to calculate either the contribution from other factors that led to improved life expectancy or the percentage of the R&D impact that comes from UK research.

www.lse.ac.uk/collections/LSEHealth/pdf/ LSEHealthworkingpaperseries/LSEHWP5.pdf

11 Johnston SC, Rootenberg JD, Katrak S, Smith WS, Elkins JS. Effects of a US National Institutes of Health programme of clinical trials on public health and costs. Lancet 2006;367:1319–27. This paper differs from the approaches used in the Funding First and Access Economics work. It adopts a hybrid approach to assess the costs and benefits of phase III randomised trials funded by the US National Institute of Neurological Disorders and Stroke. Yearly total incremental net benefits of the programmes are calculated by combining trial costs and treatment costs with a monetary value for the QALYs gained from the implementation of the trial findings. The study uses a value of a QALY based on the US GDP per capita, suggesting that this reflected “the average yearly economic productivity of a US resident, regardless of employment or age”. The study also takes into account any savings made as a result of implementing the trial findings. This work is an advance on many previous approaches but has the weakness of giving insufficient attention to other research that was reporting about the same time. 12 Cutler D, Rosen A, Vijan S. The value of medical spending in the United States, 1960–2000. NEJM 2006;355:920–7. This attempt to value medical spending in the USA 1960–2000 does not really seem explicitly to discuss the value of research. This is despite Cutler being one of the Funding First team and using somewhat similar approaches to the Funding First work, including using CVD as one of its main examples and the value of a statistical life. They claim that at least 50% of the life expectancy gains since 1950 are due to medical advances, but discuss the increased expenditure on health care and not the role of research.

http://content.nejm.org/cgi/content/ abstract/355/9/920

13 Luce B, Mauskopf J, Sloan F, Ostermann J, Paramore LC. The return on investment on health care: 1980–2000. Value in Health 2006;9:146–56. (See also the accompanying editorial: Buxton MJ. Substantial returns to health care spending: but do we spend too little or too much? Value in Health 2006;9:144–5.) The paper concludes that the value of improved health in the US population in 2000 compared with 1980 significantly outweighs the additional healthcare expenditure. (One of the four conditions for which detailed analysis is undertaken is heart attacks.) As Buxton points out in the editorial there is a danger that the returns in the form of health improvements are claimed by those estimating the returns on research and those estimating the returns to healthcare services. 14 Salter A, Martin B. The economic benefits of publicly funded basic research: a critical review. Research Policy 2001;30:509–32. This review identifies a range of benefits to an economy from publicly funded research (of any kind, not just medical). While finding that none of the included studies provided a simple and comprehensive model, it commends the progress made by Mansfield (see 15 below) in measuring the benefits from basic research. 15 Mansfield E. Academic research and industrial innovation. Research Policy 1991;20:1–12. Mansfield E. Academic research and industrial innovation: an update of empirical findings. Research Policy 1998;26:773–6. In these papers Mansfield surveyed large US corporations in seven industries for data concerning the proportion of each firm’s new products and processes that could not have been developed, without substantial delay, in the absence of recent academic research. Using figures for the value of sales of research-based products, and knowledge of the level of spending on basic research in the developed countries, he estimated a worldwide social rate of return for research conducted in 1975–78 of 28%. The pharmaceutical industry was the one most dependent on basic academic research. 16 Hanney S, Gonzalez-Block M, Buxton M, Kogan M. The utilisation of health research in policy-making: concepts, examples, and methods of assessment. Health Research Policy & Systems 2003;1:2. Reviews previous work on how health research can impact on policy, discusses possible ways in which the research contribution to policymaking can be assessed, and shows how research impact on health policy can be a key step towards achieving wider health and economic gains. Depending partly on the type of policy being made, and the type of research involved, various factors can impinge on policymaking in addition to research. These include: interests, values, established positions within institutions and personal ambition. The role that

research can play is also influenced by institutional arrangements and capacity to absorb research that can be assessed through the ‘receptors and interfaces’ model.

www.health-policy-systems.com/content/ pdf/1478-4505-1-2.pdf

17 Stoneman P. Government spending on Research and Development in the UK. Fiscal Studies 1999;20:223–59. Explores the extent of UK government spending on R&D placed in its recent historical context. Also considers patterns of expenditure, the aims behind it and the pay-off. Shows that in real terms in the ten years to 1998 the UK government reduced its expenditure, but this was mostly in defence R&D, and the percentage on health R&D increased. 18 Silverstein S, Garrison H, Heinig S. A few basic economic facts about research in the medical and related life sciences. FASEB Journal 1995;9:833–40. The authors collated data from many previous studies showing the benefits to the USA from medical research. Inevitably there is considerable inconsistency in the approaches used in the original studies: some concentrate on claimed direct cost savings to the health care system and some include indirect cost savings in the form of reductions in the level of earnings lost (i.e. human capital approach). Of the claimed $69 billion annual savings (discounted to a present level in 1995 of $1.1trn), most came from mental health ($34bn) with CVD second ($12bn). 19 Callon M, Bowker G. Is science a public good? Fifth Mullins Lecture, Virginia Polytechnic Institute, 23 March 1993. Science, Technology, & Human Values 1994;19:395–424. A complex discussion on the nature of public goods that highlights all the investment and accumulated capabilities (skills, equipment, learning processes etc.) that might be needed by others to be able to make use of knowledge. This raises doubts in practice about viewing knowledge as a pure nonrival good. 20 Belkhodja O, Amara N, Landry R, Ouimet M. The extent and organizational determinants of research utilization in Canadian Health Services Organizations. Science Communication 2007;28:377–417. Suggests that research experience is one of two factors with the greatest impact on research utilisation, especially in regional health authorities and hospitals. (The relevance of this is that it can be argued that funding research widely within a health care system helps develop the capacity of that system to absorb the findings of research from wherever they may come.)

21 Grant J, Cottrell R, Cluzeau F, Fawcett G. Evaluating “payback” on biomedical research from papers cited in clinical guidelines: applied bibliometric study. BMJ 2000;320:1107–11. As part of the examination of the papers cited in UK clinical guidelines the team analysed the location of the authors of the cited papers. 25% were based in the UK whereas only about 10% of global biomedical papers (in ISI journals) are from the UK: “the preferential citing of UK papers may provide good evidence for supporting a local science base”. 22 Jones T, Hanney S, Buxton M, Burns T. What British psychiatrists read: Questionnaire survey of journal usage among clinicians. British Journal of Psychiatry 2004;185:251–7. One of a series of studies showing that British clinicians (in this case psychiatrists), not surprisingly, disproportionately read UK-based journals to inform their clinical practice. Similar findings are reported from other countries. The series of papers also shows a natural bias towards UK research being published in UK journals. 23 UK Evaluation Forum. Medical Research: Assessing the benefits to society. London; 2006. Three bodies (The Academy Medical of Sciences, the Medical Research Council and the Wellcome Trust) came together to support the UK Evaluation Forum and this report led to them funding the current project. The report reviewed previous work and concluded: “We recommend that research funders should support research to assess the economic impact of UK medical research, which should include critiques of existing economic approaches”.

Date 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

29 30 3 36 47 52 54 62 74 93 107 113 120 124 129 138 150 163 191 202 228 251 288 298 305 309 321 316 345 368

UK health expenditure – organisations Wellcome DH Trust 2 BHF NHS (not NHS)

5 5 4 5 4 6 6 10 10 12 12 17 17 24 13 29 35 68 43 54 72 92 167 17 242 230 168 222 212 354 480 388 419

0 0 0 1 0 1 1 1 2 2 3 4 4 6 6 14 7 8 815 10 15 13 15

Cardiovascular – activity funding FCHealth

18 4 137 6 144 151 158 166 174 183 191 201 210 221 231 242 254 266 279 293 307

347 408 401 403 410 423 445 461 533 575 586

27 7 29 30 31 33 34 28 12 29 27 26 24 49 47 50 58 71 59 55 61 59 56 62 59 60 63 55 59 53 60 54 48

84 8 87 91 95 99 103 107 112 116 121 126 131 137 142 157 16 187 212 208 190 18 186 187 186 205 222 243 250 257 280 306 317 327

WT 0 0 0 0 0 0 0 0 0 0 1 0 1 1 1 1 1 1 2 1 3 5 6 17 13 11 9 13 18 30 36 26 47

39 3 3 4 5 6 6 9 10 11 11 11 11 14 14 14 15 17

BHF DH – A Total 0 0 0 0 0 1 0 1 0 1 1 5 9 0 5 9 1 5 9 1 6 11 1 6 12 2 6 15 2 6 14 3 6 18 4 6 21 4 6 22 6 6 24 6 9 26 7 9 28 8 9 33 8 10 33 10 11 39 15 11 46 13 11 47 15 32 0 13 0 11 0 9 0 13 0 18 0 30 0 36 0 26 0 47 0 0 0 0 0 0

Cardiovascular – support funding Funding DH – I Councils Total

10 10 11 11 12 13 13 14 14 15 16 17 17 18 19 20 21 22

8 8 8 9 9 9 10 10 10 11 11 12 12 13 14 17 19 19

17 18 19 20 21 22 23 24 25 26 27 28 30 31 33 37 40 41

Grand Grand total total exc. Pharma exc. pharma. industry pharma. deflated

12 16 10 20 24 31 11 38 44 54 61 72 82 92 100 111 144 171 186 213

26 27 29 31 33 37 37 42 46 48 51 55 58 65 67 76 86 88

144 132 121 116 110 102 93 99 103 103 104 108 109 113 109 115 123 121

Data provided by the MRC.

Data provided by the Wellcome Trust.

Maddock (1975) estimates MRC expenditure in 1973/74 to have been £24.8m.

Figure taken from Maddock (1975).

Data for 1972–84 was provided in two-year bands; we have taken the average for annual estimates.

Data for 1975–92 was estimated by back-casting the figures provided for 1995–2005 using an exponential function.

Data for 1975–80 based on projecting data for 1973 and 1981–2005 using a linear function.

The 1975 to 1989 time series is estimated from a back projection of the 1989 to 1992 UFC data for biomedical subjects for the UK, and the 1993 to 2005 HEFCE QR data for biomedical subjects in England. The latter is then inflated by circa 12% to make a UK-wide estimated.

10 Data for 1976 and 1977 interpolated using figures for 1975 and 1978. 11 Taken as the average of the preceding and following points. 12 Data for 1981–85 entered by hand from Annual Review of Government-Funded R&D (Cabinet Office, 1984). 13 Data for 1985–87 was estimated using an exponential function to project funding for the period 1975–84, for which we were provided with data. 14 BHF figures for 1986–88 were estimated by projecting those for 1979–85 using a linear function. 15 BHF grant funding for 1989–92 was provided for a separate project currently being undertaken by RAND Europe. 16 Numbers for these years are taken from UFC figures and are UK-based. 17 Calculated by taking the average of figures for 1992 and 1994. 18 From 1993 onwards, data is provided by HEFCE and is England-based.

Annex to Chapter Four More detailed, qualitative analysis of the importance of UK research in UK guidelines We aimed in this report to explore illustrative examples of the quantitative data in a more detailed and qualitative way. This Annex describes such a detailed approach in relation to the quantitative data described in Chapter 4 on the nationality of research referenced in guidelines. In Chapter 4 we used quantitative bibliometric analysis of the research papers in guidelines to inform the estimate of the proportion of the health gain attributable to UK research. This Annex describes a more detailed and qualitative analysis of the references in one of the NICE guidelines for which the proportion of UK references is reported in Chapter 4. Various ways of undertaking this more detailed analysis were considered. One possible approach would have been to develop a scoring approach drawing on our earlier review of ways of analysing the importance of a citation to the paper in which it was cited (Hanney et al., 2005, 2006). That approach would have combined some judgement about the importance of the specific citation to the guideline and a more detailed quantitative analysis involving how many papers were cited to support a specific point in the guidelines and counting the number of times a paper was cited in the guideline. This approach, however, was rejected as being too mechanistic. Instead, for this Annex a more wide-ranging and less mechanistic approach was adopted in relation to the hypertension guideline produced for NICE by the Royal College of Physicians – Hypertension: Management in adults in primary care; pharmacological update (National Collaborating Centre for Chronic Conditions, 2006). This guideline was selected as the most feasible on which to trial an approach to qualitative analysis. It is probably the most focused of the CVD guidelines listed in Chapter 4 and has the fewest references. It is a 2006 update of a previous 2004 guideline but it concentrates solely on pharmacological interventions. Of the 79 references in this updated guideline, the first 42 papers describe the 20 trials included in the review of pharmacological interventions. The remaining 36 references (plus one duplicate) cover a range of issues, including: trials that were excluded; and the cost and utility data used in the construction of the guideline. We first analyse the list of 20 trials (42 papers) included in the guideline and then consider the remaining papers. We use the reference numbers used in the guideline.

The papers from the 20 trials drawn on in the guideline Nine of the 20 trials have been identified as having at least some UK involvement. They are listed on Table A4.1 below in approximate order of level of UK involvement (many of them were international studies).

The dates of the relevant papers are also given. From these 20 trials the team constructing the guideline attempted to make 10 head-to-head drug comparisons, but there was only evidence from these trials for nine of the comparisons. These comparisons are important as they form a key element in the recommendations made in the guideline. Details of the comparisons to which the UK trials contributed are also given in Table A4.1. The table highlights the complexities. The main nonpharma funding bodies in the UK were probably responsible for just two studies in this guideline that is explicitly on pharmacological interventions. But in total nine out of 20 of the included trials had some UK involvement (as did 14 out of 42 papers in which they are described) but to varying degrees: • two of the 20 trials (two of the 42 papers) were conducted and funded from the UK • one trial (two papers) had a UK lead author in an international study funded by the pharmaceutical industry • one trial (two papers) had a UK co-author in a European study funded by the EU • one trial (one paper) had more authors from the UK than anywhere else but was led from Sweden and funded by the pharmaceutical industry • four of the trials (seven papers) had at least one UK co-author in an international study funded by the pharmaceutical industry. UK research played some part in six out of nine comparisons, but the two solely UK MRC-funded studies together constituted two-thirds of the data for the first head-to-head drug comparison (and the third paper had UK pharma involvement). The UK-led INSIGHT study constituted one of five studies (and less than 20% of the data) for the seventh comparison. It is also worth noting that the two solely UK MRC papers were early papers and therefore perhaps were influential on some subsequent work and of all the papers with at least one UK address; the data from CWTS show that these two were the papers cited most often (1,700 times) and fourth most often (1,154 times) respectively. It is also worth noting that the data from CWTS reported in Chapter 4 suggest that the papers on the hypertension guideline are cited much more often than the average for the field. The recommendations in the guideline drew on a detailed analysis that incorporated the head-to-head drug comparisons, cost data, utilities data etc. Therefore the recommendations are not directly referenced. It is possible, however, to see that some of the main recommendations seem to have drawn on the evidence from the UK trials. In particular, comparisons one and six seem influential in the first recommendation that “the first choice for initial therapy should be either a calciumchannel blocker or a thiazide-type diuretic” (p. 17).

Annex to Chapter Four

Table A4.1: The trials included in the hypertension guideline that have some UK involvement, and the comparisons to which they contribute Trial/papers – refs with some UK involvement

MRC (paper 34-1985)

Location of authors, in author order

Main funder

Role of the study in the head-to-head comparisons undertaken in the guideline analysis

Constituted one of three trials in the first comparison (between beta blockers [BBs] and thiazide-type diuretics). Meta-analysis showed no significant difference between two drugs in terms of mortality. This trial (but not others) found higher incidence of stoke associated with BBs. This trial and MRC-0 found BBs associated with more withdrawals.

Constituted one of three trials in the first comparison (between BBs and thiazide-type diuretics). Meta-analysis showed no significant difference between the two drugs in terms of mortality. This trial (but not others) found higher incidence of MI associated with BBs. This trial and MRC found BBs associated with more withdrawals.

Constituted one of five trials in the seventh comparison (between calcium-channel blockers [CCBs] and thiazide-type diuretics). Metaanalysis showed no difference for mortality, MI or stroke. CCBs associated with higher heart failure, but (from this and two other trials) lower diabetes.

Constituted one of three trials in the sixth comparison (between CCBs and BBs). Meta-analysis showed no difference in mortality or MI but CCBs reduced incidence of stroke. Was the sole study to show CCBs associated with reduced incidence of diabetes.

Constituted one of three trials included in the eighth comparison (between antihypertensive drug and placebo for ISH). Meta-analysis showed antihypertensive drug therapy associated with reduced incidence of stroke and MI but no significant difference in mortality rates.

Only trial (therefore Level 11 evidence) in fourth comparison (between antiotensin-11 receptor antagonists [ARBs] and CCBs). ARBs associated with higher incidence of MI. No significant difference for stroke reduction, mortality or heart failure.

Constituted one of three trials in the sixth comparison (between CCBs and BBs). Meta-analysis showed no difference in mortality or MI but CCBs associated with reduced incidence of stroke. Much smaller study than other two in sixth comparison.

MRC-0 (paper 35-1992)

INSIGHT (papers 30-2000, 31-2001)

UK, France, Netherlands, Italy, Israel, Spain

ASCOT (paper 22-2005)

Sweden, UK (most authors), Norway, USA, Iceland, Denmark, Finland, Ireland

SYST-EUR (papers 1-1991, 2-1997) [one other paper (3-1996) was on a sub-group study not including UK]

Belgium, eight other European countries including UK, plus Israel

VALUE (paper 26-2004)

USA, Norway, Switzerland, USA, UK, Italy

ELSA (paper 28-2002)

Italy, USA, Germany, Sweden, Spain, UK, France

LIFE (papers 19-2002, 20-2002 and 21-2002). Paper 16-2004 was a sub-group study. Four other papers (14, 15, 17, 18) had no UK authors

Sweden, USA, UK, Norway, Finland, Denmark

Constituted the only trial in the third comparison (between ARBs and BBs. No differences in MI, revascularisation procedures, heart failure or angina. But ARBs associated with reduced incidence of stroke and diabetes, and fewer study drug withdrawals.

HAPPHY (paper 29-1987) [but only UK address is ICI not academic]

Sweden, UK, Germany, Finland

AB Astra and ICI

Constituted one of three trials in the first comparison (between BBs and thiazide-type diuretics). Meta-analysis showed no significant difference between the two drugs in terms of mortality.

The remaining papers in the guideline Moving on from the 42 papers in the 20 included trials, a further nine papers (43–51) relate to five excluded trials (most of which had been included in the previous 2004 version of the guideline). One of these trials (papers 43, 44) was led from the USA but had one UK author and was funded by the pharmaceutical industry. Two papers (52, 53) describe sub-studies of the LIFE trial; both had one UK co-author on this US-led, Merckfunded study. Eight further papers (54–61) consider the evidence for patients’ age and ethnicity. Six were entirely UK papers (55, 56, 58–61) and another paper (57) had an incomplete reference but is linked to the ASCOT trial, which involved more UK authors even though it was

led from Sweden. Funding, from a pharmaceutical company, was acknowledged on just one of these papers (56), but presumably the others were most likely to have relied on some element of funding from UK public sources such as the authors’ hospitals or universities. Paper 60 was a review and paper 61 was itself a guideline from the British Hypertension Society published as a journal article (it appears twice, as references 61 and 63). In the Appendices four papers (62, 64–66) describe the cost data used by the Guideline Development Group to conduct their economic analysis. The NICE reference case for NICE guideline development states that “costs should be measured from an NHS and personal social services perspective”. Not surprisingly all four papers

were from the UK, with two being funded by the Health Technology Assessment Programme, which is part of the Department of Health NHS R&D Programme. Eight papers (67–74) were noted that had estimated drug-related adverse events and quality of life but none included data in a form suitable for estimation of utilities. Of these just two had UK addresses. Paper 67 was a UK-led international study funded by Hoffmann-La Roche. Paper 74 was a Belgium-led study funded by Menarini International. Finally, five further papers (75–79) provide some utility data but just one (79) was a UK study, with entirely UK addresses and funding from the MRC and the NHS.

research probably constitutes well over 10% of the research in the key trials (i.e. two out of 20 trials were solely UK, and they were described in important papers, and there was some contribution to another seven out of the 20). One of the nine head-to-head drug comparisons relied heavily on UK research and there was some contribution to five others. UK research on costs is the only research evidence used in this field and is of major importance because it provides the most relevant data for the context within which the guideline will be applied.

Comparisons with the quantitative data from CWTS As reported in Chapter 4, the quantitative data from CWTS show that 88% of the papers cited on the hypertension guideline were analysed and the proportion of UK addresses of all the authors was neither the highest nor lowest among our sample of guidelines. More detailed analysis of the data reveals that at 14.3% the proportion of UK addresses on this guideline was the second lowest. It also shows that the proportion of UK addresses, most unusually, was third rather than second, with Sweden being second to the USA with 18.5% of the addresses.

Some conclusions and implications for our analysis of the contribution made by UK research The more detailed resource-intensive analysis in this Annex is inevitably richer than the bibliometric analysis described in Chapter 4. For example, the bibliometric analysis cannot distinguish between trials included in the guideline and those trials that are listed as being excluded but nevertheless appear in the reference list. But the key question is what we can learn from this greater detail about the nature of the contribution made by UK research. • The detailed analysis highlights the complexity beneath the seemingly straightforward issue of the proportion of papers with authors with UK addresses. There are varying degrees of UK involvement in the research. A few studies are clearly fully funded by traditional UK funders and conducted solely in the UK. But more of the trials cited on this guideline, which unusually was exclusively on pharmacological interventions, are international with varying degrees of UK involvement and often funded by the pharmaceutical industry. • Chapter 4 provides various possible figures that could be used for the contribution from UK research, including: 10% (UK contribution to global research) and 17% (average proportion of UK papers on the guidelines). For this hypertension guideline UK

Annexes to Chapter Five Annex 5A: Studies used to generate QALYs gained for each intervention in each patient group Table A5A: Summary of studies used to generate QALYs gained for each intervention in each patient group Patient groups/interventions

QALYs gained

Time horizon

Treatment of acute MI: Community resuscitation

Patients with cardiac arrest of cardiac origin

No resuscitation

Up to 24 years

Naess and Steen (2004)

Assume HRQOL of 0.6 from study based on lowest value in Ward et al. (2007). 1,066 of 2,831 cardiac arrest patients survived from ROSC until hospital admission for a mean time of 532 days. Assume all would have died without resuscitation.

Patients with cardiac arrest of cardiac origin alive at arrival at hospital

Assume HRQOL of 0.6 from study based on lowest value in Ward et al. (2007). 269 of 1,066 cardiac arrest patients alive at admission survived for a mean time of 6.13 years. Assume all would have died without resuscitation.

Patients with No chest pain with thrombolytic electrocardiographic therapy signs that were typical of myocardial infarction

Patients with acute coronary syndromes

Standard therapy

Patients with acute myocardial infarction

6 months Hartwell et al. Table 18, p. 38. (2005)

The 1 year Netherlands

Vermeer et al. Table 3, ‘Quality adjusted (1988) data’ for ‘All patients’.

Cannot find studies comparing aspirin versus usual care; use same estimates as for Secondary prevention of CHD post-MI (this was also assumed in the IMPACT study). Main et al. (2004)

Table 29, p. 43.

Secondary prevention of CHD post-MI: Aspirin

Patients aged 35–84 years with coronary disease and who survived the first month with it

No treatment

Up to 25 years

Gaspoz et al. (2002)

Table 2. QALYs gained = 121,768,000 (‘current use of aspirin’) minus 115,535,000 (‘zero utilisation’) in a population of ‘about 6.8 million people’ are estimated to have CHD, and each year about 700,000 to 900,000 new cases are estimated to occur.

Patients with prior occlusive vascular events

Karnon et al. (2005)

Table 4, ‘Baseline’ analysis.

Patients discharged following MI, without absolute contraindications for beta blocker use

Phillips et al. (2000)

Table 3, QALYs gained = 42,000 (‘Current’ minus ‘Zero’ beta blocker use) (Size of single cohort = 296,613, online technical appendix 8)

Survivors of myocardial infarction with an ejection fraction <=40%

Tsevat et al. (1995)

Table 4, ‘Limited benefit model’ for patients age 60.

Patients who have had coronary events

No statin therapy

Ward et al. (2007)

Table 63, p. 101, figures for men aged 85.

Rehabilitation

Cannot find studies investigating warfarin for secondary prevention of CHD; use same estimates as for stroke. Patients who had had an acute coronary syndrome

Conventional care

Briffa et al. (2005)

p. 453, final column, third paragraph.

Annex to Chapter Five

Revascularisation: CABG surgery

Patients appropriate for CABG only

Medical management

Griffin et al. (2007)

Table 3, ‘Adjusted MD’ values in those ‘Appropriate for CABG’ only.

Patients appropriate for PCI only

Table 3, ‘Adjusted MD’ values in those ‘Appropriate for PCI’ only.

