Journal of Health Economics 47 (2016) 20–33
`
`Contents lists available at ScienceDirect
`
`Journal
`
`
`
`of
`
`
`
`Health
`
`
`
`Economics
`
`j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / e c o n b a s e
`
`Innovation in the pharmaceutical industry: New estimates of R&D
`costs夽
`Joseph A. DiMasi a,∗, Henry G. Grabowski b, Ronald W. Hansen c
`
`a Tufts Center for the Study of Drug Development, Tufts University, United States
`b Department of Economics, Duke University, United States
`c Simon Business School, University of Rochester, United States
`
`a
`
`
`
`r
`
`
`
`t
`
`
`
`i
`
`
`
`c
`
`
`
`l
`
`
`
`e
`
`
`
`i
`
`
`
`n
`
`
`
`f
`
`
`
`o
`
`a
`
`
`
`b
`
`
`
`s
`
`
`
`t
`
`
`
`r
`
`
`
`a
`
`
`
`c
`
`
`
`t
`
`The research and development costs of 106 randomly selected new drugs were obtained from a survey
`of 10 pharmaceutical firms. These data were used to estimate the average pre-tax cost of new drug and
`biologics development. The costs of compounds abandoned during testing were linked to the costs of
`compounds that obtained marketing approval. The estimated average out-of-pocket cost per approved
`new compound is $1395 million (2013 dollars). Capitalizing out-of-pocket costs to the point of marketing
`approval at a real discount rate of 10.5% yields a total pre-approval cost estimate of $2588 million (2013
`dollars). When compared to the results of the previous study in this series, total capitalized costs were
`shown to have increased at an annual rate of 8.5% above general price inflation. Adding an estimate of
`post-approval R&D costs increases the cost estimate to $2870 million (2013 dollars).
`© 2016 Elsevier B.V. All rights reserved.
`
`Article history:
`Received 15 August 2014
`Received in revised form 28 January 2016
`Accepted 29 January 2016
`Available online 12 February 2016
`
`JEL classification:
`L65
`O31
`
`Keywords:
`Innovation
`R&D cost
`Pharmaceutical industry
`Discount rate
`Technical success rates
`
`1. Introduction
`
`We provide an updated assessment of the value of the resources
`expended by industry to discover and develop new drugs and bio-
`logics, and the extent to which these private sector costs have
`changed over time. The costs required to develop these new prod-
`ucts clearly play a role in the incentives to invest in the innovative
`activities that can generate medical innovation. Our prior studies
`
`夽 We thank the surveyed firms for providing data, and individuals in those firms
`who kindly gave their time when we needed some of the responses clarified. All
`errors and omissions are the responsibility of the authors. The Tufts Center for
`the Study of Drug development (CSDD) is funded in part by unrestricted grants
`from pharmaceutical and biotechnology firms, as well as companies that provide
`related services (e.g., contract research, consulting, and technology firms) to the
`research-based industry. Tufts CSDD’s financial disclosure statement can be found
`here: http://csdd.tufts.edu/about/financial disclosure. The authors and Tufts CSDD
`did not receive any external funding to conduct this study. The R&D cost and expen-
`diture data for individual compounds and companies are proprietary and cannot be
`redistributed. Other data used were obtained from subscription databases and the
`Food and Drug Administration (FDA) and other websites.
`∗ Corresponding author at: Tufts Center for the Study of Drug Development,
`Tufts University, 75 Kneeland Street, Suite 1100, Boston, MA 02111, United States.
`Tel.: +1 617 636 2116; fax: +1 6176362425.
`E-mail address: joseph.dimasi@tufts.edu (J.A. DiMasi).
`
`http://dx.doi.org/10.1016/j.jhealeco.2016.01.012
`0167-6296/© 2016 Elsevier B.V. All rights reserved.
`
`also have been used by other researchers, including government
`agencies, to analyze various policy questions (US Congressional
`Budget Office, 1998, 2006).
`The full social costs of discovering and developing new com-
`pounds will include these private sector costs, but will also include
`government-funded and non-profit expenditures on basic and
`clinical research that can result in leads and targets which drug
`developers can explore. These additional costs can be substantial.1
`However, it is difficult to identify and measure non-private expend-
`itures that can be linked to specific new therapies. Thus, we focus
`here on the private sector costs.
