THE OXFORD HANDBOOK OF
`
`THE ECONOMICS
`
`OFTHE
`
`BIO PHARMACEUTICAL
`
`INDUSTRY
`
`
`
`Edited by
`
`PATRICIA M. DANZON
`
`AND
`
`SEAN NICHOLSON
`
`OXFORD
`
`UNIVERSITY PRESS
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 1 of 30
`
`

`

`OXFORD
`
`l/NIVEl!SITY PRESS
`
`Oxford University Press, Inc., publishes works that further
`
`
`
`
`
`Oxford University's objective or excellence
`
`
`in research, scholarship, and education.
`
`Oxford New York
`
`Auddand CapcTown DarcsSalaam llongKong K.irnchi
`
`Kuala Lumpur Madrid Melbourne Mexico City Nairobi
`New Delhi Shanghai Taipei Toronto
`
`
`
`With offices in
`
`Argcn!in,\ Austria Brnzil Chile Czech Republic France Greece
`
`Portugal
`Singapore
`
`Cuatemala Hungary Italy Japan
`Poland
`
`South Korea Switzerland Thailand
`Ukraine
`Turkey
`Vietnam
`
`
`
`
`
`Copyright. CQ 2012 by Oxford University Press
`
`Publish<.'d by Oxford University Press, Inc.
`
`
`
`
`198 Madison Avenue, New York, New York m016
`www.oup.com
`
`
`
`Oxliml is a registered trademark of Oxford University Press
`
`
`
`
`
`All rights reserved. No part of this puhli.:alion may ht' reproduced,
`
`
`
`
`
`
`
`stored in a retrieval system, or transmitted, in any form ()f hy any means,
`
`
`
`
`electronic, mechankal, photocopying, recording, or otherwise,
`
`
`
`
`without the prior permission ofOxlilrd University Press.
`
`Library of Congress Cataloging-in-Puhlicalion Data
`
`
`
`
`
`
`
`The Oxford handbook of the economics of the hiopharmaeculical industry/
`
`
`edited by Patricia M. Danzon and Scan Nicholson.
`p.cm.
`Includes bibliographical rcli.:rcnccs and indexes.
`
`
`
`
`
`
`ISBN 978-0-19-974299-8 (cloth: alk. paper) 1. Pharmaceutical industry.
`
`2. Biopharmaceutics-·Fconomic aspects. I. Danwn, Patrida Munch, 1946,
`
`
`
`
`II. Nicholson, Scan. III. Title: Economics of the biopharmaccutical industry.
`
`I 11)9665.5.094 2012
`338-4'76157--de23
`2011040815
`ISBN 978-0-19-974299-8
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 2 of 30
`
`

`

`CONTENTS
`
`Contributors
`
`1.Introduction
`PATRICIA M. DANZON AND SEAN NICHOLSON
`
`vii
`
`1
`
`PART I PHARMACEUTICAL INNOVATION
`
`
`
`2. R&D Costs and Returns to New Drug Development:
`A Review of the Evidence
`JOSEPH A. DIMASI AND HENRY G. GRABOWSKI
`
`3.Financing Research and Development
`
`SEAN NICHOLSON
`
`21
`
`47
`
`4. Cost of Capital for Pharmaceutical, Biotechnology,
`
`
`and Medical Device Firms
`SCOTT E. HARRINGTON
`
`75
`
`5.The Regulation of Medical Products
`
`
`ANUP MALANI AND TOMAS PHILIPSON
`
`6.Incentives to Innovate
`
`DARIUS LAKDAWALLA AND Nr.EllAJ SooD
`
`Exclusivity7- Patents and Regulatory
`
`
`REBECCA S. EISENBERG
`
`100
`
`143
`
`PART II THE MARKET FOR PHARMACEUTICALS
`
`
`
`
`8.Pricing and Reimbursement in US Pharmaceutical Markets 201
`£RKST R. BERNDT AND JOSEPH P. N EWTIOUSE
`
`9.Regulation of Price and Reimbursement for Pharmaceuticals
`266
`
`
`
`PATRICIA J\1. DAKZON
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 3 of 30
`
`