Cannot find studies comparing aspirin versus usual care; use same estimates as for secondary prevention of CHD post-MI (this was also assumed in the IMPACT study).

Main et al. (2004)

Patients with nonST-elevation acute coronary syndromes

No use of GPAs

Palmer et al. (2005)

Table 2. Strategy 1 versus strategy 4.

Treatment of chronic stable angina: Aspirin

Patients with CHD

Table 63, p.101, figures for men aged 85

Patients with arrhythmia

Amiodarone treatment

Buxton et al. (2006)

Table 85, p.106, ‘UK average patient’

Persons with symptomatic heart failure and left ventricular ejection fractions <=35%

Belgium, Canada and the USA

Glick et al. (1995)

Table 1, Within-trial model, Total.

Patients with chronic heart failure

Varney (2001)

Assume HRQOL of 0.6 from study based on lowest value in Ward et al. (2007). This is multiplied by LYG under the ‘Limited benefits’ scenario (0.228 years; p. 368 section 3.1). Note that this assumes no morbidity benefit.

Patients with severe heart failure and a left ventricular ejection fraction of <=35%

Standard therapy + placebo

16 countries 35 months

Glick et al. (2002)

Table 1, Total.

Cannot find studies for statins in this patient group; use same estimates as for secondary prevention of CHD post-MI (this was also assumed in the IMPACT study).

Patients with heart Optimal UK failure with a marker for pharmaceutical cardiac dyssynchrony therapy and left ventricular systolic dysfunction

Fox et al. (2007)

Table 58, p. 67, ‘Mixed’

Table 64, p. 83, ‘Mixed’.

Treatment of arrhythmia: ICD Treatment of heart failure:

Patients with endstage heart failure

Clegg et al. (2005)

Cannot find specific figures for heart transplant; see Table 52, p. 104 and use figures for Medical group, assuming 7.2 months pretransplant life expectancy for no transplant.

Treatment of acute stroke: Aspirin

Cannot find studies comparing aspirin versus usual care; use same estimates as for secondary prevention of stroke.

Patients aged 18 years and above with acute stroke

USA and Canada

Samsa et al. (2002)

Table 3, ‘Trial+long term’.

Patients with acute ischaemic stroke

Standard care

Sandercock et al. (2002)

Table 10, p. 76, ‘Base case’.

Patients with acute stroke

Contemporary conventional care

Chambers et al. (1998)

Patients with acute stroke (but not subarachnoid haemorrhage)

Do not scan anyone

Wardlaw et al. (2004)

Table 3, Comparator minus S12.

Secondary prevention of stroke: Rehabilitation therapy

Cannot find studies for rehabilitation in this patient group; use same estimates as for Secondary prevention of CHD post-MI

Patients aged 70 who had survived an initial stroke and who were suitable for treatment with an antiplatelet therapy

No treatment strategy

Beard et al. (2004)

Cannot find studies for statins in this patient group; use same estimates as for secondary prevention of CHD post-MI. 40–80 year old men and women after their first idiopathic venous thromboembolic event or pulmonary embolism

3 month conventional therapy with warfarin (standard therapy after stroke)

Aujesky et al. (2005)

Figures for men aged 40 years, 6 months conventional therapy versus 3 month conventional therapy (standard therapy after stroke). Cannot find studies in this patient group; use same estimates as for beta blockers for secondary prevention of CHD post-MI.

Primary prevention of CVD: Treatment of hypertension

65 year old women with an annual CVD risk of 2%, HF risk of 1% and diabetes risk of 1.1%

Thiazide-type diuretics

National Table 3, p. 13, figures for Collaborating women, C versus D. Centre for Chronic Conditions (2006)

Treatment of hypercholesterolaemia

Patients aged 55 years without CHD

Davies et al. (2007)

Figures for women receiving PRA (least effective).

70 yr old patients with AF at moderate risk of stroke

O'Brien et al. (2005)

Table 2, base case.

Patients aged 50–60 years with no known CVD

Annemans et al. (2006)

Table 5, risk level = 5%.

55–64 year old

Not quitting smoking

Wang et al. (2008)

Table 14, p. 38, use 0.99 value (most conservative).

Primary prevention of stroke:

Smoking cessation: Quitting smoking

CHD = coronary heart disease, MI = myocardial infarction, ICD = implantable cardioverter defibrillator, CRT-P = cardiac resynchronisation therapy device, CRT-D = CRT device plus ICD, CVD = cardiovascular disease.

Annex 5B: Estimating the numbers of eligible patients in each patient group Table A5B delineates the data sources and assumptions used to generate the numbers of eligible patients in each patient group over the period 1985–2005. Table A5B: Data and assumptions used to generate the numbers of eligible patients Patient groups/interventions

Data and assumptions

Treatment of acute MI

Data for 1998–2005 obtained from Hospital Episodes Statistics website (www.hesonline.nhs.uk/Ease/ servlet/ContentServer?siteID=1937&categoryID=192) based on the total number of emergency admissions for primary diagnosis I21 (acute myocardial infarction). These are figures for England; UK figures are computed by multiplying the numbers of admissions in England by the ratio of UK population to the English population. Figures for 1985–87 were computed by applying the ratio of admissions to UK population size to the population size in 1985–97. This applies to all treatments of acute MI except for community resuscitation, hospital resuscitation and primary angioplasty.

Community resuscitation and hospital resuscitation

As for treatment of acute MI, except the analysis is based on the numbers of emergency admissions for primary diagnosis I46 (cardiac arrest).

Data on numbers receiving surgery for 2005 obtained from the Myocardial Infarction National Audit Project (MINAP Steering Group, 2007). Prior to 2001 the numbers receiving surgery was zero. We assumed a constant rate of increase between 2001 and 2005.

Secondary prevention of CHD post-MI

Data for 2003 obtained from the Health Survey for England (www.dh.gov.uk/en/Publicationsandstatistics/ Publications/PublicationsStatistics/DH_4098712), based on the proportion who reported having ever had a heart attack. This proportion was then applied to UK population figures. In earlier and later years the same proportion was applied to the UK population figures in each year.

Revascularisation CABG surgery

Data on numbers receiving surgery for 1985–99 obtained from Society of Cardiothoracic Surgery in Great Britain and Northern Ireland website (www.scts.org/). From 2000 to 2005 the number each year was held constant at the 1999 value.

Data on numbers receiving surgery for 1991 to 2005 obtained from British Cardiovascular Intervention Society website (www.bcis.org.uk/). Assume increase from 1991 to 1992 applies to each preceding year.

Treatment of unstable angina

As for treatment of acute MI, except the analysis is based on the numbers of emergency admissions for primary diagnosis I20 (angina pectoris).

Treatment of chronic stable angina

As for secondary prevention of CHD post-MI, except the analysis is based on the proportion who reported having ever had angina.

Treatment of arrhythmia ICD

Data on total numbers of ICDs for 1985–2005 obtained from report for European Heart Rhythm Association. (www.ic.nhs.uk/webfiles/Services/NCASP/audits%20and%20reports/ReportforEuropeanHeartRhythmAsso ciation-AnnualReport2005.pdf)

Treatment of heart failure

Data for 1994–98 obtained from Key Health Statistics from General Practice (National Statistics, 2000). These are figures for England; UK figures are computed by multiplying the prevalence in England by the UK population. For 1985–93 we assume the rate is the same as in 1994. For 1999–2005 we assume the rate is the same as in 1998. This applies to all treatments for heart failure except for CRT-P and CRT-D.

CRT-P and CRT-D

Data on total numbers of CRT-Ps and CRT-Ds for 1985–2005 obtained from Report for European Heart Rhythm Association.

Heart transplant

As for treatment of acute MI, except the analysis is based on the numbers of emergency admissions for main procedure K02 (other transplantation of heart).

Treatment of acute stroke

As for treatment of acute MI, except the analysis is based on the numbers of emergency admissions for primary diagnoses I60–I69.

Secondary prevention of stroke

As for secondary prevention of CHD post-MI, except the analysis is based on the proportion who reported having ever had stroke.

Primary prevention of CVD

Data for 1985–2005 based on the UK population (Office for National Statistics, 2007).

Primary prevention of stroke

As for heart failure, except the analysis is based on the numbers of people with atrial fibrillation.

Smoking cessation

Data for 1985–2005 from the General Household Survey (www.statistics.gov.uk/ssd/surveys/general_ household_survey.asp) on the number of smokers. Data on the numbers of people who died from smoking each year were taken from Peto et al. (2006). The numbers of quitters in year x+1 was calculated as the number of smokers in year x minus the number of smokers in year x+1 minus the number of people who died from smoking in year x+1.

Annex 5C: Adjusting for overlapping patient groups We calculate the numbers of patients in each patient group who are eligible for treatment. It is likely that in some cases patients feature in more than one patient group, in which case the total number of eligible patients will be overestimated. To account for this we adjusted for overlapping patent groups using the assumptions outlined in Table A5C, which are based on the IMPACT study.

Table A5C: Assumptions used to adjust for overlapping users Patient group

Revascularisation

1. Assume that 20% of angioplasty patients go on to have CABG surgery; hence multiply angioplasty patients by a factor of 0.80. 2. Subtract numbers of primary angioplasty patients from numbers receiving angioplasty.

1. Subtract numbers of patients being treated for unstable angina. 2. Subtract 50% of patients receiving CABG surgery. 3. Subtract 50% of patients receiving secondary prevention of CHD post-MI.

1. Subtract 50% of cases (since assume these are due to CHD and are considered elsewhere). 2. Subtract heart transplant patients.

1. Subtract those receiving antihypertensive drugs elsewhere in the model from those receiving antihypertensive drugs as primary prevention of CVD. 2. Subtract those receiving lipid-lowering drugs elsewhere in the model from those receiving lipid-lowering drugs as primary prevention of CVD.

Annex 5D: Estimating the uptake rates for specific interventions in each patient group

Table A5D: Data and assumptions used to generate uptake rates Patient groups/interventions

Figures for 1994–95 from Norris (1998), Table 2. Assume constant after 1995. According to Tunstall-Pedoe et al. (1992) cardiopulmonary resuscitation was “largely neglected” before 1982. Assume zero in 1985 and linearly interpolate to ` 1994 value.

Figures for 1994–95 from UKHAS (1998), p. 117 column 1. Assume constant after 1995. According to Tunstall-Pedoe et al. (1992) cardiopulmonary resuscitation was “largely neglected” before 1982. Assume zero in 1985 and linearly interpolate to 1994 values.

Figures for 2001 and 2005 from MINAP Steering Group (2007), Table 1. Linearly interpolate for 2002–04. Thrombolysis used since 1950s; assume constant rate at 2001 value.

Figures for 1994–98 from Key Health Statistics from General Practice. Aspirin introduced for reducing MI in 1988 by FDA (www.bayeraspirin.com/pain/asp_history.htm); assume zero up to this point and linearly interpolate after this point. Assume constant after 1998.

According to UKMI (2002) clopidogrel use is predicted to be 30–50% of acute coronary syndrome cases within five years of introduction in this indication (in 2002). Assume 0% use in 2001.

Have actual numbers of cases – see data for eligible patients (assume all cases are new cases).

Figures for 1994–98 from Key Health Statistics from General Practice. Figures for 2000 from EUROASPIRE II (2001), Table 9. Figures for 2001 and 2005 from MINAP Steering Group (2007). Aspirin introduced for reducing MI in 1988 by FDA (www.bayeraspirin.com/pain/asp_history.htm); assume zero up to this point and linearly interpolate up to 1994. Linearly interpolate in 1999.

According to Main et al. (2004), p. 6 in 2001 aspirin = 91% of antiplatelet prescriptions and clopidogrel = 4%. According to UKMI (2002) clopidogrel use in 2001 was double the value in 2000. Assume zero use before 2000 and constant 4% use relative to aspirin’s 91% use after 2001.

Figures for 2000 from EUROASPIRE II (2001), Table 9. Beta blockers first introduced in 1984 (www.medscape.com/ viewarticle/542689_2) hence assume zero for 1985 and linearly interpolate to 2000. Assume constant rate after 2000.

Figures for 2000 from EUROASPIRE II (2001), Table 9. Landmark clinical study published in 1988 (Swedberg and Kjekshus) hence assume zero uptake before then and linearly interpolate to 2000. Assume constant rate after 2000.

Figures for 1994–98 from Key Health Statistics from General Practice. Assume no use prior to 1990 (Department of Health, 2007) and linearly interpolate to 1994. From Department of Health (2007) assume constant increase after 1998.

Figures for 2000 from EUROASPIRE II (2001), Table 9. Anticoagulants used since 1950s; assume constant rate at 2000 value.

Figures for 2000 from EUROASPIRE II (2001), Table 5. Landmark clinical study published in 1983 hence assume zero uptake at 1985 and linearly interpolate to 2000. After 2000 assume constant rate at 2000 value.

Have actual numbers of cases – see data for eligible patients (assume all cases are new cases)

Figures for 2000 from Capewell et al. (2006), Table 1. Landmark clinical study published in hence assume zero uptake before then and linearly interpolate to 2000. After 2000 assume constant rate at 2000 value.

Figures for 1994–98 from Key Health Statistics from General Practice. Assume no use prior to 1990 (Department of Health, 2007) and linearly interpolate 1994. From Department of Health (2007) assume constant increase after 1998.

Treatment of arrhythmia: ICD

Treatment of heart failure: ACE inhibitors

Figures for 1994–98 from Key Health Statistics from General Practice. Landmark clinical study published in 1988 (Swedberg and Kjekshus) hence assume zero uptake before then and linearly interpolate to 1994. Assume constant rate after 1998.

Figures for 1994–98 from Key Health Statistics from General Practice. Beta blockers first introduced in 1984 (www.medscape.com/viewarticle/542689_2) hence assume zero for 1985 and linearly interpolate to 1994. Assume constant after 1998.

Figures for 1994–98 from Key Health Statistics from General Practice. Diuretics first introduced in 1984 (www.medscape. com/viewarticle/542689_2) hence assume zero for 1985 and linearly interpolate to 1994. Assume constant after 1998.

Assume same as for CHD, but constant after 1998.

Figures for 1994–98 from Key Health Statistics from General Practice based on use in patients with atrial fibrillation. Aspirin introduced for reducing MI in 1988 by FDA (www.bayeraspirin.com/pain/asp_history.htm); assume zero up to this point and linearly interpolate after this point. Assume constant after 1998.

Figures for 1994–98 from Key Health Statistics from General Practice based on use in patients with atrial fibrillation. Anticoagulants used since 1950s; assume constant rate prior to 1994 at 1994 value and after 1998 use constant 1998 value.

Figures for 2001 and 2004 from CEEU (2007), p. 79. Landmark clinical study published in 1993 hence assume zero uptake before then and linearly interpolate to 2001. Linearly interpolate between 2001 and 2004. Assume 2004 value in 2005.

Figures for 2001 and 2004 from CEEU (2007), p. 79. Landmark clinical study published in 1984 hence assume zero uptake in 1985 and linearly interpolate to 2001. Linearly interpolate between 2001 and 2004. Assume 2004 value in 2005.

Figures for 2001 and 2004 from CEEU (2007), p. 79. Landmark clinical study published in 1978 hence assume zero uptake in 1985 and linearly interpolate to 2001. Linearly interpolate between 2001 and 2004. Assume 2004 value in 2005.

Figures for 2001 and 2004 from CEEU (2007), p. 79. Landmark clinical study published in 1983 hence assume zero uptake in 1985 and linearly interpolate to 2001. Linearly interpolate between 2001 and 2004. Assume 2004 value in 2005.

Figures for 1994–98 from Key Health Statistics from General Practice based on use in patients with atrial fibrillation. Warfarin used since 1950s; assume constant rate prior to 1994 at 1994 value and after 1998 use constant 1998 value.

CEEU (2007) p. 19: in 2006, on discharge, 70% of patients were on antihypertensive medication (assume for 2005). Beta blockers first introduced in 1984 – assume zero for 1985. www.medscape.com/viewarticle/542689_2

Figures for whole population for 1994–98 from Key Health Statistics from General Practice. Beta blockers first introduced in 1984 (www.medscape.com/viewarticle/542689_2) hence assume zero for 1985 and linearly interpolate to 1994. Assume constant after 1998. Assume constant rate of increase after 1998.

Figures for whole population for 1994–98 from Key Health Statistics from General Practice. As for statins assume no use prior to 1990 (Department of Health, 2007) and linearly interpolate 1994. From Department of Health (2007) assume constant increase after 1998.

Primary prevention of stroke: Warfarin

Smoking cessation: Smoking cessation

Adjust actual numbers of smokers by proportion who had contact with medical services who told respondents to quit smoking for medical reasons, obtained from Health Survey for England (www.dh.gov.uk/en/Publicationsandstatistics/ Publications/PublicationsStatistics/DH_4098712; variable name: drsmoke); this variable is available for 1993–99 and 2003–05 (range of values: 0.22–0.25). Applied 1993 value to 1985–92 and applied 2003 value to 2000–02. Then, multiply the result by the proportion of deaths due to cardiovascular disease published by Peto et al. (2006). These estimates were available for the year 2000 and were estimated to be 0.27 (we applied this proportion to every year).

Annex 5E: Adjusting for compliance We searched the economic evaluation studies used to obtain the QALYs gained estimates to ascertain whether or not non-compliance had already been accounted for in the estimates generated by the cost-effectiveness models used. In some of these non-compliance was accounted for in which case it was not considered further because its effects were already included in the estimates of the QALYs gained. Also, for some interventions we were able to find data directly on numbers of users, thereby already accounting for compliance. Where it was not accounted for we applied compliance rates used in the IMPACT study. These were:

• 100% in hospital patients • 50% in asymptomatic community patients • 75% in symptomatic community. We assumed that compliance rates were constant over time. Table A5E summarises the assumptions used.

Table A5E: Compliance rates used in the analysis Patient groups/interventions

Compliance rate (%)

Assumed to be 100% based on IMPACT study

Compliance included in QALY estimates; rate set to 100%

Assumed to be 75% based on IMPACT study

Have numbers of new users directly; rate set to 100%

Revascularisation:

Treatment of unstable angina:

Rehabilitation therapy

Compliance included QALY estimates; rate set to 100%

Treatment of hypertension

Assumed to be 50% based on IMPACT study

Have numbers of quitters directly; rate set to 100%

Heart transplant: Heart transplant Treatment of acute stroke:

Secondary prevention of stroke:

Primary prevention of CVD:

Annex 5F: Adjusting for polytreatment We adjust for polytreatment because patients in a particular patient group may receive more than one intervention and this is not reflected in the numbers of new users. Ideally we would have data on the numbers of patients receiving every combination of the interventions for each patient group, but invariably we do not have these data. Hence, we need to account for the use of multiple interventions or we may overestimate the QALYs gained.

we then move down the list and identify the intervention with the next highest QALYs gained and allocate this to the new users who did not receive the intervention with the highest QALYs gained identified in the previous step

we repeat this process until all new users have been allocated a QALY gain

the above process is then repeated for every year.

We adjust for polytreatment in three ways:

This option is used to generate the lower estimates in our analysis.

1 we assume there is no polytreatment 2 we assume there is maximum polytreatment and that the QALYs gained from each intervention are additive 3 we assume there is maximum polytreatment and that the QALYs gained from each intervention are not at all additive. Suppose that according to our calculations there are 1,225,500 patients in the secondary prevention of CHD post-MI group who are new users of statins over the period 1985–2005 and 1,737,500 new users of aspirin. With no polytreatment (option 1) we assume that the two sets of new users do not overlap and none of the new users receiving statins is a new user of aspirin, and vice versa. With maximum polytreatment (options 2 and 3) we assume that the 1,225,500 new users of statins are a subset of the total new users of aspirin and that there are no new users of statins who are not new users of aspirin. With option 2 we assume that the QALYs gained from each intervention are additive. Therefore the 1,225,500 new users of statins (who are also new users of aspirin) receive the QALYs gained from statins plus the QALYs gained from aspirin. The 512,000 new users of aspirin who are not new users of statins receive only the QALY gains associated with aspirin. Note that options 1 and 2 yield the same estimates of the total QALYs gained because in each case these are computed by multiplying the QALYs gained for each intervention by the numbers of new users and summing the totals for each specific intervention to generate a total for the patient group. This is method is used to generate the central estimates in our analysis. With option 3 we assume as with option 2 that there is maximum polytreatment, but the effects of the different interventions are not additive. In this case we allocate the highest QALY gains to each patient from the different treatments they receive. We calculate the total QALYs gained as follows: 1

in each year, we order the specific interventions for each patient group in increasing order of the numbers of new users

moving down the ordered list of specific interventions we identify the intervention with the highest QALYs gained and allocate these to all those who received that intervention (plus, assuming maximum polytreatment, all other new users falling above in the ordering)

In order to ensure that the total incremental costs are commensurate with the total QALYs gained for all three options, the total incremental costs are allocated on the same basis as the total QALYs gained. Hence, the above approaches are used to generate the central and lower estimates of the total incremental costs.

Annex 5G: Estimates of total QALYs gained by year used to generate Figure 5.1 Table A5G: Total QALYs gained by year QALYs gained (000s) 1986

Annex 5H: Studies used to generate incremental costs for each intervention in each patient group Table A5H: Summary of studies used to generate incremental costs for each intervention in each patient group Raw incremental costs

Cost base year

Gage et al. (2002)

Not the same study used to obtain QALYs gained; UK study

Cannot find studies for thrombolysis in this patient group; use same estimates as for tPA in acute stroke

Cannot find studies for aspirin in this patient group; use same estimates as for secondary prevention of stroke

Table 29, p. 43

Hartwell et al. (2005)

Table 18, p. 38

Table 4, ‘Baseline’ analysis

Cannot find studies for beta blockers in this patient group; use same estimates as for treatment of heart failure

Patient groups/interventions

Incremental costs (2005 UK£) Notes

Table 4, ‘Limited benefit model’ for patients age 60

Table 63, p. 101, figures for men aged 85

Cannot find studies for warfarin in this patient group; use same estimates as for secondary prevention of stroke

Jolly et al. (2007)

Not the same study used to obtain QALYs gained; UK study, Appendix 2

Table 3, ‘Adjusted MD’ values in those ‘Appropriate for CABG’ only

Table 3, ‘Adjusted MD’ values in those ‘Appropriate for PCI’ only

Cannot find studies for Aspirin in this patient group; use same estimates as for Secondary prevention of stroke

Table 2, strategy 1 versus strategy 4

Table 85, p. 106, ‘UK average patient’

16 countries 35 months (US$)

Cannot find studies for ACE inhibitors in this patient group; use same estimates as for secondary prevention of CHD post-MI

Cannot find studies for statins in this patient group; use same estimates as for secondary prevention of CHD post-MI

Table 64, p. 83, ‘Mixed’

Cannot find specific figures for heart transplant; see Table 52, p.104 and use figures for Medical group, assuming £1806 pre-transplant costs for no transplant.

USA and Canada (US$)

Table 3, ‘Trial+long term’

Table 9, p. 70, ‘Base case’. Note that costs are measured over 1 year – this is conservative since lifetime costs are more negative

Table 3, Comparator minus S12

Cannot find studies for rehabilitation in this patient group; use same estimates as for secondary prevention of CHD post-MI

Figures for men aged 40 years, 6 months conventional therapy versus 3 months conventional therapy (standard therapy after stroke)

Cannot find studies in this patient group; use same estimates as for beta blockers treatment of heart failure

National Collaborating Centre for Chronic Conditions (2006)

Table 3, p. 13, figures for women, C versus D

Figures for women receiving PRA (least effective)

Table 2, base case

Table 16, p. 39, use 54.88 value; this is associated with a success rate at 12 months of 5.86% and produces highest cost overall given success rate at 12 months; therefore compute total costs as 54.88*New users*1/0.0586

CHD = coronary heart disease, MI = myocardial infarction, ICD = Implantable cardioverter defibrillator, CRT-P = Cardiac Resynchronisation Therapy device, CRT-D = CRT device plus ICD, CVD = cardiovascular disease.

Annex 5I: Estimates of total incremental costs by year used to generate Figure 5.2 Table A5I: Total incremental costs by year Incremental costs (ÂŁ million) 1986

Annex 5J: Sensitivity analysis We undertook a sensitivity analysis to generate upper and lower estimates of the total QALYs gained and incremental costs around our central estimates. These were calculated as follows:

so by the medical services. Hence, to generate an upper estimate of the total QALYs gained we assumed that 100% of people who quit smoking did so as a result of medical research.

• Upper estimate: smoking cessation is a relatively important component of the total QALYs gained. We assumed in the central estimate that 22–25% of people who quit smoking did so as a result of medical research, based on the numbers of quitters who were told by medical services to quit smoking for medical reasons. This may underestimate the contribution of medical research because some people may quit smoking for medical reasons without being told to do

• Lower estimate: we calculated the lower estimate of the total QALYs gained by assuming there was maximum polytreatment and that the QALYs gained from each specific intervention are not at all additive (see Annex 5F for a more detailed description of the method). The same approach was used to generate upper and lower estimates around the central estimates of the total incremental costs.

