THE R&D COST
`OF A NEW
`MEDICINE
`
`Jorge Mestre-Ferrandiz,
`Jon Sussex and Adrian Towse
`Office of Health Economics
`
`THE R&D COST OF A NEW MEDICINE
`
`JORGE MESTRE-FERRANDIZ, JON SUSSEX AND ADRIAN TOWSE
`
`WCK1107
`Wockhardt Bio AG v. Janssen Oncology, Inc.
`IPR2016-01582
`
`1
`
`

`

`2
`
`

`

`THE R&D COST
`THE R&D COST
`OF A NEW
`OF A NEW
`MEDICINE
`MEDICINE
`
`JORGE MESTRE-FERRANDIZ,
`JORGE MESTRE—FERRANDIZ,
`JON SUSSEX AND ADRIAN TOWSE
`JON SUSSEX AND ADRIAN TOWSE
`
`DECEMBER 2012
`DECEMBER 2012
`
`
`HE5
`
`Office of Health Economics
`50 years of research and expertise
`
`3
`
`

`

`Office of Health Economics
`Southside
`7th Floor
`105 Victoria Street
`London SW1E 6QT
`United Kingdom
`www.ohe.org
`
`©2012
`All rights reserved
`Printed in the United Kingdom
`ISBN 978-1-899040-19-3
`
`This publication has undergone a rigorous peer review by the independent OHE Editorial Board
`and other experts in the field. The views expressed are those of the authors and do not necessarily
`represent those of the OHE.
`
`ABOUT THE AUTHORS
`
`Dr Jorge Mestre-Ferrandiz is Director of Consulting at the Office of Health Economics. His
`research interests include pharmaceutical pricing and reimbursement systems worldwide, the
`economics of the pharmaceutical industry, the economics of innovation and incentives for
`encouraging medical R&D.
`
`Jon Sussex is Deputy Director at the Office of Health Economics. His areas of expertise include
`health care expenditure; efficiency, competition and incentives in health care systems; the role of
`the private sector in publicly funded health care; and the economic regulation of the
`pharmaceutical industry.
`
`Prof Adrian Towse is Director of the Office of Health Economics. His current research includes
`the use of “risk-sharing” arrangements between health care payers and pharmaceutical companies,
`including value-based pricing approaches; the economics of pharmacogenetics for health care
`payers and the pharmaceutical industry; economic issues that affect both R&D for and access to
`treatments for diseases prevalent in the developing world; the economics of medical negligence;
`and measuring productivity in health care.
`
`ACKNOWLEDGEMENTS
`
`This work was partially funded by an unrestricted research grant from AstraZeneca. We also
`would like to thank CMRI for allowing us to access their data. We are grateful to Prof Pedro-Pita
`Barros, Universidade Nova de Lisboa; Prof Martin Buxton, Brunel University; Prof Patricia
`Danzon, University of Pennsylvania; and Dr Nancy Mattison, The Mattison Group, for
`comments on an earlier draft.
`
`i
`
`4
`
`

`

`CONTENTS
`
`List of Illustrations and Tables
`Executive Summary
`Introduction and Context
`Context: Setting the Scene
`The Cost of a New Medicine—the Mean
`Policy Discussions around the Cost Estimates
`Factors Affecting Development Costs
`Out-of-Pocket Costs
`Success Rates
`Development Times
`Cost of Capital
`Sensitivity Analysis: Impact of Changing the Key Variables
`A New Estimate for Drug Development Costs: Our Analysis
`Data and Methods
`Results
`Comparing Our Analysis with the Published Analyses
`Sensitivity Analysis
`Therapeutic Areas
`Success Rates
`Development Times
`Overall R&D Costs
`Compound Origin: Self-Originated versus Licensed-In
`Firm Size
`Biologics and Biopharmaceuticals
`Drivers of Trends in R&D Costs
`Drivers of Out-of-Pocket Costs
`Drivers of Failure Rates
`Drivers of Development Times
`A New Drug Development Paradigm?
`Summary and Conclusions
`Annex 1. List of Papers Providing Quantitative Evidence
`Annex 2. Sensitivity Analysis
`Glossary
`References
`
`ii
`
`iii
`v
`1
`2
`6
`16
`18
`18
`19
`24
`25
`28
`30
`34
`37
`39
`40
`42
`42
`47
`50
`53
`56
`58
`61
`62
`65
`67
`68
`71
`73
`74
`79
`81
`
`5
`
`