`

`vi
`
`CONTENTS
`
`g Countries10. Drugs and Vaccines for Developin
`
`
`ADRIAN TOWSE, ERIC KEUFFEL, HANNAH E. KETTLER,
`AND DAVID B. RIDLEY
`
`11.Insurance and Drug Spending
`
`MARK V. PAULY
`
`302
`
`336
`
`12. Consumer Demand and Health Effects of Cost-Sharing
`
`
`DANA P. GOLDMAN AND GEOFFREY F. JOYCE
`
`Theory and Practice
`13.Measuring Value: Pharmacoeconomics
`
`
`394
`
`ADRIAN TOWSE, MICHAEL DRUMMOND,
`AND CORINNA SORENSON
`
`14. Price Indexes for Prescri
`
`
`
`
`ANA AIZCORBE AND NICOLE NESTORlAK
`
`of the Issuesption Drugs: A Review
`
`15.Empirical Evidence on the Value of Pharmaceuticals
`
`
`
`CRAIG GARTHWAITE AND MARK DUGGAN
`
`and Consumers16.Promotion to Physicians
`
`
`
`DON KENKEL AND ALAN MATHIOS
`
`17.The Economics of Vaccines
`
`FRANK A. SLOAN
`
`493
`
`524
`
`18.Mergers, Acquisitions, and Alliances
`552
`HENRY G. GRABOWSKI AND MARGARET KYLE
`
`Index
`
`579
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 4 of 30
`
`

`

`CHAPTER 2
`
`R&D COSTS AND
`
`RETURNS TO NEW
`
`DRUG DEVELOPMENT:
`
`A REVIEW OF THE
`
`EVIDENCE
`
`JOSEPH A. DIMAS! AND
`
`HENRY G. GRABOWSKI
`
`ECONOMIC analyses of research and development (R&D) costs and returns in
`
`
`
`
`
`
`
`
`pharmaceuticals have received prominent attention by scholars and policy
`
`
`
`
`
`
`makers. Investment cycles in pharmaceuticals span several decades. Trends in
`
`
`
`
`future R&D costs and returns shape the incentives for companies to pursue R&D
`
`
`
`
`opportunities for new medicines. Economic studies provide a basis for evaluat­
`
`
`ing all the factors affecting R&D costs and returns and can be useful in assessing
`
`
`
`
`productivity changes in the pharmaceutical and biopharmaceutical industries
`
`
`(Munos 2009). They also can be used to consider how various policy actions
`
`
`
`
`
`
`(e.g., price regulation) affect innovation incentives (Giaccotto et al. 2005; Vernon
`2005).
`This chapter reviews the extensive literature on R&D costs and returns. The
`
`
`
`
`
`
`
`
`
`
`first section focuses on R&D costs and the various factors that have affected the
`
`
`
`
`trends in real R&D costs over time. The second section considers economic stud­
`
`
`
`
`ies on the distribution of returns in pharmaceuticals for different cohorts of new
`
`
`
`
`drug introductions. It also reviews the use of these studies to analyze the impact
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 5 of 30
`
`

`

`22
`
`PHARMACEUTICAL INNOVATION
`
`of policy actions on R&D costs and returns. The final se,tion concludes and dis­
`
`
`
`
`
`
`
`
`
`
`cusses open questions for further research.
`
`
`
`PHARMACEUTICAL INDUSTRY R&D COSTS
`
`Estimates of the cost of developing new drugs have varied methodologically and
`
`
`
`
`
`
`
`
`in terms of coverage, but taken together, they paint a picture of substantially rising
`
`
`
`
`costs for more than half a century. The resource cost increases are dramatic, even
`
`
`
`
`
`
`after adjusting for inflation. This section briefly reviews the literature on pharma­
`
`
`
`ceutical R&D costs and then describes some of the more recent results.
`
`Approaches to Estimating Pharmaceutical
`
`
`Industry R&D Costs
`Early attempts to examine at least some of the costs of new drug development were
`
`
`
`
`
`
`
`quite limited in that they did not account for important aspects of the drug devel­
`
`
`
`opment process, such as non-drug-specific R&D, expenditures on drug failur es,
`
`
`
`
`
`
`and the length of the development process and its relationship to opportunity
`
`
`
`
`costs. DiMasi et al. (1991) referenced and discussed the early economk literature
`
`
`
`on the R&D costs of new drug development. One of the earliest of these studies
`
`
`
`(Schnee 1972) examined data on 17 new chemkal entities (NCEs) from the 1950s
`
`
`
`
`and 1960s for a single firm. However, only out-of-pocket (cash outlay) costs were
`
`
`
`
`considered (i.e., the time ,osts of R&D investments were not evaluated), and nei­
`
`
`
`This was fol­were counted. ther fixed discovery costs nor the costs of drug failures
`
`
`
`
`
`lowed by several studies that also focused on individual drug out-of-pocket costs;
`
`
`
`
`taken together with the Schnee estimate of an average cost of $0.5 million per NCE,
`
`
`
`
`these studies suggested that R&D costs increased substantially from the 1950s to
`
`
`
`the late 1960s (Mund 1970; Baily 1972; Sarett 1974; Clymer 1970).
`
`
`
`
`The early literature also included two attempts to develop R&D cost estimates
`
`
`
`
`
`from published aggregate industry data on R&D expenditures and lists of approved
`
`
`NCEs (Mund 1970; Baily 1972). These studies assumed fixed lag times between
`
`
`
`
`industry R&D expenditures and new drug approvals. Although these approaches
`
`
`
`
`
`implicitly accounted for the costs of drug failures, neither of them included cap­
`
`
`
`
`
`italization of costs or accounted for varying lag times between expenditures and
`approvals.
`the full costs of drug development
`The first study that attempted to capture
`
`
`
`
`
`
`(cash outlays for investigational drug failures as well as successes, fixed discovery
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 6 of 30
`
`