`The methodological approach used in this paper follows that
`used for our previous studies, although we apply additional statis-
`tical tests to the data (Hansen, 1979; DiMasi et al., 1991, 1995a,b,
`2003, 2004; DiMasi and Grabowski, 2007). Because the methodolo-
`gies are consistent, we can confidently make comparisons of the
`results in this study to the estimates we found for the earlier stud-
`ies, which covered earlier periods, to examine and illustrate trends
`
`1 For example, for fiscal year 2013, the United States National Institutes of Health
`(NIH) spent nearly $30 billion on the activities that it funds (http://officeofbudget.od.
`nih.gov/pdfs/FY15/Approp%20%20History%20by%20IC%20through%20FY%202013.
`pdf).
`
`Abraxis EX2016
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`J.A. DiMasi et al. / Journal of Health Economics 47 (2016) 20–33
`
`21
`
`in development costs. These studies used compound-level data on
`the cost and timing of development for a random sample of new
`drugs first investigated in humans and annual company pharma-
`ceutical R&D expenditures obtained through surveys of a number
`pharmaceutical firms.
`We analyze private sector R&D activities as long-term invest-
`ments. The
`industrial R&D process
`is marked by substantial
`financial risks, with expenditures incurred for many development
`projects that fail to result in a marketed product. Thus, our approach
`explicitly links the costs of unsuccessful projects to those that are
`successful in obtaining marketing approval from regulatory author-
`ities. In addition, the pharmaceutical R&D process is very lengthy,
`often lasting a decade or more (DiMasi et al., 2003). This makes
`it essential to model accurately how development expenses are
`spread over time.
`Given our focus on resource costs and how they have changed
`over time, we develop estimates of the average pre-tax cost of
`new drug development and compare them to estimates covering
`prior periods. We corroborated the basic R&D cost results in this
`study by examining the representativeness of our sample firms and
`our study data, and by incorporating a number of independently
`derived results and data relating to the industry and the drug devel-
`opment process into analyses that provide rough comparators for
`at least components of our cost results. The details of those analyses
`are provided in our online supplement.
`The remainder of this paper is organized as follows. We briefly
`discuss the literature on pharmaceutical industry R&D costs since
`our 2003 study in Section 2. Section 3 briefly outlines the standard
`paradigm for the drug development process. In Section 4 we
`describe the survey sample data and the population from which
`they were drawn, and briefly outline the methodology used to
`derive full R&D cost estimates from data on various elements of the
`drug development process. We present base case pre- and post-
`marketing approval R&D cost estimates in Section 5. Sensitivity
`analyses are presented in Section 6. We describe the representa-
`tiveness of our data, various approaches to validating our results,
`and responses to various critiques in Section 7. Finally, we summa-
`rize our findings in Section 8.
`
`2. Previous studies of the cost of pharmaceutical
`innovation
`
`Much of the literature on the cost of pharmaceutical innovation
`dating back decades has already been described by the authors in
`their previous two studies (DiMasi et al., 1991, 2003). The interested
`reader can find references and discussions about the prior research
`in those studies. The earliest studies often involved a case study
`of a single drug (typically without accounting for the cost of failed
`projects) or they analyzed aggregate data. We will focus here on
`studies and reports that have emerged since DiMasi et al. (2003)
`that involve the use of new data for at least some parts of the R&D
`process. The basic elements of these analyses are shown in Table 1.
`Adams and Brantner (2006, 2010) sought to assess the validity
`of the results in DiMasi et al. (2003) with some alternative data.
`Specifically, in their 2006 article, they used a commercial pipeline
`database to separately estimate clinical approval and phase attri-
`tion rates, as well as phase development times.2 They found a
`similar overall cost estimate ($868 million versus $802 million in
`year 2000 dollars).3 The authors followed that study with another
`
`2 For mean out-of-pocket phase costs, they used the estimates in DiMasi et al.
`(2003).
`3 The Adams and Brantner (2006) study used records in the pipeline database that
`were reported to have entered some clinical testing phase from 1989 to 2002. Thus,
`they did not follow the same set of drugs through time. The data for the commercial
`
`study that featured clinical phase out-of-pocket cost estimates
`derived from regressions based on publicly available data on com-
`pany R&D expenditures (Adams and Brantner, 2010). They found
`a somewhat higher overall cost estimate ($1.2 billion in year 2000
`dollars).4
`In a paper authored by two of the authors of this study (DiMasi
`and Grabowski, 2007), we provided a first look at the costs of
`developing biotech products (specifically, recombinant proteins
`and monoclonal antibodies). The methodological approach was the
`same as that used for our studies of traditional drug development.