`

`LIST OF ILLUSTRATIONS AND TABLES
`
`Figure 1.1. European and US R&D spending
`
`Table 1.1. Number of new chemical or biological entities (1990–2009)
`
`Table 1.2. Number of new chemical or biological entities (2005–2009)
`
`Figure 1.2.
`
`Structure of the paper
`
`Figure 2.1. The R&D process
`
`Table 2.1.
`
`Comparison of models calculating mean R&D costs
`
`Figure 2.2.
`
`Sample of NMEs: time period when first tested in humans
`
`Table 2.2.
`
`Estimates of the full cost of bringing an NME to market (2011 US$m)
`
`Figure 2.3. Mean R&D costs per successful NME by middle year of study data
`(2011 US$m)
`
`Figure 2.4. Analysis by Paul et al (2010)
`
`Table 2.3. Out-of-pocket mean development costs (2011 US$m)
`
`Table 2.4.
`
`Probability of success (percentages)
`
`Figure 2.5. Trends in attrition rates
`
`Figure 2.6. Number of NMEs required per phase for one successful NME, based on
`recent estimates for probability of success (high and low estimates)
`
`Table 2.5. Development times (months)
`
`Table 2.6.
`
`Cost of capital used in the literature
`
`Figure 3.1. Milestones and intervals used in CMRI’s programmes
`
`Figure 3.2. Relating CMRI’s milestones to “standard” clinical phases
`
`Table 3.1.
`
`Summary of the data on cost per interval (2011 US$m)
`
`Table 3.2.
`
`Success rates by interval
`
`Table 3.3. Development times by interval
`
`Table 3.4. Hypothetical out-of-pocket spending needed for one successful medicine
`(2011 US$m)
`
`Table 3.5.
`
`Capitalised cost per successful medicine (2011 US$m)
`
`Figure 3.3.
`
`Interval times in years, CMRI data
`
`Table 3.6. Out-of-pocket costs for Phases I–III (2011 US$m)
`
`Figure 3.4. Capitalised total cost per successful medicine by cost of capital (2011 US$m)
`
`Table 3.7.
`
`Sensitivity analysis: effect on our base case cost estimate per new medicine
`
`Table 4.1.
`
`Success rates for selected therapeutic area—DiMasi (2001) versus
`DiMasi et al (2010)
`
`2
`
`2
`
`3
`
`5
`
`7
`
`9
`
`10
`
`11
`
`12
`
`15
`
`18
`
`20
`
`21
`
`22
`
`24
`
`26
`
`33
`
`34
`
`35
`
`36
`
`36
`
`37
`
`38
`
`39
`
`40
`
`41
`
`41
`
`43
`
`iii
`
`6
`
`