`

`R&D COSTS AND RETURNS TO NEW DRUG DEVELOPMENT 23
`
`and preclinical development costs, and time costs) was that of Hansen (1979). The
`
`
`
`
`
`
`
`study used Tufts Center for the Study of Drug Development (Tufts CSDD) survey
`
`
`
`data from a dozen pharmaceutical firms to obtain a random san1ple of their inves­
`
`
`
`tigational drugs and aggregate annual data on their R&D expenditures broken
`
`
`
`
`down by development phase and compound source (self-originated or licensed-in).
`
`
`
`
`Hansen found an average capitalized cost of $54 million in 1976 dollars for develop­
`
`
`ment thal occurred in the 1960s and up to the miJ-197os. As with most subsequent
`studies,
`
`
`
`
`Hansen estimated the R&D cost per approved drug, taking into consider­
`
`
`
`ation cosls incurred on failed drugs and adjusting historical costs to take account
`
`of the opportunity costs of time.
`Following the Hansen (1979) study, Wiggins (1987) applied a regression analysis
`
`
`
`
`
`
`
`
`
`
`
`using industry-reported aggregate annual R&D expenditure data combined with
`
`
`
`
`the development time profile used by Hansen. Wiggins found a capitalized cost per
`
`
`
`
`approved new drug of $125 million in 1986 dollars for drugs approved from 1970
`
`
`
`
`to 1985. However, implicit in the analysis was the a-;sumption of a fixed lag rela­
`
`
`
`
`tionship for the time between R&D expenditures and ultimate new drug approval.
`
`This was not a shortcoming with the Hansen approach.
`Since Hansen's (1979) study, the survey approach has been dominant, with
`
`
`
`
`
`
`
`
`
`similar sludies from DiMasi and associates that found increasingly higher R&D
`
`
`
`
`
`cost estimates for later periods. Specifically, DiMasi et al. (1991) reported an aver­
`
`
`
`age R&D cost of $231 million in year 1987 dollars ($318 million in year 2000 dol­
`
`
`
`
`
`lars), and Di Masi et al. (2003) reported an average R&D cost of $802 million in
`
`
`
`
`
`year 2000 dollars. Companion studies to these two survey-based articles exam­
`
`
`
`
`ined how R&D costs varied by therapeutic category (DiMasi el al. 1995; DiM2si
`
`
`
`
`et al. 2004). Gilbert et al. (2003), using an internal Bain Consulting develop­
`
`
`ment model, found an estimate of su billion for 1995 to 2000 approvals, but
`
`
`
`
`the methodology was not described in any great detail and the results included
`
`
`
`
`
`launch costs. In two recent papers, Adams and Brantner (2006, 2010) attempted
`
`
`
`
`to validate the results reported by DiMasi et al. in 2003 using public data and
`
`
`
`
`found general support for them. fa1rlier, the Congressional Office of Technology
`
`
`Assessment concluded that the results in the 1991 Diivfasi et al. study were rea­
`
`sonable (U.S. Congress, OTA 1993).
`
`
`
`
`The highest estimate to date in the literature of the expected, fully capital­
`
`
`
`
`
`ized cost of developing a single approved drug was $1.8 billion in year 2008 dol­
`
`
`
`
`
`lars (Paul et al. 2010). The authors obtained this result by using a mathematical
`
`
`
`
`model, some recent industry benchmark data on part of the process, and some
`
`
`
`
`internal data from a single firm. The most recent full capitalized R&D cost esti
`
`
`
`mates hased on industry surver data were reported by DiMasi and Grabowski
`
`
`
`
`
`(2007), although they focused on "biotech" drug development. The DiMasi et
`
`
`
`
`al. (2003) and DiMasi and Grabowski (2007) findings are discussed in some
`
`
`
`
`detail later in this chapter, along with some comparisons to the earlier findings
`
`
`
`
`to illustrate the extent to which pharmaceutical R&D costs have changed over
`lime.
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 7 of 30
`
`