`We used some data from DiMasi et al. (2003) combined with new
`data on the costs of a set of biotech compounds from a single large
`biopharmaceutical company. Biotech drugs were observed to have
`a higher average clinical success rate than small molecule drugs, but
`this was largely offset by other cost components. We found that the
`full capitalized cost per approved new compound was similar for
`traditional and biotech development ($1.3 billion for biotech and
`$1.2 billion for traditional development in year 2005 dollars), after
`adjustments to compare similar periods for R&D expenditures.
`The other studies shown in Table 1 are discussed in detail in
`the online supplement. One important finding emerging from the
`survey of cost studies in Table 1 is that clinical success rates are sub-
`stantially lower for the studies focused on more recent periods. This
`observed trend is consistent with other analyses of success prob-
`abilities (DiMasi et al., 2010; DiMasi et al., 2013; Hay et al., 2014;
`Paul et al., 2010) and our analysis below. Average R&D (inflation-
`adjusted) cost estimates are also higher for studies focused on more
`recent periods, suggesting a growth in real R&D costs. While sug-
`gestive, these studies are not strictly comparable to our earlier
`analyses of R&D costs given methodological differences and data
`omissions that are discussed in the online supplement (Appendix
`A).
`
`3. The new drug development process
`
`The new drug development process need not follow a fixed
`pattern, but a standard paradigm has evolved that fits the pro-
`cess well in general. We have described the process in some
`detail
`in previous studies, and the FDA’s website contains a
`schematic explaining the usual set of steps along the way from
`test
`tube
`to new compound approval
`(http://www.fda.gov/
`Drugs/DevelopmentApprovalProcess/SmallBusinessAssistance/
`ucm053131.htm). Marketing approval applications
`for
`inves-
`tigational compounds submitted to the FDA
`for review by
`manufacturers are referred to as new drug applications (NDAs)
`or biologic license applications (BLAs), depending on the type of
`product.
`In basic form, the paradigm portrays new drug discovery and
`development as proceeding along a sequence of phases and activ-
`ities (some of which often overlap). Basic and applied research
`initiate the process with discovery programs that result in the
`synthesis or isolation of compounds that are tested in assays and
`animal models in preclinical development. We do not have the level
`
`pipeline databases are also thin prior to the mid-1990s. The DiMasi et al. (2003)
`study covered new drugs that had first entered clinical testing anywhere in the
`world from 1983 to 1994 and followed the same set of drugs through time.
`4 However, the authors interpreted their estimate as a marginal, as opposed to
`an average, drug cost. The concept, though, of marginal cost has an unclear mean-
`ing here. With high fixed costs and a development process that varies by drug, it is
`difficult to understand what marginal pharmaceutical R&D cost means in this con-
`text. It seems that the relevant marginal concept here is marginal profitability. The
`marginally profitable drug could have a very high or a very low cost. What’s more,
`marginal profitability may only have meaning at the firm, not the industry, level.
`The cost of a marginally profitable drug in the pipeline of a firm may be high for one
`firm and low for another firm.
`
`Abraxis EX2016
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`22
`
`J.A. DiMasi et al. / Journal of Health Economics 47 (2016) 20–33
`
`Table 1
`Prior studies and analyses of pharmaceutical R&D costs (2003–2012).