`

`LIST OF ILLUSTRATIONS AND TABLES
`
`Table 4.2.
`
`Probability of market entry
`
`Figure 4.1. Clinical phase transition probabilities for oncology compounds
`
`Figure 4.2. Transition probabilities: oncology versus other drugs, 1993–2002
`
`Figure 4.3. Mean clinical and approval phase times for approved NMEs by therapeutic
`class, 2005–2009
`
`Table 4.3. Durations by disorder and primary indication
`
`Table 4.4. Development and regulatory approval times: oncologic versus other drugs
`
`Table 4.5.
`
`Average clinical period cost per approved new drug by therapeutic
`class (2011 US$m)
`
`Table 4.6.
`
`Costs for new drugs by disorder and primary indication
`
`Table 5.1.
`
`Current and maximum-possible success rates by source of molecule for
`compounds first tested in humans from 1993 to 2004
`
`Figure 5.1.
`
`Phase transition probabilities and clinical approval success probabilities by source
`of compound, for compounds first tested in humans from 1993 to 2004
`
`Table 7.1.
`
`Capitalised costs per investigational biopharmaceutical compound
`(2011 US$m)
`
`Table 7.2.
`
`Pre-approval out-of-pocket outlays per approved new molecule (2011 US$m)
`
`Table 7.3.
`
`Pre-approval capitalised cost per approved new molecule (2011 US$m)
`
`Table 8.1. Distribution of trials included on ClinicalTrials.gov (August 2012)
`
`Figure 8.1.
`
`Proposed flexible blueprint
`
`Figure 8.2.
`
`“Quick win, fast fail” model
`
`Table A1.
`
`List of papers providing quantitative evidence
`
`Figure A2.1. Capitalised total cost per successful medicine by cost of capital (2011 US$m)
`
`Table A2.1. Current, maximum and minimum success rates
`
`Table A2.2. Capitalised cost of a successful medicine under current, maximum and
`minimum success rates (cost of capital = 11%; US$ 2011 prices)
`
`Table A2.3. Capitalised cost of a successful NME altering success rates by ±1%
`(cost of capital = 11%; 2011 prices)
`
`Table A2.4. Effects of altering current success rates by ±10%
`
`Table A2.5. Sensitivity analysis for cycle times (2011 prices)
`
`Table A2.6. Sensitivity analysis: effect on our base case cost estimate per new medicine
`
`45
`
`46
`
`46
`
`48
`
`49
`
`49
`
`50
`
`51
`
`54
`
`54
`
`59
`
`59
`
`60
`
`63
`
`69
`
`70
`
`73
`
`74
`
`75
`
`75
`
`76
`
`76
`
`77
`
`78
`
`iv
`
`7
`
`

`

`EXECUTIVE SUMMARY
`
`Introduction and Context
`How much it costs to research and develop a successful new medicine has been an important
`policy issue at least since the 1960s. Cost estimates matter not just because of intellectual
`curiosity or for industry understanding of its performance, but because they are a key aspect of
`the international debate about the reasonableness of pharmaceutical prices and the magnitude of
`the long-term investments involved.
`
`Debate continues about whether the R&D productivity of the biopharmaceutical industry has
`fallen. Calculations based on annual rates of R&D spending and the number of new molecular
`(chemical or biological) entities launched suggest a declining trend in R&D productivity. But it
`takes a long time to develop a new drug, so comparing current R&D spending levels with the
`current number of new approvals is an inaccurate measure of R&D productivity, at best.
`
`Mean Research and Development Costs in the Literature
`Published estimates of the mean (average) cost of researching and developing a successful new
`medicine suggest an increase in cost over the last decade—from the estimate of US$802m by
`DiMasi et al (2003) at 2000 prices (US$1,031m at 2011 prices) to the estimate by Paul et al
`(2010) of US$1,867m at 2011 prices. In this study, we present a new estimate, US$1,506m at
`2011 prices, which lies within this range. Our analysis explores how these costs have been evolving
`and for what reasons.
`
`Mean estimates of R&D costs per new medicine, and in particular drawing conclusions based on
`comparisons between estimates, should be treated with caution because of important differences
`in the studies, particularly in the use of different databases of drugs. Moreover, important
`differences exist across subgroups of drugs—for instance, by therapeutic area, by firm size and by
`compound origin.
`
`Key Components of R&D Costs
`Four main variables determine the capitalised cost of a new drug estimate: out-of-pocket costs,
`success rates, development times and the cost of capital. Given the long timescales required to
`develop a new drug and the associated risks, we need to allow for both failures and the cost of
`capital to compute the total cost of a new successful drug, i.e. not just the out-of-pocket costs.
`Capitalised cost is the standard accounting treatment for long-term investments. This recognises
`the fact that investors require a return on research that reflects alternative potential uses of their
`investment.
`
`Out-of-pocket costs: Out-of-pocket development costs, before adjusting for failures, appear to
`have increased over time since the first DiMasi et al (1991) article. The most recent estimates, by
`Paul et al (2010) and Adams and Brantner (2010), are very similar for out-of-pocket
`development costs (from Phase I through III) at around US$215-220m in 2011 US$. The
`studies are less consistent in their estimates of the magnitude of the cost of the different clinical
`trial phases.
`
`Success rates: The most recent estimates of probability of success for Phase I, Phase II and Phase
`III are between 49% and 75%, 30% and 48%, and 50% and 71%, respectively. Overall, the
`cumulative clinical success rate appears to have decreased over time. Pammolli, Magazzini and
`
`v
`
`8
`
`