Table A5J: Results of sensitivity analysis, 1985–2005 Total QALYs gained (000s)

Total monetised QALYs gained (£ million)

Total incremental cost (£ million)

Non-smoking related activities

Total QALYs gained (000s)

Treatment of arrhythmia:

Treatment of heart failure:

Heart transplant:

Early diagnosis and treatment 14.2

Central estimate:

Upper estimate: Non-smoking related activities

Lower estimate: Patient groups/interventions Treatment of acute MI:

Secondary prevention of CHD post-MI:

Smoking cessation:

Quitting smoking

Annex to Chapter Six Literature review on R&D spillovers 1. Introduction In this Annex we focus solely on the non-health benefits of medical research and development (R&D). It is extremely important to consider non-health as well as health gains. Returns to investment in medical research by one organisation may benefit not only that organisation but also other organisations in the medical sector, in other sectors, and also in other countries (although this last is outside the scope of the current research project). That is, there are spillovers. In this paper, we present the findings of our review of the research spillovers literature in general and as applied specifically to medical research. The structure of this Annex is as follows. In Section 2 we outline the methodology used for the literature review. Section 3 describes the concepts of knowledge and information and goes on to discuss the differences between private and social rates of return to R&D, to introduce the concept of spillovers. Section 4 summarises the literature analysing the sources of spillovers. Section 5 discusses the main transmission mechanisms for spillovers identified by the literature. In Section 6 we comment on the literature on the economics of the geography of innovation, highlighting how spillovers have helped develop this strand of work. Section 7 summarises the main results.

2. Methodology We have adopted a selective approach for the literature search, mainly driven by: • our accumulated knowledge and experience on the topic • core references from past reviews of the value of the pharmaceutical industry, which have also helped us in identifying key authors in the fields. This is a strategy suitable for methodological reviews where conventional keyword based search strategy (such as R&D) may result in a very large number of references – too large to be handled within the time constraints of this project. Nevertheless, we searched the following databases: British Library Integrated Catalogue – search on Serials/ Periodicals; economic working papers database – EconPapers; EconLit; and PubMed. We looked for papers that combined the following keywords (or nearest equivalents): ‘medical’; ‘research’ (or ‘R&D’ or ‘research and development’ or ‘medical research’); ‘spillovers’ (or ‘externalities’ or ‘synergies’); ‘social rate of return’ (or ‘rate of return’). Part of the remit was to explore the factors attracting private R&D to the UK, including public medical research, so we also carried out a literature review combining the following keywords: ‘location’; ‘research’ (or ‘R&D’ or ’research and development’ or ‘medical research’); ‘companies’ (or ’firms’ or ‘enterprises’ or ‘business’ or ‘economic activity’ or ‘manufacturing’ or

‘clinical trials’); ‘productivity’; ‘competitiveness’; ‘nations’ (or ‘countries’ or ‘regions’). As an illustration of the large number of hits obtained with the different databases, the search criteria (‘medical’ and ‘research’ and ‘spillovers’ or ‘externalities’ or ‘synergies’) yielded 100, 829 and 345 hits in the economic working papers database (EconPapers), PubMed and the British Library Integrated Catalogue respectively. The references listed in previously known and identified literature also provided an additional indirect route to the ‘grey literature’, which is information produced at all levels of government, academia, business and industry in formats not controlled by formal publishing, monographs and books. Overall, we have identified and reviewed 139 papers/ reports.

3. Measuring rates of return from R&D Our starting point is defining the basic characteristics of ‘knowledge’. 3.1 Knowledge: basic characteristics Arrow (1962) suggests that knowledge is inherently a public good. He also argues that knowledge differs from the typical factors of production (labour, capital) in two ways. First, it is non-excludable, non-rivalrous or non-exhaustible. Thus, knowledge developed for any particular application can easily spill over and have economic value in very different applications. This leads to what has become known as the appropriability problem for any organisation generating new information. This problem may diminish firms’ incentives to invest in R&D. Second, compared with other factors of production knowledge has a greater degree of uncertainty, a higher extent of asymmetries and a greater cost of transacting new ideas. It is also useful to distinguish between tacit and codified knowledge. Tacit knowledge can be difficult to write down in such a way that is meaningful and readily understood. It needs face-to-face and even nonverbal communication as well as reciprocity, all of which may be ineffective or infeasible over longer distances. It has also been referred to as “sticky knowledge” (von Hipple, 1984). Codified knowledge, on the other hand, can be written down – it is more akin to information than to tacit knowledge. It is better structured and less ambiguous. The cost of transfer of codifiable knowledge is lower than for tacit knowledge and is not bounded by close proximity of the source of knowledge. This distinction is critical in giving rise to the concept of ‘localised knowledge spillovers’, as discussed later.

Annex to Chapter Six

3.2 Private versus social rate of return from R&D Early attempts to identify the link between national R&D spending and economic performance employ macroeconomic models of growth which integrate R&D stock into the traditional production function as an input – treated as a residual factor accounting for growth in earlier works (Solow, 1957) and as a source of endogenous growth in more recent models (Romer, 1986; Lucas, 1988). Empirical estimation of the impact of R&D stock on economic performance has proved to be problematic due to difficulties in: • measuring R&D stock and economic output • developing reliable econometric models using aggregate data which do not allow for variations across sectors and firms.

based on the ‘economic rent’ that the companies earn for the UK. Specifically, they estimate the net additional income brought to the UK by these companies’ activities in excess of the income they would be expected to generate in the next best alternative use(s) to which labour and capital would be diverted, if, hypothetically, these companies ceased to operate in the UK. One of the key elements generating the economic rent is the spillovers generated by the R&D investment carried out by these companies in the UK. The underlying framework behind these authors’ methodology for estimating the value of R&D spillovers in particular is illustrated in Figure A6.1. Here it is applied to pharmaceutical R&D in particular – but it could be applied to R&D carried out in any sector.

Figure A6.1 Spillovers from private pharmaceutical R&D

In general, the main limitation of these models is that they work as a ‘black box’ producing financial measures of social return to R&D investment, but unable to explain how R&D spending and innovation generate economic development. Three main methodological approaches have been used to assess the value and benefits from research: econometric studies, surveys and case studies. The first method relies on the analysis of large databases; surveys have been conducted of R&D managers; while case studies attempt to trace all the antecedents to an innovation. Each method has its own advantages and disadvantages. In this Annex we draw upon the evidence generated by all three methods. At the more general level, it is possible to distinguish between two types of return: • private or direct return to investment, meaning the economic benefits generated by a specific R&D project and accrued by the organisation originally involved, through royalties and/or sales of a new product or process • social or indirect return to investment, meaning economic and non-economic benefits spilling over for third parties to exploit, e.g. new knowledge and economic conditions that stimulate and enhance innovation and technical progress. The difference between the social and the private rate of return represents R&D spillovers arising from innovative initiatives or projects undertaken by a public or private organisation and not captured by the originator. The starting point for our work on estimating the magnitude of the impact of R&D spillovers from public and charitable medical research in the UK is Garau and Sussex (2007), who calculated the value of two major British based, research intensive, pharmaceutical companies to the UK’s economy. The estimation is a practical application of an economic methodology

As shown in Figure A6.1, R&D investment undertaken in one company not only generates benefits to itself (the ‘private’ return – shown in purple above) but also could bring about economic benefits to other companies and organisations operating in the same industry (shown in blue), and to organisations in other sectors altogether (shown in green). The sum of all three ‘rings’ represents the total social return to the investment. Garau and Sussex (2007) estimate, among other things, the spillovers generated by the (private) pharmaceutical R&D carried out by two companies in the UK. The range of estimates is based on empirical economic literature looking at the rate of return to private R&D. The total social rate of return is found to be around 51%, of which only 14% is captured by the investing firm, 26% is captured by other firms in the same sector, and 11% is captured in other (non-pharmaceutical) sectors of the UK economy. As highlighted by the authors, this estimate is highly uncertain, however, owing to the wide range of values attributed to the social rate of return to R&D in different studies. The work carried out in Garau and Sussex (2007) is based on the methodology described in the joint Association of the British Pharmaceutical Industry (ABPI), Department of Health, Department of Trade and Industry (DTI) and HM Treasury study undertaken for the UK’s Prime Minister’s Pharmaceutical Industry

Competitiveness Task Force (PICTF) in 2000 and published in the December 2001 PICTF report Value of the Pharmaceutical Industry to the UK Economy (PICTF, 2001). The PICTF report presented an estimate of the economic value added by the pharmaceutical industry as a whole to the UK economy.

4. Sources of spillovers Before discussing in detail the sources of R&D spillovers identified by the literature, we present the conceptual framework that forms the basis for the qualitative and quantitative analysis carried out on this topic. 4.1 Conceptual framework It is extremely important to consider wider economic gains as well as health gains when assessing the economic benefits of medical research. These economic gains are additional to the health gains: they refer to the income for UK residents that is generated by public medical research investment. This is distinct from the monetary value of health (QALY) gains that were estimated in Chapter 5. Investment in medical research by one organisation, public or private, may benefit not only that organisation but also other organisations in the medical sector, in other sectors, and also in other countries: i.e. ‘spillovers’. Spillovers should not be viewed as accidental: they can be, and are, a deliberate policy objective of spending on public research. Our interest lies in spillovers generated by public medical R&D. Given the link between R&D and knowledge, we do not distinguish here between R&D spillovers and knowledge spillovers but refer to both taken together. For illustrative purposes, we take a simplified view of the inputs and outputs of the R&D process. ‘Public research’ refers to R&D carried out or funded by public and charitable organisations, and includes university research. Private R&D is carried out by privately owned enterprises. These inputs, either in isolation or in combination, lead to some ‘output’: new products, new patents or better performance (measured in a number of ways) by firms. These outputs generate additional GDP, i.e. income, and economic rent for the residents of the UK. As a preliminary step to quantify the spillovers generated by the R&D analysed in this work (see Chapter 6 of this report), we conducted a literature review on the topic of spillovers. The review was originally intended to cover only public medical research. That literature was relatively scarce, however, so we extended the search to cover public and private R&D in general, without focusing on medical research in particular. The literature on R&D spillovers, as a general concept, is relatively abundant. However, the literature is not clear about defining and outlining the conceptual framework

within which spillovers exist and transmit. The aim of this section is to combine and synthesise the existing literature to present a (simplified) conceptual framework that will help the reader understand better what spillovers are and how they work in practice. This section has been divided into three main elements, each discussing in turn the three sources of spillovers described below (sections 4.2, 4.3 and 4.4 of this Annex respectively). We distinguish between public and private R&D as sources of spillovers. The first ‘source’ refers to the degree of complementarity between public and private R&D, and, in particular, whether and how much private R&D is generated as a result of public R&D. Section 4.2 of this Annex relies mainly on the literature on biomedical R&D, given the distinct nature of the R&D process for pharmaceuticals. An additional interesting result identified from this literature is that the relationship between public and private pharmaceutical R&D is bi-directional, i.e. private R&D helps generate public R&D, as well as the other way around, and it is the interaction between the public and private research sectors that helps to generate and develop new technologies. However, and as discussed in more detail later, the quantitative evidence available looks primarily at the first effect, i.e. how much private R&D is done thanks to public R&D. This strand of the literature looks at the relationship at an aggregate level, rather than looking at individual organisations. Another part of the literature looking at the public– private research relationship explores how new firm start-ups and the location decisions of new firms are affected by university and other public research. This literature looks at other sectors other than the biomedical sector. Several authors argue that the decision of new firm start-ups where to locate is influenced by traditional regional characteristics (such as size and concentration of population) as well as by the opportunity to access knowledge generated by universities and other public laboratories. The main differences between these two strands of the literature (other than the sectors analysed) is that the former literature is more aggregated in that it does not explore how individual firms/agents respond. The latter literature looks at decisions taken by individual organisations (in this case, firms) as a result of public R&D. In addition, the latter can be described as the stimulation of creating capacity to do R&D (via new firms), while the former is more of a stimulus of R&D itself. The second source of spillovers is public research (see Section 4.3 of this Annex). The literature here primarily focuses on university research, which we take as ‘public’ R&D. The literature for this section focuses on a micro level, in that it explores the reactions and decisions taken by individual organisations, such as individual firms, in response to public research.

Figure A6.2 Sources of spillovers: public and charitable R&D and its interaction with private R&D

Private R&D is the third source of spillovers identified by the literature (see Section 4.4). Figure A6.2 represents graphically public research as a source of spillovers and its complementarity with private R&D. Sections 4.2 to 4.4 in effect describe Figure A6.2 in more detail, referring back to the three sources of spillovers mentioned above. 4.2 Complementarity between public and private R&D From our literature review we have grouped a selection of papers into two broad approaches to the analysis of the relationship between public and private R&D. On the one hand, we have grouped together the literature looking at the biomedical sector. Then there is another strand of the literature that explores the links between public research (in particular, university research) and the decision to locate new firm start-ups in all sectors, not just the biomedical sector. We take these two strands of the literature in turn.

4.2.1 The biomedical sector

Within the biomedical sector, public sector research plays an important role in the discovery of new drugs and other health care technologies, but the reality of the interaction between the public and private sectors is much more complex than a simple basic/applied dichotomy would suggest. As some authors have argued, the conventional picture of public research as providing a straightforward input of basic knowledge to downstream, applied private research may be quite misleading. Thus, the relationship between the public and private sectors in the pharmaceutical industry is not well described as a ‘cascade model’. Referring to Figure A6.2, the arrows headed ‘A1’ and ‘A2’ illustrate the two-way relationship between public and private R&D. Arrow ‘A1’ illustrates the fact that some private R&D takes place in the UK thanks to public R&D in the UK. However, some private R&D would take place even if all public R&D activities were to be eliminated. Moreover, and as the literature suggests, we need to take into account the public sector R&D generated due to the private R&D – the arrow labelled ‘A2’. Note that arrow ‘A2’ stems from both ‘new’ and ‘existing’ private

R&D; i.e. there might be a feedback loop back to public R&D from the private R&D that has been generated in the first place from public R&D. The literature has focused primarily on estimating quantitatively how much private R&D is generated by publicly funded R&D. We found two particularly relevant published empirical studies of the relationship between publicly funded and privately funded R&D in the pharmaceutical industry: Ward and Dranove (1995) and Toole (2007). Both studies refer to publicly funded medical research in the USA, by the National Institutes of Health (NIH), and to the impact on the sum of R&D expenditure in the USA by all pharmaceutical companies and R&D expenditure by US-based pharmaceutical companies worldwide. A great deal of caution must be exercised when considering the relevance of these studies to the question of the impact of UK public and charitable medical research on UK private sector R&D. The scales of both publicly funded medical research and private sector R&D activity in the pharmaceutical and other health industries are several times greater in the USA than in the UK. The opportunities for interactions between the two are therefore much greater in the USA than in the UK. However we were unable to find empirical data on the public/private R&D linkages for the UK or for an economy closer to it in scale. Ward and Dranove (1995) used annual data for the period 1966–88. Over that period NIH-funded R&D grew at an average rate in money of the day terms by 8–11% p.a. depending on therapeutic category. Pharmaceutical industry R&D (in money of the day) grew over the same period at an average rate of 11–19% p.a. depending on therapeutic category. From these data Ward and Dranove estimated that a 1% increase in publicly (NIH) funded basic research expenditure in the USA in a particular therapeutic category would, after a lag of seven years, cause a 0.76% increase in private industry R&D spend in that same therapeutic category and a 1.71% increase in private industry R&D spend in other therapeutic categories. Thus a 1% increase in NIH spend across all therapeutic areas leads to a 2.5% increase in total private pharmaceutical R&D spend by members of the US trade association, Pharmaceutical Manufacturers of America. Industry data cover “both domestic and overseas R&D by US manufacturers” but “only US-based R&D for foreign manufacturers” (p. 76). Ward and Dranove did not attempt to estimate whether there was any impact of public medical research on non-pharmaceutical sector private R&D. The recent studies published by Toole, based on US data, show that basic research supported by government and public agencies, mainly undertaken in university and non-profit laboratories, stimulates and supports private investment on R&D in the pharmaceutical and biotech sector. Toole (2007) concluded that public medical research complements, rather than crowds out, private pharmaceutical industry R&D investment. Unlike Ward and Dranove, Toole

distinguishes different strengths of impact for NIH funded basic laboratory research and for the clinical human research they fund. But, unlike Ward and Dranove, Toole does not distinguish the impact of public research spend on private R&D spend within the same therapeutic category from that on private R&D spend in different therapeutic categories. Like Ward and Dranove, Toole also did not attempt to estimate the impact of public medical research on non-pharmaceutical sector private R&D. Toole used data for the period 1981–96 for NIH spend disaggregated into seven therapeutic categories, and for 1981–97 for US pharmaceutical industry R&D spend defined as US and worldwide spending by US companies and spending in the USA by non-US companies (i.e. the same definition used by Ward and Dranove, and again from the trade association of the pharmaceutical industry in the US, now named Pharmaceutical Research and Manufacturers of America, PhRMA). Over the 1981–96 period, NIH expenditure was growing at an average rate of around 3% p.a. in real terms, and pharmaceutical industry R&D grew faster than NIH in most therapeutic categories. The study is based on an empirical model where the level of private investment is a function, among other things, of basic scientific knowledge generated by public research. Toole found that a 1% increase in NIH expenditure on basic research leads to a 1.69% increase in pharmaceutical industry R&D after a lag of eight years. Toole also found a U-shaped pattern of response: with private R&D increasing in years 1, 2, 7 and 8 after the increase in NIH spend, but with no great response in years 3–6. The response to public clinical research is shown to be smaller than for basic research but achieved sooner. Toole estimates that a 1% increase in NIH expenditure on clinical research leads to a 0.40% increase in pharmaceutical R&D, and that this is achieved within three years. Moreover, Toole’s results indicate that “a dollar increase in public basic research stimulates an additional $8.38 in pharmaceutical investment after eight years” (Toole, 2007). Similar arguments have been presented by the US Congressional Budget Office (CBO, 2006): “It is seldom possible to identify particular cases in which the private sector would have performed research if the government had not. Thus, most of the available empirical evidence is based on aggregate studies. On balance, that evidence suggests a positive relationship between public and private pharmaceutical R&D” (p. 31). For the UK in particular, recent data published by the Association of the British Pharmaceutical Industry (ABPI) shows the extensive collaborative research links between the pharmaceutical industry and the UK university science base. The ABPI figures, gained from a survey of 11 major pharmaceutical companies operating in the UK, show that 606 PhD studentships and 327 postdoctoral grants were conducted in collaboration with 78 British universities in 2007 (ABPI, 2008). However, the figures are down from 2003 – by

nearly 14% on PhD studentships and almost 25% for postdoctoral grants. According to the ABPI, there are three main types of collaboration between industry and universities: PhD studentships, where students carry out research projects jointly between a university and a company; postdoctoral grants, where jointly funded research programmes are undertaken between companies and universities, including exchanges of personnel; and industrial placements, where undergraduate students work within companies for usually one year as part of their degree studies. Crespi and Geuna (2004) focus on the impact of higher education sector R&D spending on science productivity (not specifically on the pharmaceutical sector) and attempt to estimate any crowding-in or crowding-out effect of different sources of funding. Because of the need for more robust country level data, however, they do not find any significant evidence of this phenomenon. 4.2.2 New firm start-ups

As mentioned above, the second strand of the literature looking at the public–private R&D relationship explores how new firm start-ups and the location decisions of new firms are affected by the presence of university research. Several authors argue that the decision of new firm start-ups where to locate is influenced by traditional regional characteristics (such as size and concentration of population) as well as by the opportunity to access knowledge generated by universities. For instance, Audretsch and Lehmann (2004a) and Audretsch et al. (2003) show, based on German data, that university spillovers play an important role and have a strong influence in shaping strategic locational decisions of firms. Two main results can be highlighted from this work. First, the number of entrepreneurial start-ups is greater in those regions with a greater presence of knowledge inputs. Second, the ‘type’ of knowledge generated and the mechanism by which the knowledge is transmitted are important. The authors make a distinction at two levels – between research carried out in the social sciences and natural sciences areas, and between human capital and publications – which relates back to whether the knowledge is tacit or codifiable. For those university outputs and spillover mechanisms that are more tacit in nature (social sciences and human capital), geographic proximity plays a greater role in accessing and absorbing university spillovers. As a conclusion, the authors argue that new firm start-ups’ decisions to locate are influenced by traditional regional characteristics as well as by the opportunity to access knowledge generated by universities. But the impact of university output on new firm location depends on both the type of knowledge and mechanism used to access that knowledge. Mansfield (1995) distinguishes between basic and applied research, and argues that firms tend to trade off faculty quality for geographical proximity, particularly in the case of applied research. For basic research,

firms seem to pay less attention to location in choosing universities to work with and support, perhaps because in many kinds of applied R&D it is very useful for academics and firm personnel to interact and work together face-to-face, whereas in basic research ties may be weaker and more sporadic. This relates back to the distinction between tacit and codified knowledge – the former being probably more akin to applied research and the latter to basic research. 4.2.3 The impact of proximity on innovative activity

A recent article explores the links between university research and business innovation in the UK in particular for a number of sectors, including pharmaceuticals. Abramovski et al. (2007) examine whether firms are locating R&D facilities close to top university departments in the UK, and conclude that there is a positive correlation between the location of R&D-performing establishments and the presence of high quality relevant university research departments. They also claim that in some industries, private-sector R&D labs are disproportionately clustered around highly rated university research departments. These authors also found that the clustering of R&D facilities close to university departments is particularly strong in the pharmaceuticals and chemicals sector. For example, their results suggest that a postcode area with a university chemistry department rated 5 or 5* for its research is likely to have around twice as many labs doing R&D in pharmaceuticals and around three times as many foreign-owned pharmaceuticals R&D labs compared with a postcode area with no 5 or 5* rated chemistry departments. Interestingly, the authors also suggest that their results do not provide direct evidence on how firms indeed connect and link with universities – which is what they plan to examine in more detail in their future research. Martin and Tang (2007) reinforce the previous results in that researchers and students can spin out from universities to exploit new ideas and technologies, by establishing start-up companies. This in turn transfers skills, tacit knowledge and the other benefits of university know-how mentioned above into the commercial environment. These authors cite three examples illustrating how universities can stimulate regional and firm growth: Route 128 in Boston (MIT), Silicon Valley in California (Stanford), and the development of new ‘high technology’ firms around Cambridge University in the UK. The references found therein again reinforce the relation between new firms and their interactions with universities – with a particular focus on the biotechnology sector. According to a UK report by the then Minister for Science (Department of Trade and Industry, 1999), the biotech sector in the UK (broadly defined as the sector focusing on new technologies, including pharmaceuticals) offers an example of the development of clusters and factors encouraging them. The UK has a strong research base spread across a number of

regions supported by the presence of world leading research institutes such as the Sanger Institute and the Roslin Institute. To these areas of research strength correspond areas of concentration of biotechnology companies in East Anglia (Cambridge), South East England (Oxfordshire and Surrey) and Central Scotland. The quantitative analysis presented in Furman (2003) is based on panel data reflecting activities of more than 30 pharmaceutical firms over the period 1984–94. The analysis suggests that lab-level scientific orientation does indeed vary by region and is positively related to the strength of scientific and technical bases in the region geographically proximate to the focal laboratory. Furman proposes that the extent to which firms incorporate science in their drug discovery efforts is correlated with the strength of the scientific base in the local geographic regions in which they operate. The hypothesis is that regions that offer extensive scientific resources, such as universities, government laboratories, or a collection of private, science-oriented firms, will be more likely to generate science-oriented firms than will those with more limited scientific assets. 4.2.4 Why are public and private research complements?

The large majority of the literature and empirical evidence we found supports the view that public/charitable medical research, overall, stimulates additional private R&D rather than crowding it out. There is a much smaller and empirically less well supported literature that takes a contrary view – see David and Hall (2000), and Kealey and Al-Ubaydl (2000). Beyond the empirical estimates, it is important to understand why an increase in public expenditure on research might have a positive effect on private R&D in the biopharmaceutical sector. In other words, what are the key characteristics of R&D for health care technologies that make it so important for the public sector to intervene though specific policy mechanisms? What is the role of public funding supporting biomedical basic research? The development of innovative technology in the pharmaceutical and biotech sector is characterised by: • a high level of uncertainty due to a significant scientific challenge at early stage (basic research and preclinical) and recurrent risk of failure at clinical phases • large investment required compared to other sectors – the average out-of-pocket costs of a new product approval is $400 million (DiMasi et al., 2003), whilst in other highly innovative sectors, such as IT and communications technologies, an investment of £4 million allows companies to bring new products to the market (Cooksey, 2006).