`
`Study
`
`Study period
`
`Clinical success rate
`
`Real cost of capital
`
`Inflation adjustment
`
`Cost estimate
`
`DiMasi et al. (2003)
`Adams and Brantner (2006)
`Adams and Brantner (2010)
`DiMasi and Grabowski (2007)
`Gilbert et al. (2003)
`O’Hagan and Farkas (2009)
`Paul et al. (2010)
`Mestre-Ferrandiz et al. (2012)
`
`11.0%
`21.5%
`First-in-humans, 1983–1994
`11.0%
`24.0%
`First-in-humans, 1989–2002
`11.0%
`24.0%
`Company R&D expenditures, 1985–2001
`First-in-humans, 1990–2003 (large molecule) 30.2% (large molecule) 11.5%
`2000–2002 (launch)
`8.0%
`NA
`2009 (launch)
`NA
`NA
`≈2007
`11.7%
`11.0%
`In clinical development, 1997–1999
`10.7%
`11.0%
`
`2000 dollars
`2000 dollars
`2000 dollars
`2005 dollars
`2003 dollars
`2009 dollars
`2008 dollars
`2011 dollars
`
`$802 million
`$868 million
`$1.2 billion
`$1.2 billion
`$1.7 billion
`$2.2 billion
`$1.8 billion
`$1.5 billion
`
`of granularity to disaggregate R&D expenditure data into discovery
`and preclinical development testing costs, so for the purposes of
`this study, as in prior studies, discovery and preclinical develop-
`ment costs are grouped and referred to as pre-human costs.5
`Clinical (human) testing typically proceeds through three suc-
`cessive, sometimes overlapping phases. Historically, human testing
`has often been initiated first outside the United States (DiMasi,
`2001). For any of these clinical phases, pharmaceutical compa-
`nies may pursue development of their investigational compounds
`in multiple indications prior to and/or after the initial indication
`approval.
`
`4. Data and methods
`
`Ten multinational pharmaceutical firms of varying sizes
`provided data through a confidential survey of their new drug
`and biologics R&D costs.6 Data were collected on clinical phase
`expenditures and development phase times
`for a randomly
`selected sample of the investigational drugs and biologics of
`the firms participating in the survey.7 The sample was taken
`from a Tufts Center for the Study of Drug Development (CSDD)
`database of the investigational compounds of top 50 firms. Tufts
`CSDD gathered information on the investigational compounds
`in development and their development status from commercial
`pipeline
`intelligence databases (IMS R&D Focus and Thomson
`Reuters Cortellis database [formerly the IDdb3 database]), pub-
`lished company pipelines, clinicaltrials.gov, and web searches.
`Cost and time data were also collected for expenditures on the
`kind of animal testing that often occurs concurrently with clin-
`ical trials.8 The compounds chosen were self-originated in the
`following sense. Their development from synthesis up to initial
`regulatory marketing approval was conducted under the auspices
`of the surveyed firm. This inclusion criterion is broader than it
`might at first seem since it includes compounds of firms that
`were acquired or merged with the survey firm during develop-
`ment and drugs that originated with the survey firm and were
`co-developed (and for which full cost data were available).9
`Licensed-in and co-developed compounds without partner
`
`5 We capture out-of-pocket discovery costs with our data, but the pre-synthesis
`discovery period is highly variable with no clear starting point. For our analyses
`we began our representative discovery and development timeline at the point of
`compound synthesis or isolation. Thus, our estimates of time costs are somewhat
`conservative.
`6 Using pharmaceutical sales in 2006 to measure firm size, 5 of the survey firms
`are top 10 companies, 7 are top 25 firms, and 3 are outside the top 25 (Pharmaceutical
`Executive, May 2007).
`7 A copy of the survey instrument can be found in our online supplement
`(Appendix G).
`8 Long-term teratogenicity and carcinogenicity testing may be conducted after
`the initiation of clinical trials, and is often concurrent with phase I and phase II
`testing.
`9 The criterion also does not preclude situations in which the firm sponsors trials
`that are conducted by or in collaboration with a government agency, an individual
`or group in academia, a non-profit institute, or another firm.
`
`clinical cost data were excluded because non-survey firms would
`have conducted significant portions of the R&D.10
`We also collected data from the cost survey participants on their
`aggregate annual pharmaceutical R&D expenditures for the period
`1990–2010. The firms reported on total annual R&D expenditures
`broken down by expenditures on self-originated new drugs, biolo-
`gics, diagnostics, and vaccines. Data were also provided on annual
`R&D expenditures for licensed-in or otherwise acquired new drugs,
`and on already-approved drugs. Annual expenditures on self-
`originated new drugs were further decomposed into expenditures
`during the pre-human and clinical periods.
`The survey firms accounted for 35% of both top 50 firm phar-
`maceutical sales and pharmaceutical R&D expenditures. Of the
`106 investigational compounds included in the project dataset,
`87 are small molecule chemical entities (including three synthetic
`peptides), and 19 are large molecule biologics (10 monoclonal anti-
`bodies and nine recombinant proteins). For ease of exposition, we
`will refer to all compounds below as new drugs, unless otherwise
`indicated. Initial human testing anywhere in the world for these
`compounds occurred during the period 1995–2007. Development
`costs were obtained through 2013.