`

`EXECUTIVE SUMMARY
`
`Riccaboni (2011) find significant decreases in success rates, especially in Phase II and Phase III.
`But analyses by DiMasi and colleagues since the early 1990s suggest that success rates across the
`different phases have changed little.
`
`Development times: Overall development time (Phases I–III) appears to have remained relatively
`constant over time, at around 6.5 years (75–79 months) on average. Phase III trials tend to be
`the longest development phase, although the most recent work suggests that development times
`for Phases II and III now are similar.
`
`Cost of capital: The long timescales of pharmaceutical R&D mean that the cost of capital has a
`major impact on the final cost per successful NME. The estimated cost per successful drug is
`highly sensitive to the cost of capital applied. The more recent studies use a real annual cost of
`capital of 11%, up from the 9% used by DiMasi et al (1991).
`
`Our New Cost Estimate
`In this study, we present a new estimate for mean R&D costs per new successful drug based on
`previously unpublished information collected by CMRI in confidential surveys. Our fully
`capitalised R&D cost estimate per new medicine is US$1,506m at US$ 2011 prices (i.e. US$1.5
`b). Time costs, i.e. cost of capital, represent 33% of total cost. Our new estimate lies within the
`range of other recently reported estimates.
`
`Our overall probability of success is lower than those reported by DiMasi et al (2003) and Paul
`et al (2010). Development times in our study are similar to those reported by DiMasi et al
`(2003) and Paul et al (2010). Total out-of-pocket costs are very similar to those in Paul et al
`(2010) and slightly lower than those in DiMasi et al (2003).
`
`Mean Costs May Hide Important Differences
`Published estimates that refer to the mean cost of R&D per new medicine are just that—
`averages. The literature has shown that the costs of R&D vary with the subgroup of drugs
`included in the analysis. The most important factors affecting costs are: therapeutic area, firm
`size, and whether the molecule is a “traditional” chemical compound or a biologic.
`
`Therapeutic area: R&D costs vary substantially across therapeutic area because of considerable
`variation in three key variables: success rates, development times and out-of-pocket costs.
`
`The most recent analyses suggest that the most expensive therapeutic areas in terms of drug
`R&D costs are neurology, respiratory and oncology. This is because drugs in these categories
`experience lower success rates and longer development times. By comparison, anti-parasitics and
`drugs to treat HIV/AIDS have the lowest R&D costs because of higher success rates and shorter
`development times.
`
`Self-originated versus licensed-in: Most of the published calculations of R&D costs to date
`focus on self-originated new compounds because comprehensive data for licensed-in/acquired
`compounds are very difficult to collect. However, an increasing proportion of drugs are now
`licensed in and clinical success rates for such drugs are higher than for self-originated drugs.
`Greater success may be the results of a “screening effect” for licensed-in compounds: many of the
`
`vi
`
`9
`
`