`

`24
`
`PHARMACEUTICAL INNOVATION
`
`Risks, Times, and Costs for Traditional
`
`
`
`Pharmaceutical Industry R&D
`
`Figure 2.1 indicates how inflation-adjusted aggregate industry pharmaceutical
`
`
`
`
`
`
`
`
`
`
`R&D expenditures have changed over a long period, measured against changes over
`
`
`
`the same period in the number of US new drug approvals (new chemical entities,
`
`
`or NCEs). Given that drug development phases are lengthy, spreading over many
`
`
`
`
`
`
`years (DiMasi et al. 1991), there is a substantial lag between when R&D expendi­
`
`
`tures are made and when new drugs get approved. Nonetheless, the data in Figure
`
`
`
`2 . .1 strongly suggest that average R&D costs have risen at a rapid rate over time. A
`
`
`
`more rigorous analysis is needed to assess just how high pharmaceutical R&D costs
`
`
`have been during any period and how rapidly they have risen over time. It is also
`
`
`
`
`
`instructive to look beneath an overall estimate of drug development cost to impor­
`
`
`
`
`tant aspects of the drug development process that contribute to that cost.
`
`
`
`Technical Risks
`
`One of the most important contributors to cost of drug development is the amount
`
`
`
`
`
`
`of resources that are devoted to drugs that fail in testing at some point in the devel­
`
`
`
`
`
`opment process. The series of studies begun with Hansen (1979) involved esti­
`
`
`
`
`
`mates of the likelihood that a drug that enters the clinical testing pipeline (i.e.,
`
`
`
`phase 1) will eventually be approved for marketing by the US Food and Drug
`
`
`
`Administration (FDA) and estimates of attrition rates for drugs during the three
`
`
`
`
`
`
`clinical phases of development. Hansen used a clinical approval success rate of one
`
`
`
`in eight (12.5 percent). The second study in the series, DiMasi et al. (1991), found
`
`
`
`
`
`that the clinical approval success rate between the two study periods had increased
`
`
`
`
`substantially, to between one in five and one in four (23 percent). If nothing else
`
`
`
`had changed from one study period to the next, the estimated cost per approved
`
`60
`
`45
`
`60
`
`•
`R&D Expenditures
`
`45
`
`30
`New Drug Approvals •
`�
`
`-.
`
`
`
`•
`
`•
`
`15
`••
`•••
`
`•
`
`---........... 0
`0+---�-�----------�-----
`
`
`1963 1967 1971 1975 1979 1983 1987 1991 1995 1999 2003 2007
`
`Figure 2.1 New drug approvals and R&D spending.
`
`
`
`
`
`
`Soune: Courtcsty of Tufts Center for the Study of Drug Development (CSDD) and
`
`
`
`
`Pharmaceutical Research and Manufacturers of America (Ph RM A), 2009.
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 8 of 30
`
`

`

`R&D COSTS AND RETURNS TO NEW DRUG DEVELOPMENT
`
`25
`
`This significantly. would have declined new drug, inclusive of the cost o[ failures,
`
`
`
`
`
`
`
`
`
`did not happen because, as described later, out-of-pocket preclinical and clinical
`
`
`
`
`costs also increased substantially, as did average development times and the cost of
`
`
`
`
`capital. The result was a much higher full average cost estimate.
`
`
`
`
`The most recent study in the series, Dilvlasi et al. (2003), found that the success
`
`
`
`
`rate had worsened for drugs tested in humans between 1983 and 1994 relative to
`
`
`
`
`drugs tested on humans between 1970 and 1982, but only modestly; The estimate of
`
`
`
`
`the clinical approval success rate was 21.5 percent. The effect offailures on costs was
`
`
`
`
`
`modified somewhat by estimates showing that firms had terminated their clinical
`
`
`
`
`
`
`
`
`failures earlier. However, as discussed later, other factors contributed to produce a
`
`
`
`much higher full cost per approved drug tor the most recent period.
`
`Development Times
`
`When the R&D process for pharmaceuticals is lengthy, development cycles wi 11
`
`
`
`
`
`
`
`
`
`
`be an important indirect determinant of costs if cash nows are capitalized to the
`
`
`
`point at which revenues from the investment could be earned. As development
`
`
`
`
`times increase, so do capitalized cost estimates, other things equal. The time from
`
`
`
`
`synthesis of a new compound to first testing in humans increased by 6.6 months,
`
`
`
`
`on average, between the Hansen (1979) study and the Dil\fasi <'t al. (1991) study.
`
`
`
`
`
`The time from first human testing to regulatory approval increased by almost
`
`
`
`
`
`months, on average, between the study periods. The extra 2.3 years in average total
`
`
`
`
`time from discovery to approval for the second study period accounted for approx­
`
`
`
`
`
`imately 24 percent of the increase in average costs between the studies.
`
`
`
`
`
`In contrast, changes in development times had little impact on the increase in
`
`
`
`
`average cost between the DiMasi et al. (1991) study and the most recent study in
`
`
`
`the series, DiMasi et al. (2003). Although the time from first testing in humans to
`
`
`
`
`
`regulatory approval declined by an average of 8.6 rnonlhs between the two study
`
`
`
`
`
`periods, the total time from discovery to approval remained, on average, virtually
`
`
`
`
`
`identical at 11.8 years. The increase in the cost of capital had a much greater impact
`
`
`
`on total capitalized costs than did changes in development times.
`
`21
`
`Opportunity Costs
`
`Industrial R&D expenditures are investments, and there are potentially long lags
`
`
`
`
`
`
`
`between when the expenditures are made and when any potential returns can be
`
`
`
`
`earned. The three survey-based studies we focus un here attempted to capture
`
`
`
`
`
`these time costs, which, together with the out-of-pocket costs of development, yield
`
`
`
`
`a measure of the opportunity costs of bringing drugs from discovery to marketing
`
`
`
`
`approval. The approach is to capitalize costs to the point of first US approval using
`
`
`
`
`an appropriate discount rate. The discount rates used were estimates of the cost of
`
`
`
`
`capital for the pharmaceutical industry over the respective study periods. Average
`
`
`
`
`out-of-pocket costs by development phase were spread over average development
`
`
`
`times for each phase and capitalized to the point of marketing approval at the dis­
`count rate used for the study.
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 9 of 30
`
`