`We selected a stratified random sample of
`investigational
`compounds.11 Stratification was based on the status of testing as of
`the end of 2013. Reported costs were weighted to reflect the devel-
`opment status of compounds in the population relative to those in
`the cost survey sample, so that knowledge of the distribution of
`development status in the population from which the sample was
`drawn was needed. The population is composed of all investiga-
`tional compounds in the Tufts CSDD investigational drug database
`that met study criteria: the compounds were self-originated and
`first tested in humans anywhere in the world from 1995 to 2007.
`We found 1442 investigational drugs that met these criteria. Of
`these compounds, 103 (7.1%) have been approved for marketing,
`13 (0.9%) had NDAs or BLAs that were submitted and are still active,
`11 (0.8%) had NDAs or BLAs submitted but abandoned, 576 (39.9%)
`were abandoned in phase I, 19 (1.3%) were still active in phase I, 492
`(34.1%) were abandoned in phase II, 84 (5.8%) were still active in
`phase II, 78 (5.4%) were abandoned in phase III, and 66 (4.6%) were
`still active in phase III. For both the population and the cost survey
`sample, we estimated approval and discontinuation shares for the
`active compounds by phase so that the population and sample dis-
`tributions consisted of shares of compounds that were approved or
`discontinued in phase I, phase II, phase III, or regulatory review. The
`
`10 Large and mid-sized pharmaceutical firms much more often license-in than
`license-out new drug candidates. Firms that license-in compounds for further devel-
`opment pay for the perceived value of the prior R&D typically through up-front fees,
`development and regulatory milestone payments, and royalty fees if the compound
`should be approved for marketing. For a breakdown of new drugs and biologics
`approved in the United States in the 2000s by business arrangements among firms
`initiated during clinical development, see DiMasi et al. (2014).
`11 To ease the burden of reporting and increase the likelihood that firms would
`respond, we limited the number of compounds to be reported on to a maximum of
`15 for any firm (with fewer compounds for smaller firms).
`
`Abraxis EX2016
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`J.A. DiMasi et al. / Journal of Health Economics 47 (2016) 20–33
`
`23
`
`cost survey sample was purposely weighted toward compounds
`that lasted longer in development to increase the amount of infor-
`mation on drugs that reached late-stage clinical testing. Weights,
`determined as described above, were then applied to the com-
`pounds in the cost dataset so that the results would reflect the
`development status distribution for the population from which the
`sample was drawn.
`Some firms were not able to provide full phase cost data for
`every new drug sampled. For example, phase I cost data were avail-
`able for 97 of the 106 new drugs in the dataset (92%). Of the 82
`compounds in the dataset that had entered phase II, cost data were
`available for 78 (95%). For phase III, cost data were available for 42
`of the 43 compounds that entered the phase (98%). However, we
`had cost data for at least one phase for each of the 106 drugs in the
`sample. In aggregate, we had cost data for all phases entered for 94
`of the 106 compounds (89%).12 In addition, five compounds were
`still active in a phase at the time that data were reported. For these
`drugs it is likely that there will be some additional future costs for
`the drug’s most recent phase. Thus, for this reason our cost esti-
`mates are likely to be somewhat conservative. However, given the
`small number of drugs in this category and the fact that the impact
`would be on only one phase for each of these drugs, our overall cost
`estimates are not likely to be substantially affected.
`The methodology that we use to estimate development costs
`is the same as the approach used in our earlier studies (Hansen,
`1979; DiMasi et al., 1991, 2003). We refer the reader to the earlier
`studies and to our online supplement (Appendix A) for details. The
`methodology results in a full risk-adjusted cost per approved new
`compound that also takes into account time costs. That is, we link
`the cost of compound failures to the cost of the successes (inves-
`tigational compounds that attain regulatory marketing approval),
`and we utilize a representative time profile along with an indus-
`try cost of capital to monetize the cost of the delay between
`when R&D expenditures are incurred and when returns to the
`successes can first be realized (date of marketing approval). We
`refer to the sum of out-of-pocket cost (actual cash outlays) and
`time cost per approved new compound as the capitalized cost per
`approved new compound. The full capitalized cost estimate is built
`through a number of estimates of various components of the drug
`development process. These individual component estimates are
`interesting as objects of analysis in their own right, and we provide
`estimates for those components.