`

`EXECUTIVE SUMMARY
`
`licensed-in drugs are acquired after Phase I or Phase II testing has been conducted by the licensor
`and, thus, already have been shown to be promising candidates before being included in the
`acquiring company’s development portfolio.
`
`This difference also is apparent during early research, i.e. externally-sourced projects are
`significantly more likely to reach clinical testing than internal projects, perhaps for similar
`reasons. Indeed, some expectation of success would be necessary for a company to license or
`acquire any drug candidate.
`
`Firm size: An important variable explored in the literature is the effect of firm size on R&D
`productivity and whether R&D costs per approved drug vary with firm size. For such analyses,
`papers focusing on self-originated compounds might produce biased results because the sample
`for small firms with successful self-originated drugs would be too small to be meaningful.
`
`Results of research on the impact of firm size on R&D productivity and R&D costs are mixed.
`The evidence from the 1990s and early- to mid-2000s seems to suggest that size matters:
`multiple tangible and intangible assets are associated with fully integrated organisations, where
`core capacities can be important across diseases. It remains unclear, however, whether R&D
`productivity is greater for smaller companies than for traditional “big pharma”.
`
`Biopharmaceuticals: To address the problem of limited evidence on biological medicines,
`DiMasi and Grabowski (2007b) expanded their biotech sample from DiMasi et al (2003) with
`project level and aggregate annual expenditure data from a biotech firm. They found that the
`overall clinical success rate for biotech products is 30.2%, which is higher than the 21.5%
`estimate for traditional pharmaceutical products reported by DiMasi et al (2003). Total clinical
`and approval time is 8% longer for biopharmaceuticals, with nearly all the difference being in
`Phase I.
`
`Capitalisation increases biopharma costs relative to traditional R&D costs because of the longer
`development timeline and a slightly higher cost of capital. To account for the latter, DiMasi and
`Grabowski (2007b) use an 11.5% cost of capital for biologics compared to 11% for other
`medicines. Comparisons between biologics and other pharmaceuticals based on this one study,
`however, should be viewed with caution because the sample size for biologics still is small.
`
`Drivers of Trends in R&D Costs
`The key drivers of the main components of R&D costs present three sets of issues grouped
`around out-of-pocket costs, failure/success rates and development times.
`
`Drivers of out-of-pocket costs: A key element of R&D cost is the cost of clinical trials, which is
`affected by the cost per patient and the number of patients required to collect sufficient data.
`The complexity of clinical trials has increased over time, also increasing their costs. Two trends,
`however, appear to be helping to control trial costs. First, outsourcing to clinical research
`organisations (CROs) appears to increase the efficiency of running trials. Second, locating trials
`in emerging markets (Africa, Asia, Eastern Europe, Latin America and the Middle East) can
`reduce costs, both because local costs are lower and because patient recruitment may be faster.
`Nevertheless, although more clinical trials are being conducted in emerging markets, especially
`
`vii
`
`10
`
`

`

`EXECUTIVE SUMMARY
`
`Phase III trials, the majority of clinical trials still are conducted in the US and Western Europe,
`for reasons related to regulatory conditions, relevant expertise and infrastructure.
`
`Drivers of failure rates: Failure rates appear to have increased over time and have fluctuated
`across stages of development. This may be the result of a combination of reasons. First, regulators
`are becoming more risk averse and may be more reluctant to approve some drugs. Second, R&D
`is directed towards tougher challenges that require drugs with novel mechanisms of action and
`for which clinical endpoints may be less clear cut. Third, within companies, projects may
`advance prematurely, for various reasons, to the later stages of clinical development and then fail
`in Phase III.
`
`Other changes have been identified that could counter increased failure rates. These include
`better preclinical screening that ensures earlier project termination; integrating HTA earlier in
`the process to encourage earlier decisions about discontinuing projects for commercial reasons;
`and developing biomarkers and companion diagnostics that can lead to personalised or stratified
`medicine, with greater prospects for success. Alliances between companies, increasingly
`common, also may increase success rates.
`
`In the short run, however, important technological challenges, especially for
`personalised/stratified medicines, may actually lead to higher failure rates and costs as science
`advances and while companies learn how to better develop biomarkers and companion
`diagnostics. It might take some time for biomarkers and diagnostics to be used efficiently, but
`“learning by doing” could drive a more efficient R&D process in the long run.
`
`Drivers of development times: A number of factors affecting development times have been
`identified. First, regulators may be more risk averse, leading to increased regulatory stringency
`and longer regulatory reviews. Second, companies are directing efforts towards areas that are
`intrinsically associated with very long clinical development times. Third, clinical trials are
`becoming increasingly complex and, as a result, take longer to complete. Fourth, trade offs may
`be occurring between time and success rate—for example, longer Phase II development times
`that allow for additional scrutiny before deciding whether to advance to the next phase might
`produce higher success rates in Phase III. A better use of biomarkers might reduce overall
`development times by enabling pre-identification of patients with a high response rate and/or
`low probability of an adverse drug reaction.
`
`A new drug development paradigm? A number of alternatives have been put forward to try to
`make the R&D process more efficient and affordable. The key suggestions include focusing on
`the earlier phases to reduce technical uncertainties before undertaking the more expensive trials
`in the later development stages and allowing for greater flexibility.
`
`Conclusions
`Mean estimates of R&D costs per successful drug are useful in providing an overall picture, but
`should be treated with caution. Important cost differences exist across therapeutic areas, firm
`sizes and compound origins. In addition, many of the cost estimates in the literature focus on
`self-originated new compounds and exclude licensed-in compounds and compounds that have
`been discovered and/or developed via alliances or deals between companies, which are
`
`viii
`
`11
`
`