Investors in the pharmaceutical and biotech sectors (shareholders and venture capitalists, respectively) need to assess the expected returns on their investments, which is particularly problematic at the basic research stage where the outcomes of research efforts are very uncertain, and are keen to generate returns as quickly as possible. Venture capitalists are key funders of earlystage biotech firms but have a time horizon of three years for a particular investment, compared to 10–12 years required by companies to develop and market successful health care products (Pisano, 2006). Against this background, various reports commissioned by government authorities have attempted to illustrate and explain why publicly funded medical research can stimulate and complement private investment on R&D in the pharmaceutical and biotech sectors. Factors cited as important to private companies include: • Potential collaborations with research centres and universities supported by government funding. These are associated with “cost-sharing and risk reduction opportunities” (NERA, 2007; CRA, 2004). For example, public resources can be used to finance fixed capital costs (e.g. laboratories and other infrastructures) and private funds can cover variable costs of research projects (Crespi and Geuna, 2004). • Returns may be subsidised by public investment on research and clinical trials. The NIH in the USA is often cited as an example of government funds effectively complementing venture capitalists’ investments in the biotech sector. We have found no quantitative evidence on the magnitude of arrow A2 in Figure A6.2, although as argued by Cockburn and Henderson (1996, 1998, 2000) the private sector also invests heavily in basic research, viewing it as fundamental to the maintenance of a productive research effort. As argued by these authors, public and private sector scientists meet as scientific equals, solve problems together and regard each other as scientific peers, which is reflected in extensive coauthoring of research papers between the public and private sectors. For instance, Cockburn and Henderson (1998) show there is extensive co-authoring of research papers between the public and private sectors, and that there is some evidence that this co-authoring activity is correlated with private sector productivity. They conclude that the private sector results can have importance for public sector work. 4.3 Public research as a source of spillovers 4.3.1 Literature review: social rate of return to public research

Empirical literature provides estimates of the social return to public investment in agricultural research but not medical research. Table A6.1 summarises the main findings.

Table A6.1: Estimates of social rate of return to public R&D in the agricultural sector Study

Knutson & Tweeten (1976)

Sources: Griliches (1991), Salter et al. (2000).

Most of the works estimate the rate of return to publicly funded research (in the agricultural sector) to be in the range of 20–67%. The main limitations of these studies are that they may be biased towards those government R&D programmes that proved successful, and the preponderance of US data which may not translate directly to the situation of the UK. Nevertheless, they are indicative of the existence of a large and positive economic contribution of public research to the national economy as a whole with a social rate of return of at least 20% and probably higher. As noted by Buxton et al. (2004), and based on the review by Salter and Martin (2001), none of the studies measuring the benefits to an economy from publicly funded research provides a simple and comprehensive model. More recent contributions on the issue of measuring the rate of return of publicly funded research include the studies conducted by Mansfield, which focus on the contribution of research conducted by academic centres to innovation delivered by commercial sectors (arrow B2 in Figure A6.2). Mansfield (1991) concludes that over 10% of new products and processes marketed by surveyed firms could not have been developed (without substantial delay) in the absence of academic research, and that this proportion is as high as 27–31% in the pharmaceutical sector specifically. Mansfield provides an estimate of the social rate of return on academic research of 28%. The results of Mansfield’s later (1998) study are similar with respect to the impact of academic research on commercial innovation but provide no rate of return estimate. There is wide agreement about the importance of universities in generating economic growth. As argued by Nelson (1986), universities are recognised generators and repositories of public knowledge. Also, centres of commercial innovation and entrepreneurship are linked to proximity to universities. Hughes (2006) identifies four kinds of interaction that work at the university–industry interface: (1) basic university role of educating people, including providing suitably qualified human capital for the business sector; (2) research activity and the role it plays in increasing the stock of codified knowledge

which may have useful or commercial elements; (3) problem-solving in relation to specifically articulated business needs; (4) ‘public space’ functions. The last of these includes a wide range of interaction mechanisms between university staff and the business community: informal social interactions, specially convened meetings, conferences, centres to promote entrepreneurship and entrepreneurship activities, and the exchange of personnel including the role of internships. These may lead to the transfer of both codified and tacit knowledge and the establishment of relationships which may feed back into the other three kinds of interactions. Martin and Tang (2007) refer to many studies which, they argue, suggest that the recruitment of skilled graduates represents the most important mechanism through which firms derive economic benefits from basic research. 4.3.2 Public research spillovers: effects on innovation

The literature identifies two potential effects of university research spillovers on respectively: (1) innovation (patents/new product innovations); (2) performance/ growth of firms. Arrow B1 in Figure A6.2 represents the results of the literature that examines the use of the (modified) ‘knowledge production function’. The knowledge production function explores the relationship between knowledge inputs, such as public and/or private R&D, and (innovative) outputs, such as patents and new products. This strand of economic literature provides some evidence of the importance of geographically mediated commercial spillovers from university research, highlighting the positive effects from locating close to the sources of public research, in this case, universities. Indeed, university research has been found to have a positive and significant effect on both patents and new product innovations. Arrow B1 thus represents one ‘direct’ spillover effect of public R&D that is not generated via the private sector. Arrow B2 represents the literature that argues that public R&D, in the form of university research, has a positive impact on firms’ performance and growth. A number of papers suggest that geographic proximity and university spillovers are complementary determinants of firm performance. A combination of both factors (but not alone) results in significantly higher stock market performance. This effect depends on the type of university output. As illustrated in Figure A6.2, arrow B2 shows potentially (at least) two channels for this university spillover. One channel implies that the firm’s performance improves because public R&D improves (somehow) the productivity of the existing R&D carried out by existing firms, which in turn leads to better performance (B2A). The other channel by which public R&D can improve firms’ performance is more direct, without influencing ‘existing’ R&D (B2B). Unfortunately, the literature leaves relatively undefined

the exact mechanism by which this occurs, which is why arrow B2 has two ‘branches’. The difference between arrows B1 and B2B (both go directly from ‘public and charitable research’ to ‘outputs’ without going through ‘private R&D’) is the ‘effect’ of the public research: while the literature that gives rise to arrow B1 focuses on the relationship between public research and innovation (measured as new patents or new products), the literature underlying arrow B2B explores the link between public research and firms’ performance. For the first effect, as illustrated by arrow B1, Jaffe (1989a) uses a modified ‘knowledge production function’, where the dependent variable is state-level (US) corporate patents (a proxy for useful knowledge). The two inputs are private R&D and university research. The results provide some evidence of the importance of geographically mediated commercial spillovers from university research, which turns out to be strongest for the subsegment classified as ‘drugs’. The overall elasticity of corporate patents with respect to university R&D is around 0.1, meaning that a 1% increase in university R&D expenditure is associated with a 0.1% increase in corporate patents. Jaffe also models simultaneously industry R&D and university research. He finds an indirect or inducement effect, via increasing private R&D; and this effect is even larger. While we have included this result within the literature that supports the existence of B1 (direct effect of public research), this result could also be used to corroborate arrow A1. Indeed, given Jaffe’s elasticity of industry R&D with respect to total university research (0.704) and the elasticity of corporate patents with respect to corporate research (0.814), the implied elasticity of induced corporate patents with respect to university research is almost 0.6 (0.704 × 0.814). This author also finds evidence, albeit weak, of the impact of co-location of universities and research labs. Acs et al. (1992) replicate this analysis using actual product innovations from the USA, and find even higher elasticities and stronger support for co-location. Based on a survey of companies, Nelson (1986) also supports these results, as he claims university research is positively and significantly related to the R&D intensity of the industry in question. Arundel and Geuna (2001) use the PACE survey of Europe’s largest industrial firms to test, among other things, the importance of proximity in the transfer of knowledge from publicly funded research organisations (PROs), which includes universities, to firms. Their descriptive results show that PROs are the most important external source of knowledge for firms’ innovation. Also, their results show firms use a variety of methods to acquire different types of knowledge from PROs, including some that provide access to codified knowledge (e.g. reading publications or attending conferences), and methods that provide the opportunity to access non-codified knowledge, such as informal

personal contacts, joint research, and hiring trained scientists and engineers. Link and Rees (1990) compare university-based research relationships between small and large firms in an effort to identify one factor that might explain the noted difference in innovativeness between firms of different sizes. The underlying hypothesis of this work is that there are diseconomies of scale in large firms owing to the fact that bureaucratisation in the innovation decision making process inhibits not only inventiveness but also slows the pace at which new inventions move through the corporate system towards the market. Link and Rees argue that although large firms are more active in university-based research per se, small firms appear better able than large firms to utilise their universitybased associations to leverage their internal R&D. Gambardella (1992) offers empirical evidence for the pharmaceutical sector that in-house scientific research raises the ability of firms to take advantage of public science. Using evidence from case studies, he argues that firms with better in-house scientific research programmes have more effectively exploited outside scientific information. He also finds, based on data from the then 14 largest US-based pharmaceutical companies, that company patents are positively correlated with the scientific publications of the firms even after controlling for the scale of R&D. For new medicines in particular, Toole (2000) examines the impact of public research on pharmaceutical innovation using a production function approach, linking new molecular entities (NMEs) approved by the US Food and Drugs Administration (FDA) with the stock of public basic research (number of awards granted by the NIH) and private pharmaceutical R&D. Public research has a positive and significant impact on discovery of NMEs: elasticity estimates are in the range of 2–2.4 implying that a 1% increase in the stock of basic research leads to an average 2–2.4% increase in the number of approved NMEs. The results also show there is a lag of 17 to 19 years between investment on public research and new products approval. 4.3.3 Public research spillovers: effects on firms’ performance and growth

For the impact on firms’ performance and growth (arrow B2), and based on the German dataset used in Audretsch and Lehmann (2004a) and Audretsch et al. (2003), the empirical evidence in Audretsch et al. (2003) and Audretsch and Lehmann (2004b, 2006) suggests geographic proximity and university spillovers are complementary determinants of firm performance. A combination of both factors (but not alone) results in significantly higher stock market performance. This effect, as with the decisions to locate and start a new firm, also depends on university output. If the spillover involves knowledge in the natural sciences, geographic

proximity is less important; if in the social sciences, geographic proximity is seen to be a necessary condition for generating abnormal profits. Moreover, the smaller the distance from the nearest university and the higher the number of academic papers published, the higher the growth rate of firms. The Audretsch papers, however, leave unclear the mechanism by which university research ultimately improves firm performance. However, as highlighted by Jaffe (1989), public research has an indirect effect on corporate (i.e. private) patents, by increasing private R&D. Thus, this evidence might support more the existence of arrow B2A rather than B2B. 4.4 Spillovers from private R&D 4.4.1 Literature review: social versus private rate of return to private R&D

Table A6.2 summarises the findings of the empirical literature on the total economic returns – i.e. the ‘social returns’ – to private R&D spending. The social return is typically around 50% and greatly exceeds the private return captured by the investing organisation (typically around 20%) in every case. The difference between the social and private returns is the return captured by firms, organisations or individuals other than the firm that made the original investment. The empirical studies listed in Table A6.2 use various approaches to estimate the productivity growth at the industry and inter-industry level generated by R&D efforts, including cost function approaches (Bernstein and Nadiri, 1988; Bernstein and Nadiri, 1991), total factor productivity (TFP) (Griffith et al., 2004a) and production functions using patents as a measure of firms’ output (Jaffe, 1986; Scherer, 1982, 1984). Each study considers different sets of industries and data relating to US companies or OECD countries. Some of the references included in Table A6.2 are reviews (Garau and Sussex, 2007; Nadiri, 1993; PICTF, 2001). We conclude that the social rate of return to private sector R&D tends to be around 50% but could be significantly higher. The variability of the figures is partly due to the lack of a widely accepted theoretical model explaining accurately the complexity of the innovative process and to a paucity of reliable data providing robust support to the research question. In particular, some studies measure benefits induced by R&D spending in terms of output of new products and process, not taking into account that investment on R&D also contributes significantly to building absorptive capacity and accumulating capability to solve complex problems. With regard to the available data, one of the main limitations is related to a high degree of aggregation not accounting for inter-industry or interfirm differences. In terms of the literature measuring the spillovers from private R&D in particular, we have identified three

Table A6.2: Social return to private R&D Study

potential effects of private R&D spillovers: (1) productivity of other firms’ R&D; (2) entry decisions of potential competitors; (3) reduction of production costs. 4.4.2 Productivity of other firms’ R&D

Jaffe (1989b) carries out an analysis of firms’ patents over patent classes. He defines classes or technological groups, and measures a pool of spillovers. This pool is the sum of all other firms’ R&D, weighted by technological proximities. He finds that a firm’s R&D productivity is increased by the R&D of technological neighbours, though neighbours’ R&D lowers profits and market value of low R&D intensity firms. Moreover, if everyone increased their R&D by 10%, total patents would increase by 20%, with more than half the increase coming from the spillover effect; and if everyone increased their R&D by 10%, then aggregate profits would increase by about 3%, with about one third of the net increase coming from the spillovers. Frantzen (2000) performs a cointegration analysis on annual panel data with respect to a set of manufacturing industries in a series of OECD countries over the period 1972–91. Results show that not only own R&D capital, but also domestic and international R&D knowledge spillovers, as well as human capital, play an important role in explaining the evolution of manufacturing productivity in OECD countries. The spillover effects were shown to be both intersectoral and intrasectoral in nature and there was evidence that they are especially strong in research intensive industries. Similar results, based on the same data, are obtained in Frantzen (2002a, 2002b). Guellac and de la Potterie (2001) investigate the longterm effects of various types of R&D on multifactor productivity growth (MFP), which is argued by the authors to be the spillover effect of R&D. Econometric estimates are conducted on a panel of 16 OECD

countries, over the period 1980–98. These authors argue that business R&D has a positive and significant impact on MFP, indicating that there are substantial spillovers from business R&D – the return to the economy as a whole is larger than the private return. Also, the impact of business R&D on MFP is larger in countries where R&D intensity is higher, and the effect of government and university performed research on productivity is positive and significant. Park (2004), based on pooled time-series data from 14 OECD economies and three East Asian economies for the period 1980–95, allows for the simultaneous presence of international and intersectoral (manufacturing to non-manufacturing) R&D spillovers. He finds an asymmetry in intersectoral R&D spillovers: while manufacturing R&D has a strong intersectoral R&D spillover effect on non-manufacturing TFP, the reverse is not true. Park argues that this is because the manufacturing sector provides relatively more technology-embodied intermediate goods for the nonmanufacturing sector. Cassiman and Veugelers (2002), based on Belgian data in the 1993 Community Innovation Survey (CIS), show that incoming spillovers have a positive and significant effect on the probability of firms cooperating. Cooperating firms, because of the improved technological competence of the partners, better tap the existing base of know-how. Firms with better appropriability have a higher probability of cooperating in R&D. Moreover, firms that find the publicly available pool of knowledge more important for their innovation process are more likely to benefit from cooperative agreements with other research institutes. A number of articles analyse the spillover effects specifically in pharmaceutical R&D. Cockburn and Henderson (1994) use detailed data on R&D investments and outcomes at the level of individual research programmes conducted within ten pharmaceutical firms over a period of 17 years. They find that competitors’ research appears to be a complementary activity to own R&D: rivals’ R&D results are positively correlated with own research productivity. The authors interpret this as evidence of significant spillovers of knowledge across firms, rather than the depletion externality implied by ‘winner takes all’ models. Henderson and Cockburn (1996) focus on the ‘knowledge production function’. Knowledge is measured as grants of ‘important’ patents, and the explanatory variables include two measures of ‘spillovers’: (1) count of competitors’ output in the same narrow research area; (2) count of competitors’ output in all the other programmes in the wider therapeutic class. Results, based again on programme-level data, show that at the mean, programmes whose competitors’ programmes are in the same or related fields that are roughly 10% more productive, will be approximately 2% more productive themselves.

Henderson and Cockburn (1996) argue that the pharmaceutical industry is characterised by high rates of publication in the open scientific literature. They carried out an extensive interview programme with scientists and researchers. They note that many of the scientists interviewed stressed the importance of keeping in touch with the science conducted both within the public sector and by their competitors. Nearly all of the interviewees, according to the authors, had a quite accurate idea of the nature of the research currently being conducted by competitors, and they often described ways in which their rivals’ discoveries had been instrumental in shaping their own research. Moreover, these authors suggest that there may have been a concomitant change in the role of spillovers in shaping research productivity in the pharmaceutical industry. Some decades ago, when research in the pharmaceutical industry was done via screening compounds, there was little to be learned from others unless they found a particularly promising molecule. Now, however, pharmaceutical companies actively invest in the generation of new physiological and biochemical knowledge, so even knowledge of others’ false starts and failures may help to shape one’s own research programme. Thus, the impact of interfirm spillovers on research productivity in the pharmaceutical industry may have increased over time. Henderson and Cockburn argue that before 1978 the benefits of spillovers were realised primarily within narrow therapeutic areas, while after 1979 the opposite appears to be true. The authors argue that after 1979 programmes benefited primarily from work conducted by their competitors in related therapeutic areas, rather than research focused on the same disease targets. As the authors conclude, “modern theory ascribes a central role to interfirm spillovers as drivers of economic growth, so it is reassuring to find them present in research intensive industry such as pharmaceuticals” (p. 56). A more recent paper explores the relationship between R&D productivity and the potential for spillovers provided by local research to distributed research laboratories, drawing on a sample of multinational pharmaceutical firms in the 1980s (Furman et al., 2006). The authors first analyse the productivity of a firm’s research effort at therapeutic class level, rather than looking at the whole firm’s research efforts. This allows for more direct comparisons between firms’ research programmes as well as within-firm productivity differences across therapeutic areas that may be driven by local spillovers. Furman and colleagues also test for the importance of local spillovers by constructing local measures of knowledge generation directly, rather than inferring their impact. Furthermore, they distinguish between privately and publicly generated spillovers. The authors find that spillovers exist and are significant. Patent output at the therapeutic class level is positively correlated with a firm’s ‘exposure’ to papers related to that therapeutic

class authored by scientists based within 35 miles of where the firm conducts research. One of the novelties of Furman et al. (2006) is that the authors distinguish between privately and publicly authored papers (by institutional affiliation of the author). This distinction yields a striking result: patent output is positively and significantly correlated with publicly authored work, but negatively and significantly correlated with private sector work. Thus, this work somewhat contradicts the previous analyses suggesting that private R&D was a source of spillovers for other firms not directly carrying out that R&D. The authors do argue, however, that it is a reduced form result, so probably the result summarises a mix of factors. For instance, it may be that extensive rival publication in a particular area reduces the opportunity for any single firm to make novel discoveries that are a prerequisite to patent generation. Alternatively, extensive rival publication may signal significant future competition in a particular market, thus reducing the attractiveness of dedicating effort in the area and the incentive to generate patents after research efforts have already been expended. As suggested by the authors, these hypotheses are just a handful of many credible explanations that merit further research. 4.4.3 Entry decisions of potential competitors

In terms of the second spillover from private R&D, entry decisions of potential competitors, Aharonson et al. (2006) use detailed data on Canadian biotechnology entrants in the 1990s to explain the dichotomous variable of whether or not to enter in a postcode area in a given year. One of the explanatory variables is the intensity of incumbents’ inventive activity, measured as R&D spending and number of R&D employees. The authors argue that new entrants are influenced systematically by factors promoting the benefits of colocation, and seek locations that would allow them to benefit positively from knowledge spillovers. Stuart and Sorenson (2003) look for an explanation for firm co-location in high-tech industries. Their results suggest that areas with large populations of biotech and venture capitalist firms do enjoy a ‘regional advantage’ and that such areas experience the highest rates of biotechnology entrepreneurship. Feldman (2003) also looks at the locational dynamics of the US biotech industry. This author argues that the biotech industry is becoming more geographically concentrated and highly specialised in certain locations. There might be several reasons for this, but one put forward by the author is that “existing firms serve as anchors that attract skilled labour pools, specialised intermediate industries and provide knowledge spillovers that benefit new technology intensive firms in the region” (p. 312).

4.4.4 Reduction of production costs

For the third spillover effect of private R&D on reduction of costs, Levin and Reiss (1984) estimate that a 1% increase in the R&D spillover causes average cost to decline by 0.05%. Bernstein and Nadiri (1989) show that average cost declines (on average) by 0.2% in response to a 1% growth in R&D spillovers. However, there are differences in the impact on cost reduction across the different industries analysed (chemicals, petroleum, machinery and instruments) and in the short run and long run. For instance, in the long run, a 1% increase in the intra-industry spillover caused average cost to decline by around 0.1% for instruments and machinery and by approximately 0.2% for chemicals and petroleum (estimates twice as large as in the short run). Spilloverreceiving firms gained a 0.27%, 0.22%, 0.15% and 0.14% variable cost reduction in the petroleum, chemical, machinery, and instruments industries respectively as a result of a 1% increase in the intra-industry spillover. Arrow C in Figure A6.2 summarise these effects. Note the arrow stems from the private R&D ‘bubble’ without distinguishing between ‘new’ and ‘existing’ R&D. This is because spillover effects could arise from either or both ‘types’ of private R&D.

5. Mechanisms transmitting spillovers As we have seen, the literature is unambiguous in accepting that R&D/knowledge spillovers exist and are important. However, the literature is less clear about the mechanisms by which spillovers are transmitted. Krugman (1991) argues that knowledge flows are invisible, as they leave no paper trail, making spillovers difficult to measure. Jaffe et al. (1993) argue that on the contrary there is a visible paper trail in the form of patented inventions and new product introductions. Probably, most researchers involved in this area will agree with the assertion about the difficulty of measuring spillovers. There have been several attempts. Before going into the details of how spillovers are measured, we briefly discuss some of the mechanisms identified in the literature as facilitating spillovers’ transmission. First, there are the mechanisms facilitated by universities, which include their pool of talented graduates, the ideas generated by faculty, their high quality libraries and other facilities of research universities and their publications. Second, networking and social interactions are also deemed to be important mechanisms – and these include both the informal and formal ways of interaction. Formal means of interaction include technology transfer programmes, such as licensing from universities to firms. Both means of interaction seem to be relatively important for the pharmaceutical and biotech market.

Powell (1990) argues that social networks offer a highly feasible means of utilising and enhancing such intangible assets as tacit knowledge and technological innovation. Powell also argues that the exchange of distinctive competencies – be it knowledge or skills – is more likely to occur in networks. Breschi and Lissoni (2006) carry out a reassessment of the arguments and tests in support of exercise and magnitude of local knowledge spillovers proposed by Jaffe et al. (1993). They add both new data (Italian patents) and new measurable variables (social proximity between inventors, and inventors’ mobility across firms). Breschi and Lissoni interpret their results as an indication that localisation effects tend to vanish where citing and cited patents are not linked to each other by any network relationship. On the contrary, knowledge flows, as evidenced by patent citations, are strongly localised to the extent that labour mobility and network ties also are. These authors conclude that geography is not a sufficient condition for accessing a local pool of knowledge, which also requires active participation in a network of knowledge exchanges. Results found in Agrawal et al. (2006) are also consistent with the conjecture that social relationships facilitate knowledge spillovers. Gambardella (1992) argued that for the pharmaceutical market in particular, information exchange, rather than retaining it within one’s own organisational boundaries, is a major determinant of successful innovation. He claimed that this requires companies to be prepared to diffuse research findings in exchange for the knowledge produced by others. Thus, to become part of a network, and to be able to effectively exploit the information that circulates in the network, has become even more valuable than being able to generate new knowledge autonomously. Owen-Smith and Powell (2004) highlight the importance of knowledge networks in the biotech sector. These authors claim that within regional economies, contractual linkages among physically proximate organisations represent relatively transparent channels for information transfer. Also, they argue that spillovers that result from proprietary alliances (i.e. formal networks) are a function of institutional commitments and practices of members of the network. Feldman and Kelley (2006) argue that successful strategies for learning abut technical advances outside a company’s internal R&D efforts may depend on the breadth of collaborative links with other enterprises, connections to universities and the adoption of university norms of publishing research; and, probably more important, that knowledge flows both ways along these pathways. Along the same lines, Martin and Tang (2007) cite a number of papers highlighting how “through these networks, scientists can quickly and effectively contact acknowledged experts on a particular issue to obtain from them information or advice” (p. 11).