`
`5. Base case R&D cost estimates
`
`5.1. Out-of-pocket clinical cost per investigational drug
`
`To determine expected costs, we need estimates of the clinical
`development risk profile. We examined the dataset of 1442 self-
`originated compounds of top 50 pharmaceutical firms described
`above and estimated the phase transition probabilities shown in
`Fig. 1. The overall probability of clinical success (i.e., the likelihood
`that a drug that enters clinical testing will eventually be approved)
`was estimated to be 11.83%. This success rate is substantially lower
`than the rate of 21.50% estimated for the previous study, but con-
`sistent with several recent studies of clinical success rates.13 Such
`an increase in overall risk will contribute greatly to an increase in
`costs per approved new drug, other things equal.
`
`12 Phase cost correlation results presented in the online supplement, together with
`an examination of relative phase costs for drugs that had some missing phase cost
`data, suggest that our phase cost averages (exclusive of missing data) are conserva-
`tive.
`13 See, for example, Paul et al. (2010), DiMasi et al. (2013), and Hay et al. (2014).
`
`Fig. 1. Estimated phase transition probability and overall clinical approval suc-
`cess rates for self-originated new molecular entity (NME) and new therapeutically
`significant biologic entity (NBE) investigational compounds first tested in humans
`anywhere from 1995 to 2007.
`
`As described above, we calculated weighted means, medians,
`standard deviations, and standard errors for clinical phase costs.
`Some of the firms could not separate out long-term animal testing
`costs during clinical development, and instead, included these costs
`in their phase cost estimates by year. To be consistent, therefore,
`for those compounds where animal costs were separately reported,
`we allocated those costs to the clinical phases according to when
`the animal testing costs were incurred. Thus, the clinical phase
`costs presented in Table 2 are inclusive of long-term animal testing
`costs.14
`Weighted mean and median costs per investigational drug
`entering a phase15 increase for later clinical phases, particularly
`for phase III (which typically includes a number of large-scale tri-
`als). In comparison to our previous study (DiMasi et al., 2003), both
`mean and median phase III cost are notably higher relative to the
`earlier phases. While the ratio of mean phase III cost to mean phase
`I cost was 5.7 for the previous study, it was 10.1 here. Similarly, the
`ratio of mean phase III to phase II cost was 3.7 for the earlier study,
`but was 4.4 for this study. Mean phase II cost was also higher rela-
`tive to phase I cost in the current study compared to the previous
`one (2.3 times as high compared to 1.5 times as high).16 Thus, while
`mean cost in real dollars for phase I increased 28% relative to the
`previous study,17 phase I costs were notably lower relative to both
`phase II and phase III for the current study.
`As we will see below, the differential in cost per approved new
`drug between the two studies will be much greater than cost per
`investigational drug because of the much lower overall clinical
`approval success rate. However, our results do show that the impact
`is mitigated to some degree by firms failing the drugs that they
`do abandon faster for the current study period. The distribution
`of clinical period failures for this study were 45.9% for phase I,
`43.5% for phase II, and 10.6% for phase III/regulatory review. The
`
`14 When animal testing costs occurred in a year during which costs were incurred
`for two clinical phases, the animal costs were allocated to the two phases according
`to their relative costs for the year.
`15 Averages for unweighted costs did not differ greatly from the weighted cost
`figures. On an unweighted basis, mean phase I, phase II, and phase III costs were
`$29.7 million, $64.7 million, and $253.5 million, respectively.
`16 The ratios for median costs for the current study are 11.6 for phase III relative
`to phase I, 4.5 for phase III relative to phase II, and 2.6 for phase II relative to phase
`I. The corresponding ratios for the previous study are 4.5, 3.6, and 1.2, respectively.
`17 In real terms, median phase I cost was actually 4% lower for the current study
`compared to the previous study.