`

`26
`
`
`
`PH.ARMACEUTICAL INNOVATION
`
`The re:11 (i.e., inflation-adjusted) costs of capital used for the first two studies
`
`
`
`
`
`
`
`
`
`were 8 percent and 9 percent, respectively. The increase of one percentage point
`
`
`
`
`
`accounted for 13 percent of the increase in costs between the first two studies. The
`
`
`
`
`
`corn bi nation of longer devdopment times and a higher discount rate for l he sec�
`
`
`
`
`
`ond study accounted for 37 percent of the increase in average costs. As mentioned
`
`
`
`
`earlier, although there were some differences in development times between the
`
`
`second and third studies, the total development time was constant. Nonetheless,
`
`
`
`
`the estimated discount rate i!pplied to thr cash flows over the representative tirne
`
`
`
`
`
`
`
`profile was 2 percentage points higher for the third study (u percent versus 9 per­
`
`
`
`
`cent). However, out-of-pocket costs increased enough that the time cost share of
`
`
`
`
`total capitalized cost remained virtually the same (50 pt'rcent for the third study,
`
`
`compared with 51 percent for the second).
`
`
`Figure 2.2 shows the primary results for the Di Masi et al.
`(:wo3) study. In year
`
`
`
`
`
`
`
`
`2000 dollars, Lhe estimated preapproval capitalized cost per ,1pproved new drug
`
`
`
`was $802 million, with $403 million of that total accounted for by out-of-pocket
`
`
`
`cash outlays. Pharmaceutical R&D does not end with the approval of an NCE.
`
`
`
`
`Developmcnl often continues on new indications, new dosage strengths, and new
`
`
`
`formulations. The DiMasi et al. (200_1) study provided an esti mctte of postapproval
`
`
`
`R&D costs. It found that approximately one-quarter of the total R&D life-cycle
`
`
`
`
`cash outlays per approved new drug were incurred after a drug product contain­
`
`
`
`
`
`ing the active ingredient was first approved. Given that the analysis is focused on
`
`
`
`
`
`the point offirst marketing approval, the postapproval costs must be discounted
`
`
`
`
`back in time to the date of marketing approval. Therefore, on a capitalized basis,
`
`
`
`
`
`postapproval R&D costs account for only 11 percent of the total life-cycle R&D cost
`
`per approved drug, $897 million.
`
`Cost Trends
`
`The three survey-based studies, taken together, demonstrate that pharmaceutical
`
`
`
`
`
`
`
`
`
`industry R&D costs increased dramaLically over the first four decades of the mod­
`
`
`
`ern era of drug developrnenl-that is, since enactment of the 1962 Amendments to
`
`"'
`
`0
`0
`0
`'.:::',
`
`897
`
`Out"of..Pockct C,1pitalizcd
`
`
`
`0 Post-approval llfill Pre"approval II Total
`
`r:igure 2.2 Pharmaceutical life-cycle R&D costs.
`
`Source: 1-'rnm Di Masi ct al. 2003.
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 10 of 30
`
`