Feldman (2000) also defends the importance of social interactions, although she argues that we do not know exactly how economically useful knowledge is created as a result. Breschi and Malerba (2005) reinforce the importance of social networks and argue that one of the key issues raised by all of the approaches in the literature is that learning through networking and by interacting is seen as the crucial force pulling firms into clusters and the essential ingredient for the ongoing success of an innovative cluster. The third mechanism discussed in the literature relates to the possibility of ‘absorptive capacity’, which refers to the ability of economic agents to recognise, assimilate and apply new scientific knowledge – and then to appropriate some of the returns accruing to investments in new knowledge made externally (Cohen and Levinthal, 1989). Indeed, as these authors argue, the conventional wisdom was that R&D generated only one product: new information. They argue, however, that R&D also enhances the firm’s ability to assimilate and exploit existing information – the firm’s ‘learning’ or ‘absorptive’ capacity. This second ‘face’ of R&D is very important, as it represents an important element of a firm’s ability to create new knowledge. Griffith et al. (2004a) find strong evidence that R&D has this second face: industries lagging behind the productivity frontier catch up particularly fast if they invest heavily in R&D. Griffith et al. identify human capital as having a role in stimulating innovation and absorptive capacity. Fourth, entrepreneurship has been also identified as an important mechanism to transmit spillovers, in that knowledge spillovers are the source of knowledge creating the entrepreneurial opportunities for new firms. Partly due to the emergence of the literature identifying entrepreneurship as a mechanism to facilitate spillovers, the so-called Knowledge Spillover Theory of Entrepreneurship (KSTE) has been developed. It identifies one source of entrepreneurial opportunity: new ideas and knowledge. Under this theory, the implementation of new knowledge and new ideas created in one source but left uncommercialised or not fully pursued generates entrepreneurial opportunities to start up new firms. Thus, the creation of a new firm in a localised context is an important mechanism by which knowledge spills over. By serving as a conduit for knowledge spillovers, entrepreneurship is the missing link between investments in new knowledge and economic growth (Audretsch et al., 2006). Related to the fourth mechanism, Audretsch and Stephan (1999) try to answer the question: ‘how does knowledge spill over?’ They argue that the answer lies in the incentives confronting scientists to appropriate the expected value of their knowledge, considered in the context of their path-dependent career trajectories. In particular, Audretsch and Stephan focus on the ability of scientists to appropriate the value of the

knowledge embedded in their human capital along with the incentive structure influencing whether and how scientists choose to commercialise their knowledge. They conclude that the spillover of knowledge from the source creating it, such as a university, research institute or industrial corporation, to a new firm start-up facilitates the appropriation of knowledge for the individual scientist(s) but not necessarily for the organisation creating the knowledge in the first place. International trade is usually deemed to be one of the most important mechanisms by which spillovers are transmitted across countries, but international spillovers are outside the scope of this paper. The interested reader is referred to Coe and Helpman (1995), Keller (1998, 2002a, 2002b, 2004) and Griffith et al. (2004b).

6. The rise of the economics of geography of innovation Figure A6.2 (deliberately) ignores the geographical dimension. For instance, it does not take into consideration where the public and private R&D takes place and how close you have to be to the source of the spillover to access it. This is nevertheless an important dimension, as the evidence gathered in our literature review shows. Griliches’ (1979) seminal paper offered for the first time a tool to measure spillovers, by suggesting the production function approach to the estimation of the return to R&D. He argued that the level of productivity achieved by one firm or industry depends not only on its own research efforts but also on the size of the pool of general knowledge accessible to it. Moreover, the productivity of own research may be affected by the size of the pool or pools it can draw upon. A simple model of such a within-industry spillover effect is given by: Yi = B X i1- K i Kaμ where Yi = output of ith firm, which depends on an index of conventional inputs, X i, its specific knowledge capital, K i, and the state of aggregate knowledge in this industry, Ka. Griliches assumed that the aggregate level of knowledge capital was the sum of all specific firm R&D capital levels. Aggregating the individual production functions, the coefficient of aggregate knowledge capital ( + μ) is higher than at the micro level ( ), since the aggregate level reflects not only the private but also the social returns to R&D. This production function approach is aspatial or insensitive to issues involving location and geography. As argued by Audretsch (1998), there is considerable empirical evidence supporting the model of the knowledge production function. The empirical link between knowledge inputs and innovative output becomes stronger when the unit of observation becomes increasingly aggregated, e.g. when the

unit of observation is the country or the industry. However, at the level of the firm the link between knowledge inputs and innovative output becomes tenuous and only weakly positive in some studies, and even non-existent or negative in others. This is not surprising, as formal R&D is concentrated among the largest corporations and small firms account disproportionately more for new product innovations. Several economists (see references below) have therefore modified the model of the knowledge production function to include an explicit specification for both spatial and product dimensions. The modified knowledge production function becomes: Isi = IRD 1 * (UR si) 2 * [UR si * (GCsi) 3] *

where I = innovative output, IRD is private corporate expenditures on R&D, UR is research expenditures undertaken by universities, and GC measures geographic coincidence between university and corporate research. This co-location variable was added later. The unit of observation is at the spatial level, s, a state, and industry level, i. The estimation of this new equation shifts the model of the knowledge production function from the unit of observation being a firm to being a geographic unit: state, region or metropolitan area. This work has led to the notion of ‘localised knowledge spillovers’. The breakthrough paper on the topic is probably that by Jaffe (1989a), who argued that transport mechanisms for spillovers at the time were still not understood. As a first approach to analyse these issues, he examined (for the US) production of patents assigned to corporations by state over time, and related this to industry R&D and university research. Jaffe interprets the influence of university research on these patents at the state level (after controlling for industry R&D) as evidence of geographically mediated spillovers. Several papers have followed a similar methodology: see for instance Feldman (1993), Acs et al. (1994a), Feldman (1994), Audretsch and Feldman (1996), Audretsch (1998), Feldman and Audretsch (1999), Feldman (2000), Arundel and Geuna (2001), Audretsch et al. (2003), Baum and Sorenson (2003), Audretsch and Feldman (2004), Audretsch and Lehmann (2006). The authors of these papers, to a greater or lesser extent, argue that the empirical evidence they present suggests that location and proximity clearly matter in exploiting knowledge spillovers. They argue that one of greatest developments in the literature of the economics of innovation is that geography, the spatial context, does matter. This strand of the literature has gone one step further to analyse the different ‘elasticities’ according to firm size. Empirical results from the USA suggest that small firms are the recipients of R&D spillovers from knowledge generated in the R&D centres of their larger counterparts and in universities (Acs et al., 1994b). Such spillovers are apparently more decisive in promoting the innovative

activity of small firms than of large corporations. Audretsch and Vivarelli (1996) replicated the analysis for 20 Italian regions over nine years and found similar results. They show that while firm R&D expenditures contribute to the generation of innovative output for all firms, the spillovers from university research are apparently more important for small-firm innovation. Jaffe et al. (1993) provide an alternative analysis for looking at the importance of geography. They show, based on US data, that patent citations are highly localised. These authors also argue that it is probable that knowledge spillovers are not confined to closely related regions of technology space – in their study, approximately 40% of citations do not come from the same primary patent class. This is consistent with Jaffe (1986) who found that a significant fraction of the total flow of spillovers affecting firms’ own research productivity comes from firms outside the receiving firm’s immediate technological neighbourhood. Botazzi and Peri (2003) use the total number of patents granted to inventors residing in a European region as a measure of that region’s innovative output in order to investigate R&D externalities, identified as the effect of R&D intensity in one region on the innovative output of another region. These authors find evidence that spillovers are important. However, they find the effects of R&D in generating innovation quite localised – most of the benefits accrue to the region that employs the R&D resources and small positive externalities accrue to regions within 300 km of it. Quantitatively, they find that doubling R&D in a region would increase by 2–3% the patenting activity in another region within 300 km of distance. Closer to its border (within 100 km) and for regions of the same country the effect could be as large as 5–6%. They offer a tentative explanation for the small size and the short range: spillovers are the result of the diffusion of non-codified knowledge between people who have frequent (weekly or monthly) interactions. In Europe, the authors argue, people probably commute and interact much more within countries than across borders. In summary, results from the literature reviewed here suggest that one of the greatest developments in the study of the economics of innovation is to demonstrate that location and proximity do matter in exploiting knowledge spillovers. Geographic concentrations of knowledge are likely to create higher levels of innovation than would otherwise be achieved. Not only do product innovations exhibit a pronounced tendency to cluster in regions which contain concentrations of innovative inputs, but also innovative activity tends to cluster more in industries where knowledge spillovers play a decisive role. The propensity for innovative activity to cluster is more attributable to the role of knowledge spillovers than merely to the geographic concentration of production. The literature suggests that the role of knowledge spillovers is geographically bounded:

innovative activity is more likely to occur within close geographic proximity to the source of that knowledge – ‘localised knowledge spillovers’. Breschi and Lissoni (2001a, 2001b) have criticised this literature. They recognise that there is hardly any doubt that innovation networks are often localised. However, in their view the rationale for co-localisation may have less to do with knowledge spillovers mediated by physical proximity, than with the need to access a pool of skilled workers and to establish transaction-intensive relationships with suppliers and customers. Thus, in their critique, they argue that the notion of localised knowledge spillovers has been abused, generating great conceptual confusion. These authors do not deny that knowledge flows may be an extremely important agglomeration force, but they disagree with putting all of these flows under the common heading of localised knowledge spillovers.

7. Summary It is extremely important to consider wider economic gains as well as health gains when assessing the economic benefits of medical research. Investment in medical research by one organisation may benefit not only that organisation but also other organisations in the medical sector, and in other sectors, i.e. the whole economy. We have reviewed and presented the findings of the research spillovers literature in general and as applied specifically to medical research. The literature clearly suggests that R&D investment generates not only a direct return captured by the original investor, but also some indirect benefits ‘spilling over’ to third parties. There is evidence that R&D spending, both private and public, produces new knowledge, which can be accessed and exploited by other organisations for new technology development; that it contributes to the formation of a more skilled labour force; and more generally that it facilitates knowledge dissemination and technology transfer. The evidence shows that innovative activity tends to cluster more in industries where knowledge spillovers play a decisive role, and that new firm start-ups’ decisions to locate are influenced by the opportunity to access knowledge generated by universities. The literature is much sparser and more ambiguous about the quantification of spillovers than about the mechanisms through which they travel. There is some evidence for pharmaceutical R&D. This suggests that R&D spillovers are important, and that publicly funded research and private pharmaceutical R&D investment are complements rather than substitutes.

Annex to Chapter Seven Lag structures Given our data series for research expenditures for 1975–92 and for health gains for 1986–2005 the different lag structures we consider lead to different periods of research expenditure and of health gain being included in the estimated internal rate of return. This diagram indicates the first and last pair of years included in the analysis for each lag period. This leads to certain apparent anomalies, in terms of the effects of changing lags, particularly where there is substantial year on year variation in the data series.

Figure A7.1: Lags used for CVD 10-YEAR LAG

17-YEAR LAG

25-YEAR LAG

Research expenditure

Health gains 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

Figure A7.2: Lags used for mental health 9-YEAR LAG

12-YEAR LAG

14-YEAR LAG

Annexes to Chapter Eight Annex 8A: Breakdown of public, non-profit and private expenditure on mental health research (in £m), by year, 1970–2005, current prices

29 303 36 47 52 54 62 74 93 107 113 120 124 129 138 150 163 191 202 228 251 288 298 305 309 321 316 345 368

UK health expenditure – organisations (£m) Wellcome DH Trust 2 NHS (not NHS)

Mental Health – activity funding (£m) FC Health

5 18 4 5 45 4 6 6 10 10 12 12 17 17 24 13 29 35 68 43 54 72 92 16715 242 230 168 222 212 354 480 388 419

1376 144 151 158 166 174 183 191 201 210 221 231 242 254 266 279 293 307

84 8 87 91 95 99 103 107 112 116 121 126 131 137 142 15714 187 212 208 19016 186 187 186 205 222 243 250 257 280 306 317 327

WT 0 0 0 0 0 0 0 0 1 1 1 1 1 1 2 2 3 3 3 5 5 9 18 49 30 20 27 37 32 40 38 32 31

39 4 5 5 6 8 9 9 10 12 11 11 13 14 15 16 19 18

DH – A 0 0 0 0 0 5 6 6 6 7 7 6 7 7 7 7 10 10 10 11 13 12 12

9 11 11 12 14 15 16 17 18 21 20 24 25 27 32 34 39 48 49 30 20 27 37 32 40 38 32 31 0 0 0

Mental health – support funding (£m) Funding DH – I Councils Total

11 12 12 13 13 14 15 15 16 17 18 18 19 20 21 22 23 25

8 9 9 9 10 10 11 11 12 12 13 13 14 14 16 19 21 21

19 2010 21 22 23 24 25 26 28 29 30 32 33 35 37 41 45 45

Pharma industry (£m)

Grand total exc. pharma (£m)

Grand total exc. pharma. deflated (£m)

15 20 25 30 3911 48 56 68 77 92 104 116 127 141 182 217 235 270

28 31 32 34 37 39 42 44 45 50 50 55 59 62 69 75 84 93

155 149 137 130 123 110 104 103 101 106 102 109 110 108 113 114 120 129

3 Maddock (1975) estimates MRC expenditure in 1973/74 to have been £24.8m. 4 Figure taken from Maddock (1975). 5

Data for 1972–84 were provided in two-year bands; we have taken the average for annual estimates.

6 Data for 1975–92 were estimated by back-casting the figures provided for 1995–2005 using an exponential function. 7 Data for 1975–80 based on projecting data for 1973 and 1981–2005 using a linear function. 8

The 1975 to 1989 timeseries is estimated from a back-projection of the 1989 to 1992 UFC data for biomedical subjects in the UK, and the 1993 to 2005 HEFCE QR data for biomedical subjects in England. The latter is then inflated by circa 12% to make a UK-wide estimate.

10 Data for 1976 and 1977 interpolated using figures for 1975 and 1978. 11 Taken as average of the preceding and following points. 12 Data for 1981–85 entered by hand from Annual Review of Government-Funded R&D (Cabinet Office, 1984). 13 Data for 1985–87 was estimated using an exponential function to project funding for the period 1975–84, for which we were provided with data. 14 Numbers for these years are taken from UFC figures and are UK-based. 15 Calculated by taking the average of figures for 1992 and 1994. 16 From 1993 onwards, data is provided by HEFCE and is England-based.

Annex 8B: List of mental health guidelines analysed

List of mental health guidelines analysed Guideline

Bipolar disorder

www.nice.org.uk/nicemedia/pdf/CG38niceguideline.pdf

Publication date 2006

www.nice.org.uk/nicemedia/pdf/CG042NICEGuideline.pdf

www.nice.org.uk/nicemedia/pdf/cg023fullguideline.pdf

Depression in children and young people

www.nice.org.uk/nicemedia/pdf/CG028NICEguideline.pdf

Drug misuse: opioid detoxification

www.nice.org.uk/nicemedia/pdf/CG52NICEGuideline.pdf

Drug misuse: psychosocial interventions

www.nice.org.uk/nicemedia/pdf/CG051NICEguideline2.pdf

Mental Health NSF

www.dh.gov.uk/en/Publicationsandstatistics/Publications/ PublicationsPolicyAndGuidance/DH_4009598

Obsessive-compulsive disorder

www.nice.org.uk/nicemedia/pdf/cg031fullguideline.pdf

Antenatal and postnatal mental health

www.nice.org.uk/nicemedia/pdf/CG045NICEGuidelineCorrected.pdf

Post-traumatic stress disorder

www.nice.org.uk/nicemedia/pdf/CG026NICEguideline.pdf

Schizophrenia

www.nice.org.uk/nicemedia/pdf/cg001fullguideline.pdf

www.bps.org.uk/downloadfile.cfm?file_uuid=C11587F1-7E96-C67F-DD13357E1AA3B75D&ext=pdf (linked through the NICE website)

www.nice.org.uk/nicemedia/pdf/cg025fullguideline.pdf

Annex 8C: Studies used to generate QALYs gained for each intervention in each patient group Table A8C: Summary of studies used to generate QALYs gained for each intervention in each patient group QALYs gained per user year

Comparators

Kendrick et al. (2006)

We used the absolute QALY gain between baseline and 12 months. We calculated the integral between baseline (0.608) and 12-month (0.781) measurement of quality of life to obtain a QALY estimate for one calendar year.

Patient groups/interventions Treatment of depression:

Male and female patients, over 18 years of age, suffering from depression

SSRIs followed up by TCA in the case of dropout, TCA followed by a SSRI in case of a dropout and TCA only

Patients affected by non-psychiatric problems in need of psychiatric help

Direct CPN care, UK delayed CPN care (after 12 weeks) and continuing GP care

Patients with DSM-IV diagnoses of major depressive disorder, dysthymic disorder, panic disorder, social phobia, and generalised anxiety disorder

The study compared three first-line treatments for depression and anxiety. The options were brief therapy, cognitivebehavioural therapy and care as usual

24 weeks Gournay and Brooking (1995)

Netherlands up to Van Roijen 1.5 years et al. (2006)

We used the incremental QALY gain over standard care. Direct CPN treatment saved four days’ absence from work. Preventing absence from work was approximately valued at 0.1 of a QALY per year. No other details were given in the study. We used the incremental QALY gain over GP care. The utility score at baseline was 0.52 in the CBT group. At one-year follow-up, the utility score was 0.68 in the CBT group. We assumed a linear increase in quality of life between baseline and 12-month measurement.

Treatment of schizophrenia:

Patients with schizophrenia

One first generation USA drug (perphenazine) and four second generation drugs (olanzapine, quetiapine, risperidone, zipradone)

Rosenheck et al. (2006)

We used the absolute QALY gain between baseline and 12 month. The utility values changed from approximately 0.682, 0.695 and 0.676 to 0.725, 0.73 and 0.725 for olanzapine, quietapine and risperidone respectively. We modelled the average linear increase between baseline and 12-month measurement.

We used the incremental QALY gain over standard care. Direct CPN treatment saved four days’ absence from work. Preventing absence from work was approximately valued at 0.1 of a QALY per year. No other details were given in the study.

We used incremental QALY gain over standard care. Direct CPN treatment saved four days’ absence from work. Preventing absence from work was approximately valued at 0.1 of a QALY per year. No other details were given in the study.

Other mental disorders:

SSRI = selective serotonin reuptake inhibitors, CPN = community psychiatric nurse, CBT = cognitive behavioural therapy, FGA = first generation antipsychotics.

Annex 8D: Estimating the numbers of eligible patients, the number of initial contacts with community psychiatric nurses and the number of SSRI users Table A8D delineates the data sources and assumptions used to generate the numbers of eligible patients in each patient group, the number of new patients seeing community psychiatric nurses and the number of SSRI users over the period 1985–2005. Table A8D: Data and assumptions used to generate the numbers of eligible patients, CPN contacts and SSRI users Patient groups/interventions

Treatment of depression:

Data for 1975–98 obtained from Middleton et al. (2001). These are prescription figures from the UK. For 1999–2005, we linearly extrapolated using the rate of increase in prescriptions between 1997 and 1998. Kendrick et al. (2006) suggests that 931 prescriptions are used for 81.6 patient years. To obtain the number of users, we divided the number of prescriptions by 11.04 in each year.

Data for 1988 to 2003 obtained from Patient Care in the Community – NHS Community Mental Health Nursing (ONS and DH, 2003/2004). These figures are for England; UK figures are computed by multiplying the share in England by the UK population. The data gives the initial contact with a mental health nurse each year. Where a previous episode of care for the same patient did not end with a positive discharge from care, a new episode is recorded only if more than 6 months have elapsed since the last contact. For 2004–05 we extrapolate the data using the slope from 2003 to 2002.

Data for 1994–98 obtained from Key Health Statistics from General Practice (National Statistics, 2000). These are figures for England; UK figures are computed by multiplying the share in England by the UK population. For 1985–93 we assume the rate is the same as in 1994. For 1999–2005 we assumed that the rate is the same as in 1998.

Antipsychotics

Data for 1994–98 obtained from Key Health Statistics from General Practice (National Statistics, 2000). These are figures for England; UK figures are computed by multiplying the prevalence in England by the UK population. For 1985–1993 we assume the rate is the same as in 1994. For 1999–2005 we assumed that the rate is the same as in 1998.

Data for 1994–98 obtained from Key Health Statistics from General Practice (National Statistics, 2000). These are figures for England; UK figures are computed by multiplying the prevalence in England by the UK population. For 1985–93 we assume the rate is the same as in 1994. For 1999–2005 we assumed that the rate is the same as in 1998.

Information can be found under description of CPN user numbers for depressed patients.

Annex 8E: Estimating the uptake rates for specific interventions in each patient group Table A8E: Data and assumptions used to generate uptake rates Patient groups/interventions

Evidence from Gournay and Brooking (1994), Greenwood et al. (2000) and Hannigan (1999). Hannigan argues that the patient mix treated by CPN changed in the early 1990s to a case mix of patients with severe mental disorders. We used numbers on the patient mix from Gournay and Brooking (24% depressed before 1990 and approximately a quarter schizophrenic in 1994) and Greenwood et al. (12.6% depressed and 55% schizophrenic in 2000). Before 1990, we assumed that 24% of CPN patients were depressed. Following Gournay and Brooking, we further assumed a linear increase from 0% CPN treatment for schizophrenic patients in 1985 to 25% in 1994. After 2000, we used the 2000 rates for schizophrenic and depressed patients. The remaining percentages were attributed to CPN treatment of other mental disorders.

Data for 1993 obtained from Psychiatric Morbidity Among Adults Living in Private Households, 2000 (ONS and DH, 2001). This is a UK figure. CBT was introduced in the UK in the early 1990s. We therefore assumed the uptake rate to be zero prior to 1990 and linearly increasing between 1990 and 1994. For the period after 1998, we applied the 1998 uptake rate.

Figures for 1994–98 from Key Health Statistics from General Practice. The figures are for England. Evidence from prescription data suggested that atypical antipsychotics were only used after 1989 in the UK. We assumed a zero uptake rate in 1989 and linearly interpolate between 1989 and 1994. From Department of Health (2007) we assumed linear increase. For the period after 1993, we extrapolated using the slope from 1992 to 1993. Between 1995 and 2005, the uptake rate of first generation antipsychotics (FGA) is decreasing. Since FGA and atypical antipsychotics are substitute therapies, we deducted the decrease in FGA users from the number of atypical antipsychotic users. Following this adjustment, we assumed that all considered atypical antipsychotic users in our model, would not have received any other antipsychotic medication otherwise.

Information can be found under description of CPN uptake rates for depressed patients.

SSRI = selective serotonin reuptake inhibitors, CPN = community psychiatric nurse, CBT = cognitive behavioural therapy, FGA = first generation antipsychotics, N/A = not applicable.

Annex 8F: Studies used to generate incremental costs for each intervention in each patient group Table A8F: Summary of studies used to generate incremental costs for each intervention in each patient group Raw incremental costs

Incremental costs (2005 UK£)

Consistent with the respective QALY estimate, we used the absolute cost of SSRI over one year

Gournay and Brooking (1995)

We used the incremental cost of direct CPN treatment compared to GP treatment

Scott et al. (2003)

Consistent with the respective non-UK QALY estimate, we used the incremental cost compared to standard care from this UK study. However, the costs were for a 17-month treatment

Consistent with the respective non-UK QALY estimate, we used the absolute cost of atypical antipsychotics from this UK study

We used the incremental cost of CPN treatment compared to GP treatment

Annex 8G: Sensitivity analysis We undertook a sensitivity analysis to generate upper and lower estimates of the total QALYs gained and incremental costs around our central estimates. These were calculated as follows: Our lower estimate of the net health gain made three adjustments compared with our base case: • Our base case estimate may have overestimated the quality of life gain from SSRIs: it has been suggested that the QALY gain compared to placebo may be more modest as they are increasingly used in general practice and less severe patients (Kirsch et al., 2008). In our lower estimate, we assumed that each SSRI user gained only half the QALY gain compared to that assumed in the base case. • In our base case the cost of SSRI treatment per year was simply the prescription cost assuming that prescribing the drug had no effect on patient contacts with the health care system. In our low net benefit case, we used an estimate of the full cost of patient care, reflecting the alternative assumption that these patients would otherwise not be receiving active health service care (Kendrick et al., 2006).

• Our lower estimate also used a more conservative estimate of the uptake rate for atypical antipsychotics between 1985 and 2005. Our upper estimate made two adjustments from the base case: • Our base case estimated the QALYs gain per year as a steady increase to observed improvements at one year. This may underestimate the QALY gain for those patients on medication longer than one year. For the upper estimate, we assumed that half the years of use were for new users and the other half for patients continuing on SSRIs and enjoying the full benefit throughout the year. • Our base case estimate of QALYs gained for patients receiving atypical antipsychotics reflected a study in which patients switched from a current treatment to an atypical antipsychotic (Rosenbeck et al., 2006). The gain for these patients may have been less than for patients who were not previously treated. For this higher net benefit estimate, we used a higher estimate based on a small earlier study of previously untreated patients (Chouinard, 1997).