`
`Abraxis EX2016
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`24
`
`J.A. DiMasi et al. / Journal of Health Economics 47 (2016) 20–33
`
`Table 2
`Average out-of-pocket clinical period costs for investigational compounds (in millions of 2013 dollars).a
`
`Testing phase
`
`Mean cost
`
`Median cost
`
`Standard deviation
`
`Standard error
`
`Phase I
`Phase II
`Phase III
`
`Total
`
`25.3
`58.6
`255.4
`
`17.3
`44.8
`200.0
`
`29.6
`50.8
`153.3
`
`3.0
`6.6
`34.1
`
`Nb
`
`97
`78
`42
`
`Probability of entering phase (%)
`
`Expected cost
`
`100.0
`59.5
`21.1
`
`25.3
`34.9
`54.0
`
`114.2
`
`a All costs were deflated using the GDP implicit price deflator. Weighted values were used in calculating means, medians, and standard deviations.
`b N = number of compounds with cost data for the phase.
`
`Table 3
`Nominal and real cost of capital (COC) for the pharmaceutical industry, 1994–2010.
`
`Nominal COC (%)
`Inflation rate (%)
`Real COC (%)
`
`1994
`
`14.2
`3.1
`11.1
`
`2000
`
`14.9
`3.1
`11.8
`
`2005
`
`13.3
`2.5
`10.8
`
`2010
`
`11.4
`2.0
`9.4
`
`corresponding figures for the previous study were 36.9% for phase
`I, 50.4% for phase II, and 12.6% for phase III/regulatory review.
`
`5.2. Cost of capital estimates
`
`To account for the time value of money in our previous paper
`(DiMasi et al., 2003), we utilized an 11% real after-tax weighted
`average cost of capital (WACC). In particular, we employed the capi-
`tal asset pricing model (CAPM) to estimate the cost of equity capital.
`This was combined with the cost of debt, appropriately weighted
`with the cost of equity, to yield a representative, pharmaceutical
`industry weighted after-tax cost of capital. The resultant parame-
`ters were estimated at regular intervals from the mid-1980s to the
`year 2000, given the time period spanned by our sample of R&D
`projects.
`In the present paper, we follow the same methodology to com-
`pute WACC. In the current R&D cost analysis, we have a sample
`of new drugs that began clinical trials in 1995 through 2007 and
`which have an average introduction period in the latter part of
`the 2000 decade. Hence, a relevant time period for our cost of
`capital is the mid-1990s through 2010. Our analysis yielded an
`after-tax weighted cost of capital of 10.5%, moderately lower than
`in our last paper. This reflects the fact that the cost of equity cap-
`ital has declined in pharmaceuticals since 2000 (as well as for
`other industrial sectors). Research intensive industries, including
`the pharmaceutical industry, generally finance most of their invest-
`ments through equity, rather than through debt. This is the case
`even when the cost of debt is significantly below the cost of equity
`(Hall, 2002; Vernon, 2004). One of the primary reasons is that
`servicing debt requires a stable source of cash flows, while the
`returns to R&D activities are skewed and highly variable (Scherer
`and Harhoff, 2000; Berndt et al., 2015). Given the low debt-to-
`equity ratios that exist for pharmaceutical firms, the cost of equity
`component dominates the computed WACC values in Table 3.
`To obtain a real cost of capital, we first compute the nominal val-
`ues and then subtract the expected rate of inflation. The nominal
`cost of capital in 1994 is from a CAPM study by Myers and Howe
`(1997). The estimates for 2000, 2005, and 2010 are based on our
`own analysis, utilizing a comparable approach, with a large sam-
`ple of pharmaceutical firms.18 As this table shows, the estimated
`nominal cost of capital for pharmaceuticals was fairly stable during
`
`18 The sample is composed of all publically traded drug firms in the Value Line
`Survey which also provides beta values and the other pharma-specific parameters
`used in the CAPM calculations for the relevant years. The long-term horizon equity
`risk premium, and the yield on long-term government bonds employed in the CAPM
`analysis, are from Ibbotson Valuation yearbooks for 2000, 2005, and 2010.
`
`the period 1994–2000 (14.2–14.9%). However, it decreased during
`the decade of 2000s, particularly after the global recession occurred
`(with a value of 11.4% observed in 2010).
`As discussed in DiMasi et al. (2003), the rate of inflation was
`above historical values during the first part of the 1980s, but then
`receded back to or below historical levels throughout most of the
`1990s. Hence, we utilized the long run historical value for inflation
`for the expected inflation level in 1994 and 2000 (3.1%), as in our
`prior work. For the 2000s decade, inflation was significantly below
`historical values. In this case, we employed a 5-year lagged moving
`average to compute the expected rate of inflation in 2005 and 2010
`(calculated as 2.5% and 2.0%, res