`

`EXECUTIVE SUMMARY
`
`increasingly common. Other cost estimates do not differentiate at all between self-originated and
`licensed-in compounds, making it difficult to gauge their accuracy. These issues should be
`factored in when drawing conclusions about R&D costs.
`
`With these caveats in mind, it does appear that R&D costs per new successful medicine are
`increasing. The reasons for the increase in R&D costs are multiple—from higher cash outlays to
`higher costs of capital and higher attrition rates throughout the clinical trials process.
`
`Technological challenges are driving these cost increases in part; today’s targeted diseases currently
`are more complex than diseases targeted decades ago, producing a negative impact on R&D costs.
`In addition, as we move towards personalised/stratified medicine, the need to identify the
`appropriate patient population more narrowly increases. Some examples already exist of new
`medicines being developed alongside companion diagnostics that target very specific patient
`subpopulations; this can increase the R&D costs in the short run as companies adapt to this new
`environment. In the long run, however, this “learning by doing” has the potential to make the
`R&D process more efficient.
`
`Companies sometimes may advance projects prematurely through the R&D process when more
`time and resource invested at an earlier stage could prevent an ultimately unsuccessful and more
`expensive Phase III. Companies continue to work to improve the efficiency of R&D decisions by
`addressing those factors within their control—e.g. greater scrutiny in the early R&D stages,
`integrating health economics earlier in the process, and moving some trials to less expensive
`locations and/or using CROs to manage clinical trials. Biopharmaceutical companies also are
`trying to address the more complex scientific challenges through alliances with other companies
`and by working in collaboration with a range of other stakeholders. This allows both risks and
`rewards to be shared, rather than being borne by one party alone.
`
`The R&D costs identified in our study are driven by a combination of factors, including changes
`in technology and decisions taken by companies and regulators. Whether the current drug
`development paradigm needs revising and—if so, how—is clearly an important policy issue that
`merits further investigation.
`
`ix
`
`12
`
`