Table A8G1: Results of sensitivity analysis, 1985–2005: lower estimate 1

Patient groups/intervention

Total (£ million) 9,365

SSRIs = selective serotonin reuptake inhibitors

Table A8G2: Results of sensitivity analysis, 1985–2005: upper estimate 1

Total (£ million)

Depression: SSRIs

Schizophrenia: Atypical antipsychotics

References Abramovski L, Harrison R, Simpson H. University research and the location of business R&D. Economic Journal 2007;117(519):114–41. Access Economics. Exceptional Returns: The value of investing in health R&D in Australia. Canberra: Access Economics; 2003. www. accesseconomics.com.au/publicationsreports/showreport.php?id=33&sear chfor=2004&searchby=yearearlier [accessed 15 September 2008].

Bernstein J, Nadiri M. Interindustry R&D spillovers rates of return and production in high-tech industries. American Economic Review papers and proceedings 1988;78(2):429–34. Bernstein J, Nadiri M. Research and development and intra-industry spillovers: an empirical application of dynamic duality. Review of Economic Studies 1989;56:249–68.

Access Economics. Exceptional Returns: The value of investing in health R&D in Australia II. Canberra: Access Economics; 2008. www. accesseconomics.com.au/publicationsreports/showreport.php?id=165 [accessed 15 September 2008].

Bernstein J, Nadiri M. Product demand, cost of production, spillovers, and the social rate of return to R&D. NBER Working Paper 3625, 1991.

Acs Z, Audretsch D, Feldman M. Real effects of academic research: comment. American Economic Review 1992;82(1):363–7.

Breschi S, Lissoni F [2001a]. Knowledge spillovers and local innovation systems: a critical survey. Industrial and Corporate Change 2001;10(4):975–1005.

Botazzi L, Peri G. Innovation and spillovers in regions: evidence from European patent data. European Economic Review 2003;47:687–710.

Acs Z, Audretsch D, Feldman M [1994a]. R&D spillovers and innovative activity. Managerial and Decision Making 1994;15(2):131–8. Acs Z, Audretsch D, Feldman M [1994b]. R&D spillovers and recipient firm size. Review of Economics and Statistics 1994;76(2):336–40.

Breschi S, Lissoni F [2001b]. Localised knowledge spillovers vs. innovative milieux: knowledge ‘tacitness’ reconsidered. Papers in Regional Science 2001;80(3):255–73.

Agrawal A, Cockburn I, McHale J. Gone but not forgotten: labor flows, knowledge spillovers, and enduring social capital. Journal of Economic Geography 2006;6(5):571–91.

Breschi S, Lissoni F. Cross-firm inventors and social networks: localised knowledge spillovers revisited. Annales d’Economie et de Statistique 2006;79–80:1–21.

Aharonson B, Baum J, Feldman M. Desperately seeking spillovers? Increasing returns, industrial organisation and the location of new entrants in geographic and technological space. Working paper. Toronto: University of Toronto; 2006.

Breschi S, Malerba F. The geography of innovation and economic clustering. Some introductory notes. In S Breschi, F Malerba (eds). Introduction, in Clusters, networks and innovation. Oxford: Oxford University Press; 2005. Briffa TG, Eckermann SD, Griffiths AD, Harris PJ, Heath MR, Freedman SB, Donaldson LT, Briffa NK, Keech AC. Cost-effectiveness of rehabilitation after an acute coronary event: a randomised controlled trial. Med J Austral 2005;183:450–5.

Annemans L, Lamotte M, Kubin M, Evers T, Verheugt FWA. Which patients should receive aspirin for primary prevention of cardiovascular disease? An economic evaluation. Int J Clin Pract 2006;60:11–1137. Arrow K. Economic welfare and the allocation of resources for invention. In R Nelson (ed.). The Rate and Direction of Inventive Activity. Princeton, NJ: Princeton University Press; 1962. pp. 609–26. Arundel A, Geuna A. Does proximity matter for knowledge transfer from public institutes and universities to firms? SEWP 73. 2001. Association of the British Pharmaceutical Industry (ABPI). Threat to pharmaceutical industry support for university research. Press release 2008 22 April.

Buxton M, Caine N, Chase D, Connelly D, Grace A, Jackson C, Parkes J, Sharples L. A review of the evidence on the effects and costs of implantable cardioverter defibrillator therapy in different patient groups, and modelling of cost-effectiveness and cost–utility for these groups in a UK context. Health Technol Assess 2006;10(27). Buxton M, Hanney S. How can payback from health services research be assessed? Journal of Health Service Research and Policy 1996;1:35–43.

Audretsch D. Agglomeration and the location of economic activity. Oxford Review of Economic Policy 1998;14:18–29.

Buxton M, Hanney S, Jones T. Estimating the economic value to societies of the impact of health research: critical review. Bulletin of the World Health Organization 2004;82(10):733–9.

Audretsch D, Feldman M. R&D spillovers and the geography of innovation and production. American Economic Review 1996;86(3):630–40.

Cabinet Office. Annual Review of Government Funded R&D. London: HMSO; 1984.

Audretsch D, Feldman M. Knowledge spillovers and the geography of innovation. In JV Henderson, JF Thisse (eds). Handbook of Regional and Urban Economics. Elsevier; 2004. pp. 2713–39.

Capewell S, Beaglehole R, Seddon M, McMurray J. Explanation for the decline in coronary heart disease mortality rates in Auckland, New Zealand, between 1982 and 1993. Circulation 2004;102(13):1511–6.

Audretsch D, Keilbach M, Lehmann E. Entrepreneurship and economic growth. Oxford: Oxford University Press; 2006.

Capewell S, Critchley J, Unal B. IMPACT: A validated, comprehensive coronary heart disease model. Overview and technical appendices. Liverpool: University of Liverpool; 2007.

Audretsch D, Lehmann E [2004a]. University spillovers and new firm location. Discussion Papers on Entrepreneurship, Growth and Public Policy 2004. Audretsch D, Lehmann E [2004b]. Mansfield’s missing link: the impact of knowledge spillovers on firm growth. Journal of Technology Transfer 2004;30(1–2):207–10. Audretsch D, Lehmann E. Do locational spillovers pay? Empirical evidence from German IPO data. Economics of Innovation and New Technology 2006;15(1):71–81. www.informaworld.com/smpp/title~content=t713641545~ db=all~tab=issueslist~branches=15#v15 [accessed 15 September 2008]. Audretsch D, Lehmann E, Warning S. University spillovers: strategic location and new firm performance. DP3837. 2003. Audretsch D, Stephan P. Knowledge spillovers in biotechnology: sources and incentives. Journal of Evolutionary Economics 1999;9(1):97–107. Audretsch D, Vivarelli M. Firms size and R&D spillovers: evidence from Italy. Small Business Economics 1996;8(3):249–58. Aujesky D, Smith KJ, Roberts MS. Oral anticoagulation strategies after a first idiopathic venous thromboembolic event. Am J Med 2005;118:625–635 Baum JAC, Sorenson O. Advances in Strategic Management Geography and Strategy. Vol. 20. Greenwich, CT: JAI Press; 2003. Beard SM, Gaffney L, Bamber L, De Platchett J. Economic modelling of antiplatelet therapy in the secondary prevention of stroke. J Med Econ 2004;7:117–34.

Capewell S, Morrison CE, McMurray JJ. Contribution of modern cardiovascular treatment and risk factor changes to the decline in coronary heart disease mortality in Scotland between 1975 and 1994. Heart 1999;81(4):380–6. Capewell S, Unal B, Critchley J, McMurray JJV. Over 20 000 avoidable coronary deaths in England and Wales in 2000: the failure to give effective treatments to many eligible patients. Heart 2006;92:521–3. Cassiman B, Veugelers R. R&D cooperation and spillovers: some empirical evidence from Belgium. American Economic Review 2002;92(4):1169–84. Chambers M, Gladman J, Hutton J. Cost-effectiveness of stroke units and severity of stroke. Annu Meet Int Soc Technol Assess Health Care 1998;14:64. Charles River Associates (CRA). Innovation in the pharmaceutical sector. A study undertaken for the European Commission. 2004. www.crai. co.uk/uploadedFiles/RELATING_MATERIALS/Publications/Consultant_ publications/files/pub_4510.pdf [accessed 15 September 2008]. Chouinard G. Economic and health state utility determinants for schizophrenic patients treated with risperidone or haloperidol. Journal of Clinical Psychopharmacology 1997;17(4):298–307. Clegg AJ, Scott DA, Loveman E, Colquitt J, Hutchinson J, Royle P, Bryant J. The clinical and cost-effectiveness of left ventricular assist devices for end-stage heart failure: a systematic review and economic evaluation. Health Technol Assess 2005;9(45).

Clinical Effectiveness and Evaluation Unit (CEEU), Royal College of Physicians of London. National Sentinel Stroke. Audit Phase I (organisational audit) 2006. Phase II (clinical audit) 2006. London: Royal College of Physicians; 2007.

EUROASPIRE II Study Group. Lifestyle and risk factor management and use of drug therapies in coronary patients from 15 countries. Principal results from EUROASPIRE II Euro Heart Survey Programme. Eur Heart J 2001:22;554–72.

Cockburn I, Henderson R. Racing to invest? The dynamics of competition in ethical drug discovery. Journal of Economics and Management Strategy 1994;3(3):481–519.

European Science Foundation/European Medical Research Councils. SocioEconomic Value of Medical Research: Strategic workshop. Strasbourg, 26–27 May 2005.

Cockburn I, Henderson R. Public-private interaction in pharmaceutical research. Proc Natl Acad Sci USA 1996;93:12725–30.

Feldman M. An examination of the geography of innovation. Industrial and Corporate Change 1993;2(3):417–37.

Cockburn I, Henderson R. Absorptive capacity, coauthoring behaviour and the organization of research in drug discovery. The Journal of Industrial Economics 1998;66(2):157–82.

Feldman M. Knowledge complementarity and innovation. Small Business Economics 1994;6(5):363–72.

Cockburn I, Henderson R. Publicly funded science and the productivity of the pharmaceutical industry. Presented at NBER conference on science and public policy; 2000. Coe DT, Helpman E. International R&D spillovers. European Economic Review 1995;39:859–87. Cohen W, Levinthal D. Innovation and learning: the two faces of R&D. Economic Journal 1989;99:569–96. Comroe J, Dripps R. Scientific basis for the support of biomedical science. Science 1976;192:105–11. Congressional Budget Office (CBO). Research and Development in the Pharmaceutical Industry. Washington, DC: 2006. Cooksey D. A review of UK health research funding. London; HM Treasury: 2006. www.hm-treasury.gov.uk/media/4/A/pbr06_cooksey_final_ report_636.pdf [accessed 15 September 2008]. Cramer J, Rosenheck R. Compliance with medication regimens for mental and physical disorders. Psychiatric Services 1998;49:196–201. Crespi G, Geuna A. The Productivity of Science. Report prepared for the Office of Science and Technology, Department of Trade and Industry. Brighton: University of Sussex; 2004. www.sussex.ac.uk/spru/documents/ crespiost2.pdf [accessed 15 September 2008]. Cutler D, Rosen A, Vijan S. The value of medical spending in the United States, 1960–2000. New England Journal of Medicine 2006;355:920–7. David P, Hall B. Heart of darkness: modeling public-private funding interactions inside the R&D black box. Research Policy 2000;29(9):1165–83. Davies L, Lewis S, Jones PB, Barnes TR, Gaughran F, Hayhurst K, Markwick A, Lloyd H. Cost-effectiveness of first- v. second-generation antipsychotic drugs: results from a randomised controlled trial in schizophrenia responding poorly to previous therapy. British Journal of Psychiatry 2007;191:14–22. Dawson G, Lucocq B, Cottrell R, Lewison G. Mapping the Landscape: National biomedical research outputs 1988–95. London: Wellcome Trust; 1998. Department for Innovation, Universities and Skills. The 2007 R&D Scoreboard. London: DIUS; 2007. www.innovation.gov.uk/downloads/2007_ rd_scoreboard_analysis.pdf [accessed 15 September 2008]. Department of Health [2004a]. Health Survey for England 2003. London: Department of Health; 2004. www.dh.gov.uk/en/Publicationsandstatistics/ Publications/PublicationsStatistics/DH_4098712 [accessed 15 September 2008]. Department of Health [2004b]. National Service Framework for Mental Health: Five years on. London: Department of Health; 2004. Department of Health. Medicines Uptake in England: A quantitative analysis of variation. London: Department of Health; 2007. Department of Trade and Industry. Biotechnology Clusters: Report of a team led by Lord Sainsbury, Minister for Science. London: DTI; 1999. Chapter 2: www.berr.gov.uk/files/file28739.pdf [accessed 15 September 2008]. DiMasi J, Hansen R, Grabowski H. The price of innovation: new estimates of drug development costs. Journal of Health Economics 2003;22:151–85. DiMasi J, Hansen R, Grabowski H, Lasagna L. Cost of innovation in the pharmaceutical industry. Journal of Health Economics 1991;10:107–52. Domino ME, Swartz MS. Who are the new users of antipsychotic medications? Psychiatric Services 2008;59:507–14. Drummond ML, Davies LM, Ferris F. Assessing the costs and benefits of medical research: the diabetic retinopathy study. Social Science and Medicine 1992;34(9):973–81.

Feldman M. Location and innovation: the new economic geography of innovation. In GL Clark, MS Feldman, MP Gertler (eds). Oxford Handbook of Economic Geography. Oxford: Oxford University Press; 2000. pp. 373–94. Feldman M. The locational dynamics of the US biotech industry: knowledge externalities and the anchor hypothesis. Industry and Innovation 2003;10(3):311–28. Feldman M, Audretsch D. Innovations in cities: science based diversity, specialisation and localised monopoly. European Economic Review 1999;43:409–29. Feldman M, Kelley M. The ex ante assessment of knowledge spillovers: government R&D policy, economic incentives and private firm behaviour. Research Policy 2006;35:1509–21. Fox M, Mealing S, Anderson R, Dean J, Stein K, Prince A, Taylor R. The clinical effectiveness and cost-effectiveness of cardiac resynchronisation (biventricular pacing) for heart failure: systematic review and economic model. Health Technol Assess 2007;11(47). Frantzen D. R&D, intersectoral and international knowledge spillovers and human capital: an empirical investigation. International Economics 2000;3(4):487–505. Frantzen D [2002a]. Cross-sector and cross-country technical knowledge spillovers and the evolution of manufacturing productivity: a panel data analysis. Economie Appliquée 2002;55:31–62. Frantzen D [2002b]. R&D spillovers and scale effects. Jahrbücher f. Nationalökonomie u. Statistik 2002;222(6):727–49. Funding First. Exceptional Returns: The economic value of America’s investment in medical research. New York: Lasker Foundation; 2000. www. laskerfoundation.org/advocacy/pdf/exceptional.pdf [accessed 15 September 2008]. Furman J. Location and organising strategy: exploring the influence of location on the organisation of pharmaceutical research. In Baum and Sorenson (2003). Furman J, Kyle M, Cockburn I. Public & Private Spillovers, Location and the Productivity of Pharmaceutical Research. NBER Working Paper 12509, 2006. Gage H, Kenward G, Hodgetts T, Castle N, Ineson N, Shaik L. Health system costs of in-hospital cardiac arrest. Resuscitation 2002;54:139–46. Gambardella A. Competitive advantages from in-house scientific research: the US pharmaceutical industry in the 1980s. Research Policy 1992;21:391–407. Garau M, Sussex J. Estimating Pharmaceutical Companies’ Value to the National Economy. Case study of the British Pharma Group. London: Office of Health Economics; 2007. Gaspoz JM, Coxson PG, Goldman PA, Williams LW, Kuntz KM, Hunink MGM, Goldman L. Cost effectiveness of aspirin, clopidogrel, or both for secondary prevention of coronary heart disease. N Engl J Med 2002;345:1800–6. Glick H, Cook J, Kinosian B, Pitt B, Bourassa MG, Pouler H, Gerth W. Costs and effects of Enalapril therapy in patients with symptomatic heart failure: an economic analysis of the Studies of Left Ventricular Dysfunction (SOLVD) treatment trial. J Cardiac Fail 1995;1:371–80. Glick H, Orzol SM, Tooley JF. Economic Evaluation of the Randomized Aldactone Evaluation Study (RALES): treatment of patients with severe heart failure. Cardiovasc Drugs Ther 2002;16:53–9. Goto A, Suzuki K. R&D capital, rate of return on R&D investment and spillovers of R&D in Japanese manufacturing industries. Review of Economics and Statistics 1989;71(4):555–64. Gournay K. Role of the community psychiatric nurse in the management of schizophrenia. Advances in Psychiatric Treatment 2000;6:243–51.

Gournay K, Brooking J. Community psychiatric nurses in primary health care. British Journal of Psychiatry 1994;165:231–8.

Jaffe A [1989a]. Real effects of academic research. American Economic Review 1989;79(5):957–70.

Gournay K, Brooking J. The community psychiatric nurse in primary care: an economic analysis. Journal of Advanced Nursing 1995;22(4):769–78.

Jaffe A [1989b]. Characterizing the ‘technological position’ of firms, with application to quantifying technological opportunity and research spillovers. Research Policy 1989;18(2):87–97.

Grant J, Cottrell R, Cluzeau F, Fawcett G. Evaluating ‘payback’ on biomedical research from papers cited in clinical guidelines: applied bibliometric study. British Medical Journal 2000;320:1107–11.

Jaffe A, Trajtenberg M, Henderson R. Geographic localization of knowledge spillovers as evidenced by patent citations. Quarterly Journal of Economics 1993;108(3):577–98.

Greenwood N, Chisholm B, Burns T, Harvey K. Community mental health team case loads and case-mix. Psychiatric Bulletin 2000;24:290–3. Griffin SC, Barber JA, Manca A, Sculpher MJ, Thompson SG, Buxton MJ, Hemingway H. Cost effectiveness of clinically appropriate decisions on alternative treatments for angina pectoris: prospective observational study. BMJ 2007;334(7594):624. Griffith R, Harrison R, Van Reenen J [2004b]. How special is the special relationship? Using the impact of US R&D spillovers on UK firms as a test of technology sourcing. IFS WP 04/32, 2004. Griffith R, Redding S, Van Reenen J [2004a]. Mapping the two faces of R&D: productivity growth in a panel of OECD industries. Review of Economics and Statistics 2004;86(4):883–95. Griliches Z. Research cost and social returns: hybrid corn and related innovations. Journal of Political Economy 1958;66;419–31. Griliches Z. Research expenditure, education, and the aggregate agricultural production function. American Economic Review 1964;54(6):961–74. Griliches Z. Issues in assessing the contribution of research and development to productivity growth. Bell Journal of Economics 1979;10:92–116.

Johnston SC, Rootenberg JD, Katrak S, Smith WS, Elkins JS. Effects of a US National Institutes of Health programme of clinical trials on public health and costs. Lancet 2006;367:1319–27. Joint Economic Committee, US Senate. The Benefits of Medical Research and the Role of the NIH. Washington, DC: United States Senate; 2000. http://hsc.utoledo.edu/research/nih_research_benefits.pdf [accessed 15 September 2008]. Jolly K, Taylor R, Lip GYH, Greenfield S, Raftery J, Mant J, Lane D, Jones M, Lee KW, Stevens A. The Birmingham Rehabilitation Uptake Maximisation Study (BRUM). Home-based compared with hospital-based cardiac rehabilitation in a multi-ethnic population: cost-effectiveness and patient adherence. Health Technol Assess 2007;11(35). Karnon J, Brennan A, Pandor A, Fowkes G, Lee A, Gray D, Coshall C, Nicholls C, Akehurst R. Modelling the long term cost effectiveness of clopidogrel for the secondary prevention of occlusive vascular events in the UK. Current Medical Research and Opinion 2005;21:101–12. Kealey T, Al-Ubaydl O. Should governments fund science? Economic Affairs 2000;20(3):4–9.

Griliches Z. The search for R&D spillovers. NBER Working Paper 3768, 1991. Griliches Z, Lichtenberg F. R&D and productivity growth at the industry level: is there a relationship? In Z Griliches (ed.). R&D, Patents and Productivity. Chicago: NBER and University of Chicago Press; 1984. Guellac D, de la Potterie B. R&D and productivity growth: panel data analysis of 16 OECD countries. DSTI Working paper 2001/3. Paris: Organisation for Economic Co-operation and Development; 2001. Hanney S, Frame I, Grant J, Buxton M, Young T, Lewison G. Using categorisations of citations when assessing the outcomes from health research. Scientometrics 2005;65(3):357–79. Hanney S, Grant J, Wooding S, Buxton M. Proposed methods for reviewing the outcomes of research: the impact of funding by the UK’s Arthritis Research Campaign. Health Research Policy and Systems, 2004;2:4. www. health-policy-systems.com/content/2/1/4 [accessed 15 September 2008]. Hanney S, Home P, Frame I, Grant J, Green P, Buxton M. Identifying the impact of diabetes research. Diabetic Medicine 2006;23:176–84.

Keller W. Are international R&D spillovers trade-related? Analysing spillovers among randomly matched trade partners. European Economic Review 1998;42(8):1469–81. Keller W [2002a]. Trade and the transmission of technology. Journal of Economic Growth 2002;7:5–25. Keller W [2002b]. Geographic localization of international technology diffusion. American Economic Review 2002;92(1):120–42. Keller W. International technology diffusion. Journal of Economic Literature 2004;42:752–82. Kendrick T, Peveler R, Longworth L, Baldwin D, Moore M, Chatwin J, Thornett A, Goddard J, Campbell M, Smith H, Buxton M, Thompson, C. Cost-effectiveness and cost-utility of tricyclic antidepressants, selective serotonin reuptake inhibitors and lofepramine. British Journal of Psychiatry 2006;188:337–45. King DA. The scientific impact of nations. Nature 2004;430:311–6. Kirsch I, Deacon BJ, Huedo-Medina TB, Scoboria A, Moore T, Johnson B. Initial severity and antidepressant benefits: a meta-analysis of data submitted to the Food and Drug Administration. PLoS Med 2008;5:e45.

Hannigan B. Mental health nursing in the community: a United Kingdom perspective. Home Health Care Management & Practice 1999;11:60–6.

Knutson M, Tweeten LG. Toward an optimal rate for growth in agricultural production research and extension. American Journal of Agricultural Economics 1979;61:70–6.

Hartwell D, Colquitt J, Loveman E, Clegg AJ, Brodin H, Waugh N, Royle P, Davidson P, Vale L, MacKenzie L. Clinical effectiveness and costeffectiveness of immediate angioplasty for acute myocardial infarction: systematic review and economic evaluation. Health Technol Assess 2005;9(17).

Krugman P. Geography and Trade. Cambridge, MA: MIT Press; 1991. Laatikainen T, Critchley J, Vartiainen E, Salomaa V, Ketonen M, Capewell S. Explaining the decline in coronary heart disease mortality in Finland between 1982 and 1997. American Journal of Epidemiology 2005;162(8):764–73.

Heidenreich P, McClellan M. Biomedical research and then some: the causes of technological change in heart attack treatment. In K Murphy, R Topel (eds). Measuring the Gains from Medical Research: An economic approach. Chicago, IL: University of Chicago Press; 2003. pp 163–205.

Levin R, Reiss P. Tests of a Schumpeterian model of R&D and market structure. In Z Griliches (ed.). R&D, Patents and Productivity. Chicago: NBER and University of Chicago Press; 1984. pp. 175–204.

Henderson R, Cockburn I. Scale, scope, and spillovers: determinants of research productivity in the pharmaceutical industry. RAND Journal of Economics 1996;27(1):32–59.

Link AN, Rees J. Firm size, university-based research and the returns to R&D. Small Business Economics 1990;2:24–31.

Huffman WE, Evenson RE. Science for Agriculture: A long term perspective. Ames, Iowa: Iowa State University Press; 1993.

Lucas R. On the mechanisms of economic development. Journal of Monetary Economics 1988;22(1):3–42.

Hughes A. University-Industry Linkages and UK Science and Innovation Policy. Centre for Business Research, University Of Cambridge, Working Paper 326, 2006.

Luce B, Mauskopf J, Sloan F, Ostermann J, Paramore LC. The return on investment on health care: 1980–2000. Value in Health 2006;9:146–56.

Jacob R, McGregor M. Assessing the impact of health technology assessment. International Journal of Technology Assessment in Health Care 1997;13(1):68–80. Jaffe A. Technological opportunities and spillovers of R&D: evidence from firms’ patents, profits and market value. American Economic Review 1986;76:984–1001.

Maddock I. The Seventh Royal Society Technology Lecture: Science, technology and industry. Proceedings of the Royal Society of London Services A, Mathematical and Physical Sciences 1975;345(1642):295–326. Main C, Palmer S, Griffin S, Jones L, Orton V, Sculpher M, Henderson R, Sudlow C, Hawkins N, Riemsma R. Clopidogrel used in combination with aspirin compared with aspirin alone in the treatment of non-STs egmentelevation acute coronary syndromes: a systematic review and economic evaluation. Health Technol Assess 2004;8(40).