`

`R&D COSTS AND RETURNS TO NEW DRUG DEVELOPMENT
`
`27
`
`802
`
`335
`
`467
`
`Preclinical Clinical
`
`Total
`JD Hansen (1979) li!il DiMasi et al. (1991) II DiMasi ct al. (2003) I
`
`
`
`
`
`Figure 2.3 Pharmaceutical R&D costs gave increased substantially over time.
`
`From DiMasi ct al. 2003.
`Source:
`
`the Food and Drug Cosmetic Act of 1938, which, for the first time in the United
`
`
`
`
`
`
`States, required proof of efficacy as well as safety. Figure 2.3 shows how preclini­
`
`
`
`
`
`cal, clinical, and total preapproval average costs increased across the three studies.
`
`
`
`Preclinical costs are all costs incurred prior to first human testing. This includes
`
`
`
`out-of-pocket discovery costs as well as the costs of preclinical development.
`
`
`
`
`
`Clinical costs include all R&D costs incurred from initial human testing to first
`
`marketing approval.
`In constant dollars, total capitalized preapproval cost per approved new drug
`
`
`
`
`
`
`
`
`increased by a factor of 2.3 between the Hansen (1979) study and the DiMasi et al.
`
`
`
`study, and there was a similar increase of 2.5 between DiMasi et al. (1991)
`(1991)
`
`
`
`
`and DiMasi et al. (2003). However, at a more disaggregated level, there were sub­
`
`
`
`
`
`stantial differences .. From the first to the second study, preclinical costs increased
`
`
`
`
`
`somewhat more than did clinical period costs. However, between the second and
`
`
`
`
`
`third studies, clinical cost per approved drug increased substantially more rapidly
`
`
`
`
`than preclinical cost (an increase of 349 percent for the former, compared with 57
`
`percent for the latter).
`The length of time between the study periods was not identical. We can
`
`
`
`
`
`
`
`
`
`get a more precise estimate of the rate of increase in costs across the studies by
`
`
`
`
`
`
`estimating the average endpoint for analysis in each study. The endpoint is the
`
`
`
`
`date of marketing approval. The first study roughly corresponded to develop­
`
`
`
`
`
`ment that yielded approvals during the 1970s, development for the second study
`
`
`
`
`mostly resulted in approvals during the 1980s, and development for the most
`
`
`
`
`recent study was associated largely with 1990s approvals. DiMasi et al. (2003)
`
`
`
`
`found an average difference in approval dates of 9.3 years between the first and
`
`
`
`second studies and 13 years between the second and third studies. Using these
`
`
`
`
`
`time differences, we can calculate average annual rates of increase between the
`studies.
`Figure 2.4 indicates that the annual rate of increase in inflation-adjusted total
`
`
`
`
`
`
`
`
`
`out-of-pocket costs was relatively constant across the studies (7.6 percent between
`
`
`
`
`the first and second studies and 7.0 percent between the second and third studies).
`
`
`
`
`However, the rates of increase in overall costs mask substantial differences in how
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 11 of 30
`
`

`

`PHARMACEUTICAL INNOVATION
`
`l'rcdi.nical Clinical
`
`Total
`
`
`
`[J 1970s to 1980s approvals • l 980s lo I 990s approvals
`
`Figure 2.4 Annual growth rates for R&D out-of-pocket cost per approved new drug.
`
`Source:
`
`From Di Masi ct al. 2003.
`
`costs changed over time for components of the R&D process. Figure 2.4 shows that,
`
`
`
`
`
`
`
`
`
`
`whereas preclinical costs continued to increase in real terms between the second
`
`
`
`
`and third studies, the rate of increase was less than one-third that between the first
`
`
`
`
`and second studies. On the other hand, the rate of increase in clinical period costs
`
`
`
`
`
`was dramatic for the most recent study-almost twice as fast as that between the
`
`first and second studies.
`
`
`
`Large-Molecule R&D Cost Metrics
`
`Almost all prior research on pharmaceutical R&D costs has focused on synthetic,
`
`
`
`
`
`
`
`
`
`so-called small-molecule drugs, as opposed to biologics, or large-molecule drugs.
`
`
`
`Although some of the molecules for the DiMasi et al. (2003) sample were biolog­
`
`
`
`ics, the overwhelming majority of the drugs in the sample and in the pipelines of
`
`
`
`the survey firms at that time were small-molecule drugs. The study by DiMasi and
`
`
`
`
`
`Grabowski (2007) was the first to focus on so-called biotech molecules. Specifically,
`
`
`
`
`the sample they used consisted of approximately equal numbers of recombinant pro­
`
`
`
`
`
`teins and monoclonal antibodies (mAbs). Although out-of-pocket clinical costs were
`
`
`
`
`
`collected for a relatively small sample of large molecules (17), the other metrics used
`
`
`
`
`
`for the cost analys.is (development times and success and attrition rates) were deter­
`
`
`
`
`mined from large samples. The same methodology used to estimate average costs
`
`
`
`
`for the three survey-based studies of traditional pharmaceutical firm development,
`
`
`
`described earlier, was applied to the biotech sample.
`
`Figure 2.5 shows some of the main results from the DiMasi and Grabowski
`
`
`
`
`
`(2007) study. The average overall capitalized cost per approved new chemical
`
`
`
`
`entity was $1.2 billion for large molecules. The study also compared develop­
`
`
`
`ment costs for small and large molecules. First, the results from the Di Masi et al.
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 12 of 30
`
`