`

`INTRODUCTION AND CONTEXT
`
`Key Points
`How much it costs to research and develop a successful new medicine has been an important
`policy issue at least since the 1960s1. Cost estimates matter not just because of intellectual
`curiosity or for industry understanding of its performance, but because they are a key aspect of
`the international debate about the reasonableness of pharmaceutical prices and the magnitude of
`the long-term investments involved.
`
`Debate continues about whether the R&D productivity of the biopharmaceutical industry has
`fallen. Calculations based on annual rates of R&D spending and the number of new molecular
`(chemical or biological) entities launched suggest a declining trend in R&D productivity. But it
`takes a long time to develop a new drug, so comparing current R&D spending levels with the
`current number of new approvals is an inaccurate measure of R&D productivity, at best.
`
`Published estimates of the mean (average) cost of researching and developing a successful new
`medicine suggest an increased over the last decade—from the estimate of US$802m by DiMasi
`et al (2003) at 2000 prices (US$1,031m at 2011 prices) to the estimate by Paul et al (2010) of
`US$1,867m at 2011 prices. In this study, we present a new estimate, US$1,506m (at 2011
`prices), which lies within this range. Our analysis explores how these costs have been evolving
`and for what reasons.
`
`Mean estimates of R&D costs per new medicine and, in particular, drawing conclusions based
`on comparisons between estimates, should be treated with caution because of important
`differences in the studies, particularly in the use of different databases of drugs. Moreover,
`important differences exist across subgroups of drugs—for instance, by therapeutic area, by firm
`size and by compound origin.
`
`An important international policy debate continues about how much it costs to research and
`develop a successful new drug (NME2), be it chemical or biological. It is important to
`understand how much a new medicine costs to research and develop, identifying in particular
`the key drivers of changes in these costs over time, for at least two reasons:
`
`1. To help understand the reasonableness, or otherwise, of the prices sought by producers of
`new medicines
`2. Given R&D productivity challenges in recent years, companies must ensure that their
`R&D resources are spent optimally to maximise use of limited resources
`
`Monitoring and understanding R&D cost drivers is important in devising optimal responses to
`change. These responses, however, need to come from both policy makers and biopharmaceutical
`companies. On the one hand, the pharmaceutical industry is highly regulated (R&D, marketing
`authorisation, pricing) so there is a need to explore what actions regulators can take to make
`R&D more efficient—without compromising safety and efficacy and still ensuring that new
`drugs provide value for money. On the other hand, companies need to take appropriate action to
`increase R&D productivity, in part by improving the ability to assess accurately the value of
`pursuing or discontinuing projects.
`
`1 See, for example, Bogue (1965).
`2 Throughout the paper, we use the acronym “NME” (new molecular entity) to refer to a new product, whether chemical or
`biologic. Note that some other researchers instead use “NCE” (new chemical entity) and CMRI uses “NAS” (new active
`substance). See the Glossary for definitions of each term.
`
`1
`
`13
`
`

`

`INTRODUCTION AND CONTEXT
`
`With this as context, the objective of this publication is to provide an up-to-date overview of the
`literature estimating drug R&D costs. In this study, R&D costs are the costs of researching,
`developing and gaining regulatory approval for a successful new medicine. As described in more
`detail in Figure 2.1, R&D costs include pre-discovery costs (or basic research), pre-clinical (or
`discovery) costs, clinical (or development) costs, and costs associated with regulatory review. Pre-
`discovery and pre-clinical stages are sometimes jointly referred to as “discovery research” (the “R”
`of “R&D”). Clinical (or “development”) costs include Phase I, Phase II and Phase III costs (the
`“D” of “R&D”). R&D costs do not include Phase IV costs, studies done after marketing begins.
`We present some new observations on drug R&D costs, using previously unpublished evidence,
`and explore what we believe are the most important factors that are driving change in the key
`variables affecting costs.
`
`Context: Setting the Scene
`Debate continues about whether the R&D productivity of the pharmaceutical industry has
`decreased over time. The rationale of claiming it has decreased relies on comparing annual global
`pharmaceutical R&D spending (the “input”) and the number of NMEs launched (the “output”).
`Figure 1.1 shows pharmaceutical R&D spending since 1990 in Europe and in the US (in euros
`and US dollars, respectively). Rates of growth have suffered some peaks and troughs over the last
`two decades and especially over the last few years. Indeed, R&D spending in the US and Europe
`fell in 2008 and may have fallen again in 2011, but the general trend is a rising one.
`
`Figure 1.1. European and US R&D Spending
`
`Source: EFPIA (2012)
`
`While Figure 1.1 shows the evolution of the “input” variable, i.e. R&D spending, Table 1.1
`shows one way of measuring R&D “output” in the pharmaceutical industry—the number of
`new chemical or biological entities launched in the same time period.
`
`Table 1.1. Number of new chemical or biological entities (1990–2009)
`
`Number
`
`Total
`
`Average per year
`
`Source: EFPIA (2010a)
`
`1990–1994
`
`1995–1999
`
`2000–2004
`
`2005–2009
`
`162
`
`32
`
`146
`
`29
`
`215
`
`43
`
`207
`
`41
`
`2
`
`14
`
`