Mansfield E. Academic research and industrial innovation. Research Policy 1991;20(1):1–12. Mansfield E. Academic research underlying industrial innovations: sources, characteristics and financing. Review of Economics and Statistics 1995;77:55–65. Mansfield E. Academic research and industrial innovation: an update of empirical findings. Research Policy 1998;26:773–6. Mansfield E, Rapoport A, Romeo E, Villani E, Wagner S, Husic F. The Production and Application of New Industrial Technology. New York: W W Norton and Company; 1977. Martin B, Tang P. The benefits from publicly funded research. SPRU WP 161, 2007. McKeown T. The Role of Medicine. Oxford: Basil Blackwell; 1979. Middleton N, Gunnel D, Whitely E, Dorling D, Frankel S. Secular trends in antidepressant prescribing in the UK, 1975–1998. Journal of Public Health Medicine 2001;23(4):262–7. MINAP Steering Group. Myocardial Infarction National Audit Project (MINAP): How the NHS manages heart attacks. Sixth Public Report 2007. London: Royal College of Physicians; 2007. Moed HF, De Bruin RE, van Leeuwen TN. New bibliometric tools for the assessment of national research performance: database description, overview of indicators and first applications. Scientometrics 1995;33(3):381–422. Moncayo A. Chagas disease: current epidemiological trends in the Southern Cone countries. Memorias do Instituto Oswaldo Cruz 2003;98:577–91. Murphy K, Topel R (eds). Measuring the Gains from Medical Research: An economic approach. Chicago: University of Chicago Press; 2003. Mushkin S. Biomedical Research: Costs and benefits. Cambridge, MA: Ballinger; 1979. Nadiri MI. Innovations and technological spillovers. NBER Working Paper 4423 (CC), 1993. Naess AC, Steen PA. Long term survival and costs per life year gained after out-of-hospital cardiac arrest. Resuscitation 2004;60:57–64. National Collaborating Centre for Chronic Conditions. Hypertension: Management in adults in primary care: pharmacological update. London: Royal College of Physicians; 2006. National Health Service. National Service Framework for Mental Health. London: Department of Health; 1999. National Institutes of Health. Cost Savings Resulting from NIH Research Support. 2nd edn. Bethesda, MD: US Dept. of Health and Human Services, NIH; 1993. National Statistics. Key Health Statistics from General Practice 1998. Series MB6 no. 2. London: National Statistics; 2000. National Statistics. Psychiatric Morbidity among Adults Living in Private Households 2000. London: National Statistics; 2001. National Statistics. Patient Care in the Community. NHS community mental health nursing, 2003–04. London: National Statistics; 2004. Natural Environment Research Council. Economic Benefits of Environmental Science. Swindon: NERC; 2006. www.nerc.ac.uk/publications/corporate/ economic-fullreport.asp [accessed 15 September 2008]. Nelson R. Institutions supporting technical advance in industry. American Economic Review 1986;76:186–9. NERA Economic Consulting. Key Factors in Attracting Internationally Mobile Investments by the Research-based Pharmaceutical Industry. A report prepared for the UK Trade and Investment and the Association of the British Pharmaceutical Industry. London: NERA; 2007. www.nera.com/image/PUB_ MobileInvestments_Sep2007.pdf [accessed 15 September 2008]. NICE. Guide to the Methods of Technology Appraisal. London: NICE; 2004. Nordhaus WD. The health of nations: the contribution of improved health to living standards. In Murphy and Topel (2003). Norris RM. Fatality outside hospital from acute coronary events in three British health districts, 1994–5. United Kingdom Heart Attack Study Collaborative Group. BMJ 1998;316:1065–70. O’Brien CL, Gage BF. Costs and effectiveness of Ximelagatran for stroke prophylaxis in chronic atrial fibrillation. JAMA 2005;293:699–706.

OECD. Purchasing Power Parities (PPP). OECD Statistics Directorate; 2008. www.oecd.org/department/0,3355,en_2649_34357_1_1_1_1_1,00.html [accessed 15 September 2008]. Office for National Statistics. Annual Abstract of Statistics 2007 Edition. No. 143. London: Office for National Statistics; 2007. Office for National Statistics. General Household Survey. London: Office for National Statistics; 2008. www.statistics.gov.uk/ssd/surveys/general_ household_survey.asp [accessed 15 September 2008]. Owen-Smith J, Powell W. Knowledge networks as channels and conduits: the effects of spillovers in the Boston biotechnology community. Organization Science 2004;15(1):5–21. Palmer S, Sculpher M, Philips Z, Robinson M, Ginnelly L, Bakhai A, Abrams K, Cooper N, Packham C, Alfakih K, Hall A, Gray D. Management of nonST-elevation acute coronary syndromes: how cost-effective are glycoprotein IIb/IIIA antagonists in the UK National Health Service? Int J Cardiol 2005;100:229–40. Park J. International and intersectoral R&D spillovers in the OECD and East Asian economies. Economic Inquiry 2004;42(4):739–57. Peterson W. Return to poultry research in the United States. Journal of Farm Economics 1967;49:656–69. Peto R, Lopez AD, Boreham J, Thun M. Mortality from Smoking in Developed Countries 1950–2000. Second edition. 2006. www.deathsfromsmoking.net [accessed 15 September 2008]. Pharmaceutical Industry Competitiveness Task Force (PICTF). Value of the Pharmaceutical Industry to the UK Economy. London: Office of Health Economics; 2001. www.advisorybodies.doh.gov.uk/pictf/value.pdf [accessed 15 September 2008]. Phillips KA, Shlipak MG, Coxson P, Heidenreich PA, Hunink MGM, Goldman PA, Williams LW, Weinstein MC, Goldman L. Health and economic benefits of increased -blocker use following myocardial infarction. JAMA 2000;284:2748–54. Pisano G. Can science be a business? Lessons from biotech. Harvard Business Review 2006;84(10):114–25. Powell WM. Neither market nor hierarchy: network forms of organisation. Research in Organisations 1990;12:293–336. Raiten D, Berman S. Can the Impact of Basic Biomedical Research be Measured? A case study approach. Bethesda, MD: Life Sciences Research Office, FASEB; 1993. Romer P. Increasing returns and long-run growth. Journal of Political Economy 1986;94(5):1002–37. Rosenbeck R, Leslie D, Sindelar J, Miller EA, Stroup S, McEvoy J, Davis S, Davis CE, Keefe RSE, Swartz M, Perkins D, Hsiao JK, Lieberman J. Costeffectiveness of second-generation anti-psychotics and perphenazine in a randomized trial of treatment for chronic schizophrenia. American Journal of Psychiatry 2006;163:2080–9. Royal Netherlands Academy of Arts and Sciences. The Societal Impact of Applied Research: Towards a quality assessment system. Amsterdam: Royal Netherlands Academy of Arts and Sciences; 2002. Salter A, D’Este P, Martin B, Geuna A, Scott A, Pavitt K, Patel P, Nightingale P. Talent, not Technology: Publicly funded research and innovation in the UK. Brighton: SPRU; 2000. www.sussex.ac.uk/spru/documents/ saltertalentshort.pdf [accessed 15 September 2008]. Salter A, Martin B. The economic benefits of publicly funded basic research: a critical review. Research Policy 2001;30:509–32. Samsa GP, Matchar DB, Williams GR, Levy DE. Cost-effectiveness of ancrod treatment of acute ischaemic stroke: results from the Stroke Treatment with Ancrod Trial (STAT). J Eval Clin Prac 2002;8:61–70. Sandercock P, Berge E, Dennis M, Forbes J, Hand P, Kwan J, Lewis S, Lindley R, Neilson A, Thomas B, Warlaw J. A systematic review of the effectiveness, cost-effectiveness and barriers to implementation of thrombolytic and neuroprotective therapy for acute ischaemic stroke in the NHS. Health Technol Assess 2002;6(26). Scherer FM. Interindustry technology flows and productivity growth. Review of Economics and Statistics 1982;64:627–34. Scherer FM. Using linked patent and R&D data to measure interindustry technology flows. In Z Griliches (ed.). R&D, Patents and Productivity. Chicago: NBER and University of Chicago Press; 1984.

Schmitz A, Seckler D. Mechanized agriculture and social welfare: the case of the tomato harvester. American Journal of Agricultural Economics 1970;52:567–77. Scott J, Palmer S, Paykel E, Teasdale J, Hayhurst H. Use of cognitive therapy for relapse prevention in chronic depression. British Journal of Psychiatry 2003;182:221–7.

Vermeer F, Simoons ML, de Zwaan C, van Es GA, Verheugt FW, van der Laarse A, van Hoogenhuyze DC, Azar AJ, van Dalen FJ, Lubsen J. Cost benefit analysis of early thrombolytic treatment with intracoronary streptokinase. Twelve month follow up report of the randomised multicentre trial conducted by the Interuniversity Cardiology Institute of The Netherlands. Br Heart J 1988;59:527–34. Von Hipple E. Sticky information and the locus of problem solving, implications for innovation. Management Science 1984;40:429–39.

Silverstein S, Garrison HH, Heinig SJ. A few basic economic facts about research in the medical and related life sciences. FASEB Journal 1995;9:833–40. Solow K. Technical change and the aggregate production function. Review of Economics and Statistics 1957;39(3):312–20. Stuart T, Sorenson O. The geography of opportunity: spatial heterogeneity in founding rates and the performance of biotechnology firms. Research Policy 2003;32:229–53.

Wang D, Connock M, Barton P, Fry-Smith A, Aveyard P, Moore D. ‘Cut down to quit’ with nicotine replacement therapies in smoking cessation: a systematic review of effectiveness and economic analysis. Health Technology Assessment 2008:12(2). Ward MR, Dranove D. The vertical chain of research and development in the pharmaceutical industry. Economic Inquiry 1995;33:70–87. Ward S, Lloyd Jones M, Pandor A, Holmes M, Ara R, Ryan A, Yeo W, Payne N. A systematic review and economic evaluation of statins for the prevention of coronary events. Health Technology Assessment 2007;11(14).

Sussex J (ed.). Improving Population Health in Industrialised Nations. London: OHE; 2000. Sveikauskas L. Technological inputs and multifactor productivity growth. Review of Economics and Statistics 1981;63(2):275–82. Swedberg K, Kjekshus L. Effects of enalapril on mortality in severe congestive heart failure: results of the Cooperative North Scandinavian Enalapril Survival Study (CONSENSUS). Am J Cardiol 1988;62(2):60A–66A. Toole A. The impact of public research and industrial innovation: evidence from the pharmaceutical industry. SIEPR Discussion Paper no. 00-07. Stanford Institute for Economic Policy Research; 2000.

Wardlaw JM, Seymour J, Cairns J, Keir S, Lewis S, Sandercock P. Immediate Computed Tomography scanning of acute stroke is cost-effective and improves quality of life. Stroke 2004;35:2477–83. Warry P. Increasing the Economic Impact of Research Councils. Research Council Economic Impact Group (Warry Report). London: Department of Trade and Industry; 2006. Weisbrod B. Economics and Medical Research. Washington DC: American Enterprise Institute; 1983.

Toole A. Does public scientific research complement private investment in research and development in the pharmaceutical industry? Journal of Law and Economics 2007;50:81–104.

Wellcome Trust. Putting NHS Research on the Map: An analysis of scientific publications in England, 1990–97, London: Wellcome Trust; 2001. www. wellcome.ac.uk/About-us/Publications/Books/Biomedical-science/ WTD003194.htm [accessed 15 September 2008].

Tsevat J, Duke D, Goldman L, Pfeffer MA, Lamas GA, Soukup JR, Kuntz KM, Lee TH. Cost-effectiveness of captopril therapy after myocardial infarction. J Am Coll Cardiol 1995;26:914–9.

Wooding S, Hanney S, Buxton M, Grant J. Payback arising from research funding: evaluation of the Arthritis Research Campaign. Rheumatology 2005;44:1145–56.

Tunstall-Pedoe H, Bailey L, Chamberlain DH, Marsden AK, Ward ME, Zideman DA. Survey of 3765 cardiopulmonary resuscitations in British hospitals (the BRESUS study): methods and overall results. BMJ 1992;304:1347–51. UK Evaluation Forum Medical Research: Assessing the benefits to society. London: Academy of Medical Sciences; 2006. www.acmedsci.ac.uk/index. php?pid=48&prid=1 [accessed 15 September 2008].

Zhechun Z, Critchley J, Capewell S. S-IMPACT: A validated, comprehensive ischemic stroke model. Supplementary appendix: general methods. Liverpool: University of Liverpool; 2006.

UK Medicines Information (UKMI). Clopidogrel in acute coronary syndrome. Monograph Number 3/02/01. NHS collaboration of the United Kingdom Medicines Information Pharmacists’ Group and The National Prescribing Centre; 2002. Unal B, Critchley J, Capewell S. Explaining the decline in coronary heart disease mortality in England and Wales, 1981–2004. Circulation 2004;109:1101–7. Unal B, Critchley J, Capewell S. Modelling the decline in coronary heart disease deaths in England and Wales, 1981–2004: comparing contributions from primary prevention and secondary prevention. British Medical Journal 2005;331:614. United Kingdom Heart Attack Study (UKHAS) Collaborative Group. Effect of time from onset to coming under care on fatality of patients with acute myocardial infarction: effect of resuscitation and thrombolytic treatment. Heart 1998;80:114–20. Van Leeuwen TN, Moed HF, Tijssen RJW, Visser M, Van Raan A. Language biases in the coverage of the Science Citation Index and its consequences for international comparisons of national research performance. Scientometrics 2001;51(1):335–46. Van Leeuwen TN, Visser MS, Moed HF, Nederhof AJ, Van Raan AFJ. The Holy Grail of science policy: exploring and combining bibliometric tools in search of scientific excellence, Scientometrics 2003;57(2):257–80. Van Roijen LH, Van Straten A, Al M, Rutten F, Donker M. Cost-utility of brief psychological treatment for depression and anxiety. British Journal of Psychiatry 2006;188:323–9. Varney S. A cost-effectiveness analysis of bisoprolol for heart failure. Eur J Heart Fail 2001;3:365–71. Velligan D, Lam F, Ereshefsky L, Miller A. Perspectives on medication adherence and atypical antipsychotics medications. Psychopharmacology 2003;54(5):665–7.

TAP825/27-10-08/V4c/700

• Staff intranet
• Student intranet

What do you want to do?

Medical research – just what's it worth exactly.

Public bodies, lobby groups and medical research charities have all benefited from important work to assess just how much impact health research has on society.

Brunel’s Health Economics Research Group (HERG) has developed the payback framework technique over recent years – giving research –funding groups the chance to show policymakers and the public exactly what their investments have achieved.

HERG’s 2008 report, ‘Medical Research: What’s it Worth’, has had a massive impact on policy and public spending. It has often been cited as a vital factor in convincing the Government to raise the health research budget in 2010 and to protect the science budget in its 2010 and 2013 Spending Reviews.

Elsewhere, a number of major medical research funding organisations, including Cancer Research UK, used the 2008 report as key evidence in their own response to the Spending Review.

Charities including Asthma UK and the National Breast Cancer Foundation in Australia have funded HERG to lead independent and methodological evaluations of the impacts of their research funding, helping to significantly raise the profile of the latter and give it the encouragement to substantially boost its investment in research.

Meanwhile, extensive dissemination has helped to publicise the full range of the group’s impact studies, including publications and presentations in Canada, the EU, Ireland and the UK.

On Medicine

What’s it worth? The economic case for medical research

Guest 16 Jun 2014

This has been reposted , with kind permission from the author and the Wellcome Trust.

What’s it worth, a report published today in BMC Medicine , is one of the first ever estimates of the economic gains from investment in publicly funded UK cancer research. The research was commissioned by the Wellcome Trust, Academy of Medical Sciences, Cancer Research UK and the Department of Health. Liz Allen, Head of Evaluation at the Wellcome Trust, argues the case for investing in medical research…

Bill Clinton achieved a lot in the White House. He presided over the longest period of peacetime economic growth in American history, he signed the North American Free Trade Agreement, and he was the first Democrat since FDR to win re-election. Yet when asked last month to recall his greatest accomplishment, he chose none of these things. The best thing he did as President, he told the Inclusive Capital Conference in London this year, was spending $3bn on the Human Genome Project.

The pride of place that Clinton gives to the sequence of humanity’s genetic code derives in large part from the way it has already contributed to great advances in treating cancer and rare inherited conditions, and the promise of personalised medicine. Its contribution to health, however, was not the reason he gave. He was proudest of the genome’s role in generating an entirely new industry that has become a fresh source of economic growth. That $3bn investment, by governments and charities, has brought a vast return: more than $180 billion, Clinton said, is now invested in genome companies in the US alone.

Much of that return has its foundations in British investment in genomics, and work by British scientists and the Wellcome Trust Sanger Institute. A good example is Solexa, a sequencing technology company spun out of research by Shankar Balasubramanian and David Klenerman in the mid-90s, which sold for $600 million in 2006.

Similarly, research by Mike Stratton at the Wellcome Trust Sanger Institute and Richard Marais at the Institute of Cancer Research led to the discovery of an unexpected target for drug therapy, mutations in a gene called BRAF present in malignant melanoma. The research resulted in the pharmaceutical giant Roche developing Zelboraf, a drug targeted against BRAF mutations which is now widely available to patients with advanced melanoma.

The money invested in sequencing the human genome is an example of how, when we invest in medical research, we win twice over. We make discoveries that improve human health, but we also make a substantial financial return.

Research published today in BMC Medicine provides the latest evidence that there is a strong economic argument for continuing this investment. The authors estimate that for every pound invested in cancer-related medical research there is a direct economic return of 10 pence year on year , far in excess of the UK Government’s stated minimum threshold of 3.5 pence per pound for investments.

On top of that, there is an estimated additional 30 pence of yearly return per pound in ‘spillover’ benefits, the indirect positive financial impact of public and charitable research investment – including increasing employment and leveraging private sector R&D activity. All in all, the study claims a weighty 40 pence return for every pound spent on medical research.

The research also highlights that health and economic benefits take time to come to fruition- 15 years on average – and we must invest with an eye on the long term. For every breakthrough, there are false starts and incremental, less startling findings, but they’re equally part of the process of science and we should be open about them. Funding is not the only ingredient needed to achieve results; curiosity, risk taking, persistence and some serendipity are just as vital.

Take the example of IVF. Pioneering research in the early 1960s by British scientists Robert Edwards and Patrick Steptoe led to the successful delivery of the first test tube baby, Louise Brown, in Oldham in 1978. Since Louise, there have been more than five million babies born using IVF technology around the world. In the UK alone the IVF industry is worth an estimated £500 million.

Another success story is monoclonal antibodies – the basis for a class of drugs that treat everything from breast cancer to cardiovascular disease to psoriasis. The industry was recently valued at £250 billion, but the breakthrough in the science came in the late 1970s on the back of Medical Research Council-funded research in Cambridge by Greg Winter.

In sum, medical research is that rarest of beasts in public spending – a genuine investment that can almost be guaranteed to pay off over time, in wealth as well as health. We don’t know where the next discovery will come from, but spending money on our intellectual infrastructure now ensures that the research system will continue to produce a return 15 or 20 years from now.

As all the parties plan their manifestos for a parliamentary term that will surely include continuing austerity measures, a decision to cut back investment in medical research would be a terrible mistake. And a truly false economy.

You can read more about the Wellcome Trust’s involvement with the ‘What’s it worth?’ report on their website , and you can find out more about the report published today in our guest post from Cancer Research UK .

View the latest posts on the On Medicine homepage

• Latest Posts

Latest posts by Guest ( see all )

• When culture meets epidemic: the case of Ebola - 23rd April 2015
• Sleep in the intensive care unit – it’s important, right? - 23rd March 2015
• Withdrawal of medicines that cause deaths: takes longer than you think - 5th February 2015

Recommended posts

Popular on medicine tags.

• BMC Medicine
• clinical trials
• medical evidence
• Genome Medicine
• ISRCTN Registry
• Alzheimer's
• Alzheimer's Research & Therapy
• breast cancer

Popular posts

• Most shared

Sorry. No data so far.

Most Shared Posts

• World Suicide Prevention Day 2023: the Lifeguard Pharmacy study
• Brain Tumor Awareness Month: Can proton beam therapy improve the long-term quality of life of patients with oligodendroglioma brain tumors?
• Winners of the 2022 Cardio-Oncology Journal Prize
• A unique approach. A unique population. Investing in nurses’ mental health
• Does the midwife-led continuity of carer model improve birth outcomes and maternal mental health in vulnerable women?
• March 2024  (1)
• February 2024  (2)
• January 2024  (1)
• November 2023  (1)
• October 2023  (1)
• September 2023  (7)
• July 2023  (1)
• June 2023  (1)
• May 2023  (4)
• April 2023  (1)
• March 2023  (2)
• February 2023  (7)

Browse Econ Literature

• Working papers
• Software components
• Book chapters
• JEL classification

More features

• Subscribe to new research

RePEc Biblio

Author registration.

• Economics Virtual Seminar Calendar NEW!

Medical Research: What's It Worth?

• Author & abstract
• 1 Citations
• Related works & more

Corrections

• Office of Health Economics

Suggested Citation

Download full text from publisher.

Follow serials, authors, keywords & more

Public profiles for Economics researchers

Various research rankings in Economics

RePEc Genealogy

Who was a student of whom, using RePEc

Curated articles & papers on economics topics

Upload your paper to be listed on RePEc and IDEAS

New papers by email

Subscribe to new additions to RePEc

EconAcademics

Blog aggregator for economics research

Cases of plagiarism in Economics

About RePEc

Initiative for open bibliographies in Economics

News about RePEc

Questions about IDEAS and RePEc

RePEc volunteers

Participating archives

Publishers indexing in RePEc

Privacy statement

Found an error or omission?

Opportunities to help RePEc

Get papers listed

Have your research listed on RePEc

Open a RePEc archive

Have your institution's/publisher's output listed on RePEc

Get RePEc data

Use data assembled by RePEc

Wellcome uses cookies.

Medical research, what’s it worth : estimating the economic benefits from medical research in the uk : briefing document., available online.

Attribution 4.0 International (CC BY 4.0)

You can use this work for any purpose, including commercial uses, without restriction under copyright law. You should also provide attribution to the original work, source and licence. Read more about this licence .

Medical Research, what’s it worth? : estimating the economic benefits from medical research in the UK : briefing document . Attribution 4.0 International (CC BY 4.0) . Source: Wellcome Collection.

About this work

Description.

Following a study published in 2008 which yielded the first quantitative assessment of the economic benefit of biomedical and health science in the UK, this new study was commissioned to develop the methodology and calculate the economic benefits of investment in cancer-related research by the taxpayer and charities in the UK. The results of this study strongly suggest that biomedical research provides a sizeable return on investment.

Publication/Creation

Physical description, contributors.

• Cancer Research UK sponsoring body
• Great Britain. Department of Health sponsoring body
• Wellcome Trust (London, England) sponsoring body
• Health Economics Research Group (Brunel University) author
• Office of Health Economics (London, England) author
• King's College London author
• RAND Europe author

Bibliographic information

Type/technique.

• Electronic books
• Biomedical Research economics
• Great Britain

Permanent link

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Int J Prev Med
• v.2(4); Oct-Dec 2011

Evaluation of Health Research: Measuring Costs and Socioeconomic Effects

Kerstin roback.

1 Center for Medical Technology Assessment, Department of Medical and Health Sciences, Linköping University, Sweden

Koustuv Dalal

Per carlsson, objectives:.

The topic of this work is health research evaluation including basic and clinical medical research, as well as healthcare research. The main objects are to explore possible approaches for valuing research in economic terms and to prepare an analytical model for evaluation of health research using the Swedish context. The study also aims to identify potential effects and their significance, and to provide a basis for discussions about the effects of research investments.

The study has reviewed ten articles indicating positive effects, in the form of improved health and economic growth. The study also developed a model applied to Swedish health research.

The review indicates that positive effects, in the form of improved health and economic growth, have a value that greatly exceeds the costs of the research investments. The tentative model applied to Swedish health research also indicates predominantly positive returns, but in a lower range than the review would imply. Methodological problems, however, entail major uncertainty in the cited results.

Conclusions:

Accurate determination of the economic value of research would require significantly better basic data and better knowledge of relationships between research, implementation of new knowledge, and health effects. Information in support of decisions about future allocation of research resources is preferably produced by a combination of general analyses and strategically selected case studies.

INTRODUCTION

The issue of increased healthcare demands is well known in connection with the demographic transitions of health problems such as from younger to older generations and from communicable to non-communicable diseases. Healthcare system is burdened by a rise in infectious, lifestyle and chronic degenerative diseases with elevated threats of environmental health hazards. New and hard-to-treat viruses and microbes such as HIV and Methicillin-Resistant Staphylococcus aureus (MRSA) have already challenged the research community and old public health threats such as tuberculosis reappeared with changes in living conditions. Diseases of which we have little knowledge at the local level are spreading throughout the world, and may require further readiness in the research community for new challenges. As a whole, the situations demand a strong academic knowledge and context dependent research readiness in relation to geographical and cultural issues. However, research funding should not be viewed solely as a means of averting health threats; it must also be considered as an intervention that has both positive and negative health effects as well as other societal effects.