`

`
`
`R&D COSTS AND RETURNS TO NEW DRUG DEVEIOPMENT
`
`-·-····-···-······-----------------------,
`
`615
`
`Preclinical* Clinical
`
`Total
`
`:7
`J O Biotech liJ Pharma 11111 Pharma (tim����);·
`
`
`Figure 2.5 Preapproval
`capitalized
`type.
`cost by new molecule
`
`
`
`S,1urce: From Di.Ma�i and Grabowski 2007.
`
`(2003) study were adjusted upward for inflation, because the biotech results were
`
`
`
`
`
`
`
`
`
`
`expressed in constant 2005 dollars. This yielded costs for the preclinical and clin­
`
`
`
`
`
`ical phases, and overall costs, that were significantly lower for traditional small­
`
`
`
`
`
`
`molecule development. However, the molecules used for the biotech analysis were
`
`
`
`of a later vintage than the sample used for the 2003 study. The biotech sample was,
`
`
`
`
`
`in some sense, five years more recent. Consequently, the results from the DiMasi
`
`
`
`out but also extrapolated ct al. (2003) study were not only adjusted for inflation
`
`
`
`five years using the growth rates implied by the differences between the second
`
`
`
`and third survey-based studies of traditional pharmaceutical development (see
`
`
`
`
`
`Figure 2.4). This produced an overall capitalized cost per approved new chemical
`
`
`
`
`
`entity for traditional pharmaceutical firm development similar to that for biotech
`
`
`
`
`
`
`development ($1.3 billion and $l.2 billion, respectively). However, there were sub­
`
`
`
`
`
`stantial differences by development phase. Clinical period costs were higher for
`
`
`
`traditional phannaceutical development, but preclinical phase costs were higher
`
`for biotech development.
`
`
`
`Recent Metrics and Implications for R&D Costs
`
`
`
`cover drug development literature on the costs of new The studies in the academic
`
`
`
`
`
`
`
`
`the period from the 1950s to part of the first decade of the 21st century. However, it
`
`
`
`
`
`is interesting to at least consider Lhe trends for R&D costs during more recent years
`
`
`and for the near future. Without new data on cash flows, we cannot be conclusive
`
`
`about such trends, but there are many metrics that have an impact on full costs
`
`
`
`
`
`that can be examined for recent years. Taken together, these metrics may strongly
`
`suggest a direction of change.
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 13 of 30
`
`

`

`_10
`
`
`
`PHARMACEUTICAL INNOVATION
`
`Impact of Risk and Time on Re!,"D Costs
`Before we examine recent industry benchmark data, it is instructive to get a sense
`
`
`
`
`
`
`
`
`
`
`
`
`ror the degree Lo which changes in certain key development parameters affect over­
`
`
`
`
`
`all costs. DiMasi (2002) was the first to construct various thought experiments
`
`
`
`
`that examined how much the capitalized cost per approved new drug changes in
`
`
`
`
`
`
`re<.ponse to isolated changes in individual development phase lengths, equal pro­
`
`
`
`
`portionate changes for all development phase lengths <.imultaneously, individual
`
`
`
`
`
`
`clinical-phase attrition rates, and clinical approval success rates.
`
`
`Figure 2.6 is taken from the DiMasi (2002) study. It uses the results from the last
`
`
`
`
`
`survey-based study of traditional pharmaceutical industry development (Di Masi
`
`
`
`
`
`ct al. 2003) as the base agaicsl which changes are me,1sured. The figure shows, in
`
`
`
`
`
`percentage terms, the extent to which full capitalized cost per approved new drug
`
`
`
`
`
`
`is reduced if the overall clinical approval success rate is increased from its base
`
`case value of 21.s percent to _15 percent.
`
`
`Tlw results indicate that cost per approved
`
`
`
`
`
`new drug can be reduced by approximately 30 percent if the approval success rate
`
`
`increases from approximately one in five to one in three.
`A similar improvement in average cost can be obtained instead from faster
`
`
`
`
`
`
`
`
`
`
`developn1ent times. Figure 2-7 shows that a 30 percent improvement in total capi­
`
`
`
`
`talized cost per approved new drug would occur if all development phases and the
`
`regulatory approval phase were simultaneously
`
`
`reduced by half, other things equal.
`
`
`
`
`
`Since this work, Paul et al. (2010) has presented similar results for improvements in
`parameters of their mathematical model.
`
`Trends Development Time, Suceess Rate, and Trial Complexity
`
`
`
`
`
`
`
`Although comprehensive estimates of out-of-pocket cash flows for new drug R&D
`
`
`
`
`for recent years are not available, we can examine trend data for aspects of the
`
`
`
`
`
`development process that can be substantial determinants of changes in costs. As
`
`7S1!i!
`
`}( )',!{,
`
`0 25(!'(
`) u 20'¼,
`
`1:=;c¼,
`
`10%
`
`5')6
`
`()(}})
`
`
`
`Success Hak
`
`-11111- Average phase cost
`
`
`-+-Phase cost adjusted for cost of failures
`
`Figure 2.6 Cost reductions from higher clinical success rates.
`
`
`
`
`Source: From Di fVLsi '.!002.
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 14 of 30
`
`