`

`INTRODUCTION AND CONTEXT
`
`Table 1.2 breaks down the 2005–2009 figures to show new entities reaching world markets since
`2005, year on year.
`
`Table 1.2. Number of new chemical or biological entities (2005-2009)
`
`Year
`
`2005
`
`2006
`
`2007
`
`2008
`
`2009
`
`Number
`
`30
`
`35
`
`25
`
`31
`
`25
`
`Source: EFPIA (2010a)
`
`Wang and McAuslane (2012) find that the European Medicines Agency (EMA) approved 13 and
`23 NMEs in 2010 and 2011, respectively, while the US Food and Drug Administration (FDA)
`approved 19 NMEs in 2010 and 32 in 2011.
`
`The number of new entities launched has decreased from an average of 43 per year between
`1990 and 1994 to around 30 or fewer per year over the last five years.
`
`If we use this data to crudely analyse R&D productivity, we can argue there has been a decline in
`productivity; the input (R&D spending) is growing while the output (measured by the number
`of NMEs) is decreasing. However, as explained in greater detail later in this publication, it takes
`considerable time to research, develop and gain approval for an NME; hence, the outputs we
`observe during recent years are the result of R&D that was done 10–15 years ago. It is important
`that we do not focus solely on the static relationship between R&D spending and NME count as
`the basis for conclusions about whether productivity has increased or decreased. Whether the
`productivity of recent R&D spending has decreased will not be clear until we can count the
`number of new successful entities launched in the next five to ten years.
`
`Closely related to the productivity discussion is another heavily debated issue about the
`pharmaceutical industry: how much it costs to develop a successful NME.
`
`Taking a high-level view, and without looking at the details of each of the studies that estimate
`these costs, the costs of researching and developing a new drug appear to have increased
`significantly over the last decades. Indeed, Barker (2010) argues that “the current model for
`developing new drugs is becoming unaffordable” (page 357). The latest estimate places the cost
`of a successful drug, from the start of the R&D process to marketing approval, at around
`US$1.9b (Paul et al, 2010) in 2011 prices3. In 2003 these costs were estimated to be US$1.0b in
`2011 prices (DiMasi et al, 2003), which itself was a large increase on an earlier estimate of
`US$451m in 2011 prices in the early 1990s (DiMasi et al, 1991).
`
`Later in this publication we present our own new R&D costs estimate, based on our analysis of
`unpublished data collected by CMRI. Our estimate, US$1.5b in 2011 prices, lies within the
`US$1.0b–US$1.9b range. But, as argued in the next section, we need to be cautious when
`comparing the different estimates.
`
`3 Throughout the paper, we have updated all costs estimates to 2011 prices using the US GDP implicit price deflator.
`
`3
`
`15
`
`

`

`INTRODUCTION AND CONTEXT
`
`Ideally, given that our main interest is to identify trends in real costs per successful NME, we
`would base this estimate on information from all successful NMEs over time. But such
`comprehensive information is not available. The various estimates that are available, moreover,
`use different databases, many of which cannot be compared as they contain confidential, project-
`specific information, and drugs of different vintages; conclusions based on comparisons across
`studies that use different databases should be treated with caution. To explore trends, we
`compare across studies that use similar databases and methodologies. This allows us to identify
`how the various elements that make up total R&D cost have been evolving over time.
`
`The R&D cost per new medicine approved is indeed increasing. The reasons for the increase are
`multiple—from higher cash