The main purpose of health research, covering basic and clinical traditional medical research, is to improve the health of the general population in the form of better quality of life and increased longevity. More health research leads to better health, which eventually results in productivity gains. Such effects can be estimated in research evaluations. Evaluation of health research is getting more and more important as such evaluations should not only provide information to the sceptic policy makers, but also to the research community itself. Many questions arise in this context: Is increased research funding justified from a national perspective? Will the investment be covered by the potential future return in the form of improved health and greater economic growth? Will the healthcare provided be of significantly higher quality in those countries that proportionally invest the most in relation to their economies? How much is it reasonable for a small country to contribute to the collective global knowledge base?

There is lack of models for accurate evaluation of all significant socio-economic effects pertaining to health research. The aims of the current work are to explore possible approaches for valuing research and to prepare an analytical model for evaluation of health research using the Swedish context. The analysis is an attempted to clearly identify potential effects and estimate their significance. The paper is intended to provide a basis for discussing how the effects of research investments can be valued and which effects should be included.

REVIEW OF THE HEALTH RESEARCH EVALUATION LITERATURE

A literature search for health research evaluation studies (1985-Sept 2009) was performed using Pubmed. The following sets of keywords were used in different combinations: (in the title) clinical/medical/biomedical/health research; impact, value, return/s, benefits, gains; funding, economy, economics and (in all fields) evaluation, evaluating, assessment/s, assessing, measuring and estimating. A language restriction was made to documents in English and Swedish.

The Pubmed search yielded 491 references. The rather broad search resulted in 375 non-relevant titles. For the remaining 116 relevant titles, abstracts were read and 49 articles that had analysed or discussed effects of health related research were retrieved. Documents from an international workshop at Sigtuna, Sweden, 2007 and 2009 were also added and reference lists were screened for references relevant to the current study. Final inclusion criteria were: original studies that investigated societal effects of health research including broader societal benefits, health benefits and/or survival effects. The final screening yielded 10 articles which met these criteria and were included into the current review ( Table 1 ).

Categorization of the literature

Different Approaches for Valuing Health-Related Research

The review of the selected ten articles covers different approaches for valuing medical research. Broadly, the approaches are: Disease- and technology-specific case studies, Case studies of specific research funding, General socio-economic approaches, and combinations of these three approaches. According to this categorization, the articles are presented in table 2 .

Studies in the review divided into different categories

Disease- and technology-specific case studies

In this approach, the magnitude of the healthcare investments in the studied diseases is compared to the gains achieved in terms of quality of life, survival, and any savings in terms of treatment costs. Another related approach is to track significant medical innovations to see the impact they have had on health improvements, and what research and development costs they have entailed.

Case studies of specific research funding

It is becoming increasingly common for research financiers to want to evaluate what is achieved as a result of the allocated funding. Analyses of the health economic and socio-economic effects are also increasingly sought after. Many studies consequently consist of case studies of research projects in which attempts are made to find links between the financed projects and health gains in society. These studies are also often targeted to specific disease groups.

General socio-economic approach

Health gains in society can have many causes above and beyond the triumphs of medical research, such as improved diet and living conditions, safer work environments and healthier lifestyles. It is not the case that every health gain can be attributed to health-related research; on the other hand, though, medical research also produces non-medical social effects such as increased productivity, greater competitiveness and economic growth. Health-related research thus contributes indirectly to a country's GDP.

In absence of some vital information it is difficult to calculate the net value of the social effects.

Studies that take a completely economic approach must consequently be based on rough estimates. The authors of such studies are cautious in drawing their conclusions, and are more inclined toward discussing the reasonability of the values they derive. Still, the analyses in these types of studies consistently indicate that the socio-economic benefits broadly exceed the research costs. Those studies also admit that the estimate of increased longevity and changes in medical expenditures include contributions from other sources, but argues that this has a marginal influence on the calculated values.

A Combined Approach

One study has used a combined approach for valuing health research. In the first instance it analysed the health gain due to increases in healthcare expenditures. Next, it analysed the value of advances in treatments for specific diseases such as heart attack, stroke, type-2 diabetes, and breast cancer. Finally, it calculated the value of specific innovations using published cost-effectiveness data.

ANALYTICAL MODEL FOR SWEDISH HEALTH RESEARCH

The above review indicates that studies of individual research projects and individual innovations in a country must be complemented by studies conducted from a broad socio-economic perspective which reduces the risk of double-counting and makes it possible to include the positive effects that research has in terms of growth and innovation in a society.

The suggested analytical model identifies relevant investments and effects. We have calculated a monetary value wherever possible, and provide a qualitative description in other cases. The evaluated investment (intervention) pertains to Sweden's total current investment in funding health-related research. The benefits of this intervention can thus be compared directly to the costs in order to obtain some indication of the net effect.

In the analysis, the “research financing” intervention refers to the total of private, municipal, and state expenditures for clinical research and associated health-related research conducted in Sweden. The size of the investment is calculated as the sum of the 2005 research expenditures in Sweden in the areas of healthcare, medicine, physiology, and pharmacology. The effects of the intervention are estimated for the year, 2015, based on forecasts for a ten-year period. Costs and effects arise at different times and hence monetary sums (where the input values are expressed in current prices) are converted to 2005 levels. Further discounting of future costs and effects has not been performed.

Health-related research has an effect on every citizen but the effects impact different parts of the societal structure differently. The benefits to the citizen in general come mainly in the form of improvements in quality of life and additional life-years, while the benefits to the economy may consist of higher net exports or increased sales, which in turn can contribute to corporate profits. Locally, at the individual municipality level, it may be in the form of spin-off effects in the business community. Certain effects of Swedish research will naturally extend beyond the country's borders, while there is at the same time an influx of effects from the outside world. To study the effects of Swedish research funding on Swedish society, we must thus estimate how great a share of the effects can be attributed to investments made domestically, as we can include only that part of the effects which can be ascribed to the studied intervention.

In many studies of the societal effects of research, the choice is made to study the current effects of a resource investment that was made a number of years earlier. On the other hand, we have chosen an ex-ante approach; that is, an estimation of what we believe can be achieved in the future as a result of the research being done today. We do this partly because, historically speaking, changes in the medical field have been sweeping and revolutionary. New therapies and new health threats are emerging constantly. An ex-ante approach enables us to incorporate recent historical trends in the analysis, and to take into account current changes of which we are aware. Furthermore, our analysis is intended to shed light on decisions about investments in health research. Decision-makers may therefore have an interest in a forecast, and in the possibility of influencing a future outcome. We have consequently estimated future effects based on the effects observed in modern times, transferred to a future society that has evolved in a manner consistent with current trends.

Identification of Relevant Costs and Effects

First on the cost side are the financial resources made available by various research funders. Other costs arise as a result of these investments, such as unremunerated research work and input from the people who participate in clinical studies. Research can also lead to more expensive treatments, troublesome side effects, and environmental impacts. Costs also arise whenever new methods are integrated into healthcare; there are costs associated with breaking with ingrained behaviours, and reorganisations that temporarily lower productivity may be necessary. Finally, medical innovations are often added to pre-existing methods, leading to greater complexity that may have negative effects.

The financial benefits from an investment in health-related research consist largely in the avoidance of certain costs associated with illness. The social costs of illness include the costs of diagnosis, treatment, recovery, and disease prevention; these costs can be kept down by expanding our knowledge of various diseases, and by introducing new and more effective care methods. Disease prevention simultaneously produces social and emotional gains in terms of improved health, improved quality of life and additional life-years. Process improvements also occur in healthcare as a result of greater knowledge and better medical methods. Such improvements can appear in the form of, for instance, shorter treatment times, less discomfort, and improvements in the work environment.

Two different types of production gains occur in the economy; the avoidance of production losses as the result of a healthier population 11 , 12 and increased production 13 , 14 in the form of higher employment in various sectors of the economy, including healthcare and research. Increases in productivity can also arise as a result of the generally greater impetus to innovate that derives from new research findings, and as a result of the fostering of positive attitudes toward change that emerges from the encounter between research and healthcare.

Finally, research also has repercussions in the research world, in that the general body of knowledge grows; and this should be valued as an investment in future research and the development of innovations. 11 , 14 Salter and Martin assert that a country's research activities constitute an “admission ticket” to sharing in the global knowledge base. 14 The conversion of research findings into medical innovations is becoming an increasingly international process, and a high-quality domestic research base is a prerequisite for establishing enterprises that can participate in global R & D projects intended to generate knowledge and create technologies in the medical field.

Once the relevant costs and benefits have been identified, they must be measured and valued. A large part of the effects results from research conducted by research institutions outside Sweden. According to statistics from the Global Innovation Scoreboard, Sweden incurs 1.42% of the total worldwide R & D expenditure. 15 An analysis of medical publications from various countries indicates that Sweden's share of health-related research is actually larger, at 1.93%. 16 Our impact on medical advancements is presumably even greater, given that Sweden have a relatively large domestic pharmaceutical industry. Furthermore, we believe that the local effects are relatively greater again, as regional R & D investments 17 and industry exert a local influence that makes it more likely that research results will be implemented and converted into practical healthcare. We have consequently assumed that Swedish R & D contributes 3% of the total effects of health-related research on Swedish healthcare.

In addition, a large part of the health gains achieved is the result of a generally improved standard of living, and these effects can thus not be ascribed to improvements in medicine. The share accounted for by R & D investments has previously been estimated at 25-67%. 2 , 8 , 10 An econometric study by Lichtenberg also indicates that the proportion falls within this range; the author concluded that medical innovations accounted for 40% of the increase in life expectancy that occurred between 1982 and 2001, calculated as an average for 52 countries in various stages of economic development. 1

Current researchers have their doubts about these earlier estimates, but some such estimate must nevertheless be made in order to gain an idea of the effects of the research. We have consequently made a rough estimate, and believe that the share accounted for by R & D investments in Sweden could be 50%; that is, somewhere in the middle of the range. This can be motivated by the fact that changes of standard of living cannot be expected to have as great an effect in the future as they have had historically, and by the fact that Sweden has a high level of education. However, it should be noted that the validity of using such a high percentage requires that lifestyle improvements can be attributed to health-related research, for example, advice concerning dietary and exercise habits and lifestyle coaching that can be expected to have effects in the future.

The lists of relevant costs and benefits include items of varying power, and some items cannot be valued or even ranked in order of size. We believe that there are nevertheless advantages to show that these effects exist. The fact that certain effects are reciprocally related is also problematic. For instance, an increase in “negative side effects” entails that the effect of “improved health, quality of life, and lifespan” will be lower. For informational purposes, we have chosen to list reciprocally related effects on both the costs and benefits sides.

Valuing Costs and Benefits

In the current study, costs and effects are viewed as either costs or benefits, measured in Swedish kronor [1 USD = 7.86 Swedish Kronor (SEK), 1 July 2005]. In reality, effects may be economic, physical, social, and/or emotional, and may manifest as both costs and benefits (Table ​ (Table3 3 and ​ and4). 4 ). However, our review shows that researchers generally are more interested in the benefits and may ignore some negative effects.

Identified costs and benefits

Overview of the valuation of costs and benefits; the analysis is meant as a template for further development in evaluation of societal net benefit of health related research. Developments are needed in valuation methods of both costs and benefits.

Unreported work and input on the part of doctors and clinical staff should also be included in this item. Research work is often performed during time that is compensated within the county council healthcare budget. It is difficult to find any data regarding such costs, but one study of input in clinical trials in the USA 18 indicates that these costs are significant, and that the research work is only partly compensated.

• Voluntary input (research subjects who participate voluntarily at free of charge) is estimated at SEK 570 million (approximately USD 73 million). The costs associated with voluntary research subjects are calculated as the number of days of participation multiplied by the value of a lost workday at 2005 prices. The number of lost workdays is calculated as half of the estimated number of days that volunteer research subjects contributed in 2005. This item also includes costs for individuals who perform unpaid research work during their spare time, but naturally no documentation regarding this is available.
• More expensive healthcare due to new methods is estimated at SEK 60 million (approximately USD 8 million). The Swedish Association of Local Authorities and Regions (SKL) estimates that medical/technical advances are expected to increase resource demand by roughly 0.8% per year. Given the forecast that the total resource demand increases at a rate of 1.6% per year, the increase from 2014 to 2015 will be SEK 4107 million (in 2005 kronor). 19 , 20 If we presume that medical/technical advances stands for half of this increase (1.6/0.8) and that Sweden's contribution to such advances amounts to 3%, this would yield a cost of roughly SEK 60 million that can be ascribed to research conducted in Sweden. The rest of the increase in resource demand stems from (i) demographic effects, (ii) higher standard of living and an associated higher willingness to pay, and (iii) foreign R & D.
• Negative side effects from treatments . Side effects entail suffering, death, and higher costs for the healthcare system and society. The risk of negative effects is high when innovations are initially put into use, before their long-term benefits have been assessed. However, this effect is difficult to differentiate from others. Suffering and death lead to an increased value of the item “Improved health” on the benefits side, and the costs are also included in the increased cost of healthcare.
• Environmental impacts (emissions) . A 2004 report from the Swedish Medical Products Agency 21 states that the main pharmaceutical substances which have an extensive impact on the aquatic environment are oestradiol and ethinyl-oestradiol. Subsequent reports have also sounded the alarm regarding high emissions of psychotropic drugs, antibiotics, and blood pressure-lowering medications. Changes in natural capital are difficult to calculate, making it hard to valuate environmental impacts. It is, however, important to be aware that environmental effects can have long-term implications for both health and the economy; and an estimation of environmental impacts should be included in any forecast of research effects.
• Implementation costs . New methods that necessitate reorganisation can temporarily lower productivity. Costs can also arise from the efforts which are necessary to break ingrained behaviours whenever innovations are to be integrated into the healthcare process. Innovations usually receive a positive reception in healthcare, but the addition of technologies entails increased on-the-job stress, heightens the risk of improper treatment, and leads to higher maintenance and administrative costs. 22 A cleanout of technologies should thus occur in parallel with the implementation of innovations. The effects included in this item are expressed as a poorer result in process improvements and more expensive healthcare. All benefits represent estimated future effects thought to occur during 2015.

The number of life-years gained is calculated for both men and women as the average annual increase in life expectancy multiplied by the forecast number of inhabitants in 2015. We further assume that the quality of life during the added years needs, on average, to be adjusted downward to 0.75 QALY per life-year gained, 24 as the life-years added during the final stages of life cannot be expected to be spent free of health problems. The value of an increase in life expectancy is calculated at SEK 7,057 million. This value has been adjusted as follows: half of the health gains are assumed to be a result of R & D, and the portion that can be ascribed to Swedish investments is estimated at 3%, as described above.

A general improvement in health must also be assumed, given that lifespan is increased. Register data ascertained from measurements of self-perceived quality of life (1980/81 to 1996/97) indicate that morbidity has decreased in many areas, and people are staying active well into old age. 23 However, a more recent population study from Stockholm shows conversely that self-perceived quality of life has declined (1998 to 2002). 24 This perceived poorer health status can be explained by changes in perceptions of what constitutes sickness and health, and an increase in lifestyle- and stress-related illnesses. It may therefore still be reasonable to assume that, in parallel with increased lifespan, a general elevation of health status is occurring with regard to those diseases/conditions where medical ground is being gained, and where this ground has been gained as a result of health-related research. However, “net health” is not improving, as new health threats are making their presence felt. It is not possible for us to gauge the magnitude of this health gain at the present juncture, but we do believe that it should be at least as great as the value of extended survival.

Based on the foregoing, it may seem self-evident that research has a major impact in terms of reducing the costs of diagnosing, treating, and recovering from disease. However, the resources that are freed up are usually replaced with other costs, as society's willingness to pay for healthcare does not decrease; if a given disease condition requires fewer resources, we will instead invest more in some other area. This, in turn, presumably has positive effects such as improved health, quality of life, and lifespan, but the results are not discernable as savings in healthcare costs.

• Productivity and process improvements . It is difficult to differentiate and value the significance of research in terms of the possibilities of doing healthcare work better and more efficiently. One conceivable method for gaining some indication of the contributions that research makes would be to track individual research investments that are directly intended to, for instance, facilitate healthcare work, reduce treatment times, improve patient comfort, or otherwise bring about improvements in the work environment.
• Lower production losses due to a healthier workforce (reduced absenteeism due to illness) . The value of this is viewed as the increase in production that occurs because people are able to work more. We are living longer and may spend more active years in working life, and we are also getting healthier and consequently do not need to stay home from work to the same extent as was formerly the case. We have chosen to incorporate this effect into a general production increase that generates economic growth (see below), but the question of whether this depiction of a healthier workforce is realistic may be open to discussion. For instance, to date we have not seen a decrease in absenteeism due to illness corresponding to the decline we have seen in mortality figures. We must also note here that absenteeism statistics are difficult to interpret, as the numbers are strongly influenced by the effects of policy factors and economic cycles.

GDP is defined as the value of all the goods and services produced in a country. During the 2000s, Sweden's GDP increased by roughly 3% each year. 29 Calculated conservatively, the increase in GDP from 2014 to 2015 will thus have a value of SEK 100,000 million (expressed in 2005 kronor). If we assume that half of this increase will be attributable to the effects of research, we arrive at a figure of SEK 50 000 million. We know that the cost of health-related research accounts for about 20% of Sweden's total research expenditures. This leaves SEK 10 000 million (100′000 × 50% × 20%) as the result of health-related research, which corresponds to nearly half the research investments made (cost item 1).

• Expanded knowledge base (which can lead to future innovations) . Some authors have noted that an expanded knowledge base can have significant value. 14 Research always has a value in terms of the expertise that is amassed, for instance via post-graduate education, but its value can also be seen as an investment in the ability to develop future innovations. However, here we encounter yet another problem in terms of metrics. New findings are added to our “knowledge capital” at the same time as knowledge is being “taken out” and used, presumably to equivalent degrees. Still, something has been produced, and if it is administered properly it will have major value, not least in the form of entrée to the global research society and the consequent access to a global knowledge base. We have consequently chosen to list this as a separate item in our analysis, but without venturing a guess as to its value.

DISCUSSION AND CONCLUSION

Our analytical model represents an attempt to cover all the relevant effects in a way that has not been done before. In valuing the effects of research we have, however, encountered a number of metric-related problems, and there are many effects that can be neither quantified nor valued monetarily. A summary of all the costs and benefits shows that the estimated future socio-economic benefits exceed the socio-economic costs for the estimated values. The calculated return on investment is 1.08, but it is not currently possible to draw any conclusions in this regard, as certain effects cannot be valued in an acceptable way (See table 4 ).

The above analysis requires further refinement. As mentioned earlier, the current analysis has failed to consider the environmental effects. It would further be of interest to measure the distribution of research impacts on different parts of the societal structure and to compare the effects of current research funding and a hypothetical situation in which the research funding had been altered (reduced, increased or redistributed). Furthermore, to be able to conduct reliable analyses of societal effects of health research, it is necessary to increase the overall knowledge through combined analysis of case studies and broader socioeconomic studies.

It is easy to find examples of research based innovations that have yielded manifold returns. Many have tried to demonstrate the importance of research by calculating the value generated through the use of such innovations, while failing to take the research and development costs into account. Furthermore, a pro-innovation bias is readily evident in the studied literature. The authors have accepted in advance that research and innovation are profitable, and have then either simply described the positive effects, or described the costs and effects for a number of successful medical technologies in relation to older alternatives.

Many studies attempted to describe the value of individual research investments, while few have tried to estimate the scope of the entire research investment and consider it in relation to health improvements and economic growth. The problem with case studies is that it is difficult to make calculations for a large enough number of different disease conditions to enable us to draw conclusions about the entire healthcare system, and overlapping effects often occur, with the result that the effect is overestimated. Using case studies as a template for prioritising research is also problematic, as it only later becomes apparent which research investments yielded the greatest returns. If we consider the relationship between broad-based and cutting-edge research, it is likely that both serve a function in the “research factory,” and that one cannot exist without the other.

To estimate the value of research investments, researcher must be able to isolate the effects that would not have arisen had the investment not been made. This can be difficult, and opinions will vary as to what does or does not constitute a research effect. For instance, Siegler et al. Do not believe that altered living habits, such as decreased tobacco use, constitute an effect of research. 31 There is, however, reason to presume that altered behaviours are, in part, an effect of applied research results, although the problem of determining just how much of the effect can be ascribed to research still remains.

Another problem is that of assessing which research investments have contributed to a given medical advance. How much has basic research contributed? Which basic research has contributed, and to what extent should it be factored in? In addition, research in a given field of knowledge is usually conducted in parallel in a number of countries. How then is one's own country's input to be valued? This is a line of reasoning that most of the studies overlook. Of the studies we have cited, only three incorporated the value of foreign research in their analysis; on the other hand, they did not include all costs of earlier research, an omission that would reasonably lead to an overestimation of the return on the research in question. 4 , 6 , 9 Murphy and Topel factored in all medical research in the USA, but omitted research “spill-overs” from other countries. 8 There are, however, studies that indicate that foreign research does account for a significant share. Eaton and Kortum claimed, for instance, that roughly 50% of the productivity increase in the USA was attributable to technological advances in other countries. 30

Finally, there is a methodological dilemma inherent in the fact that we can never draw comparisons between how healthy we feel at present and how healthy the population felt a number of years ago. Historically speaking, we can see increased lifespan and an increase in the number of years spent in good health. It is, however, impossible to determine whether this improved health is perceived as a higher quality of life, as higher expectations and psychological stresses in an increasingly hectic living environment may have “eaten up” those gains.

The purpose of studies of the value of health-related research is (1) to justify the size of research investments and (2) to optimise the return on those investments. The latter ideally entails that forecasts can be made to identify how and where money should be invested to achieve maximum benefit. This is not possible at present, and methods for valuing research need to be developed further. We must be able to establish and measure the health economic effects of research in a reliable manner if we are to be able to determine with some degree of certainty how research money should be invested. We also need to know more about the links and flows between basic, clinical, and applied research. The issue of how research results will be used has drawn attention in recent years. Considerable costs for the treatment of disease could have been avoided had we been able to quickly apply key research findings in practice. 6 , 32 One classic example of this is the evidence of a link between smoking and lung cancer that was presented in a preliminary report as far back as 1950, 33 but which has still not achieved its full impact.

Meltzer has asserted that health economics methods could also be used to evaluate individual research projects in advance. 34 There is currently a great deal of activity in this area at the international level, and an approach in which the value of additional information is calculated is of great interest in this context. Prospective analyses of the economic value of assessment research have begun to be applied to a limited extent, and have been shown to be capable of providing usable information about the value of research. 35 – 37 International exchanges are valuable in developing methods of evaluation and prioritisation, but local studies at the national and regional level are also needed to achieve effective local guidance.

Increasing our knowledge of the economic effects of research in a more specific way will require us to combine general analyses at the societal level with case studies. This could apply to studies of general effects in society, as well as the effects of specific research projects financed by research councils or in some other way. Luce et al. Have shown how these effects can be elucidated to a greater extent by combining different approaches. 10 The various approaches used to value research offer different advantages and disadvantages, but uncertainty is a problem that they all share. Continued research in this area should consequently be focused on improving the reliability of the input values used in the calculation models.

The paper concludes that positive effect of clinical research benefits excess costs. However, because of vast methodological problems none of the presented research evaluation approaches are sufficient to obtain confident results. The tentative model applied to Swedish health research indicates that the positive effects are predominant, but that the return is in a lower range than the studied literature would imply. Determination of the economic value of research would primarily require significantly better basic data and better knowledge of relationships between research, implementation of new knowledge, and health effects. For future work, to provide information in support of decisions about future allocation of research resources, we recommend a combination of general analyses and strategically selected case studies.

Conflict of interest statement: All authors declare that they have no conflict of interest.

Source of funding: The source of financial support of this work is The Swedish Ministry of Education and Research, through the Committee of Inquiry on the needs and prerequisites of clinical research

Estimating the economic benefits from medical research in the UK

This report is the outcome of a one-year study commissioned by the Academy of Medical Sciences, the Medical Research Council (MRC) and the Wellcome Trust.

Estimating the economic benefits from medical research in the UK (PDF)

If you cannot open or read this document, you can ask for a different format.

Email [email protected] , telling us:

• the name of the document
• what format you need
• any assistive technology you use (such as type of screen reader).

Find out about our approach to the accessibility of our website .

It compares the economic benefits accruing to the UK from publicly and charitably funded medical research in cardiovascular conditions and in mental health conditions with the cost of that research.

This is the website for UKRI: our seven research councils, Research England and Innovate UK. Let us know if you have feedback or would like to help improve our online products and services .