`

`
`
`R&D COSTS AND RETURNS TO NEW DRUG DEVELOPMENT
`
`31
`
`30%
`
`25%
`20%
`
`.g
`15%
`
`10%
`u
`5%
`
`0% 0% 5% 10%15%20%25%30%35% 40%45% 50%
`Phase time reduction
`
`, ..._ Clinical cost -a- 1btal cost
`
`Figure 2.7 Cost reductions from simultaneous percentage
`
`
`
`
`decreases in all phase lengths.
`rom DiMasi 2002.
`Source:
`
`noted earlier, lengthier average development times, other things being equal, result
`
`
`
`
`
`
`
`
`
`in higher full cost estimates, because R&D cash flows are capitalized at a given dis­
`
`
`
`
`
`
`count rate over a longer period before first marketing approval. Kaitin and DiMasi
`
`
`
`
`
`(2011) examined US clinical development and regulatory approval phase trends
`
`
`
`
`since the early 1980s (Figure 2.8). Although these data do not account for clinical
`
`
`
`
`
`
`testing periods outside the United States prior to testing in the United States nor
`
`
`
`
`for preclinical development periods, the average total time from the start of clini­
`
`
`
`
`
`
`cal testing in the United States to US regulatory approval has varied little, ranging
`
`
`
`from approximately eight to nine years for each five-year period since the early
`1980s.
`Although development times have remained relatively stable over the last few
`
`
`
`
`
`
`
`
`
`decades, the data on technical risks in drug development indicate a worsening of
`
`
`
`clinical approval conditions. The DiMasi et al. (2003) study found an estimated
`
`
`success rate of a little more than one in five (21.5 percent) for investigational drugs
`
`
`
`
`
`DiMasi 1983 and 1994. More recently, between that first entered clinical testing
`
`
`
`
`
`et al. (2010) found an estimated clinical success rate of approximately one in six
`
`
`
`
`
`(16 percent) for investigational drugs that first entered clinical testing from 1994
`
`
`
`
`through 2004 (Figure 2.9). Others have suggested even lower success rates for drugs
`
`
`tested in humans more recently (Paul et al. 2010).
`We can also gain insight into changes i n direct resource costs associated with
`
`
`
`
`
`
`
`
`
`individual investigational drugs from data on the complexity of clinical trials at a
`
`
`
`
`fairly micro level. Getz et al. (2008) examined a very large number of US-based piv­
`
`
`
`
`
`
`
`otal clinical trial protocols to determine changes in protocol complexity over time.
`
`
`
`
`Unique procedures in these protocols were counted, as well as the frequency with
`
`
`
`
`which those procedures were to be employed in each protocol. In addition, eligibil­
`
`
`
`
`
`ity criteria were examined, and a measure of investigator work effort was applied
`
`
`to the individual procedures. Data for 1992-2002 and 2003-2006 were compared.
`
`WATSON LABORATORIES, INC. , IPR2017-01621, Ex. 1084, p. 15 of 30
`
`

`

`
`
`PHARMACEUTICAL INNOVATION
`
`/]----·--------
`
`8_/ !
`
`S.7
`
`5.8
`
`_,,, , � -
`
`6.4
`
`6.5
`
`6.6
`
`6.4
`
`2.8
`
`2.7
`
`2.4
`
`1990-9tl 1995-99
`19::-:0-84 1985-89
`2000-0<1 2005-09
`
`
`L.J \pprnval
`
`Phase O Clinical Phase
`
`Figure 2.8 Mean US clinical development and regulator
`
`
`y approval phase
`times by period of approval.
`
`Suurcc: F