state art
`
`nature publishing group
`
`Trends in Risks associated With New
`Drug Development: Success Rates
`for Investigational Drugs
`JA DiMasi1, L Feldman1, A Seckler1 and A Wilson1
`
`This study utilizes both public and private data sources to estimate clinical phase transition and clinical approval
`probabilities for drugs in the development pipelines of the 50 largest pharmaceutical firms (by sales). The study
`examined the development histories of these investigational compounds from the time point at which they first
`entered clinical testing (1993–2004) through June 2009. The clinical approval success rate in the United States was
`16% for self‑originated drugs (originating from the pharmaceutical company itself) during both the 1993–1998 and the
`1999–2004 subperiods. For all compounds (including licensed‑in and licensed‑out drugs in addition to self‑originated
`drugs), the clinical approval success rate for the entire study period was 19%. The estimated clinical approval success
`rates and phase transition probabilities differed significantly by therapeutic class. The estimated clinical approval
`success rate for self‑originated compounds over the entire study period was 32% for large molecules and 13% for small
`molecules. The estimated transition probabilities were also higher for all clinical phases with respect to large molecules.
`
`IntroductIon
`Numerous studies have found that the drug development process
`is highly expensive and that these costs have trended significantly
`upward for decades.1–6 Many factors affect the cost of drug
`development, but two of the key basic elements are time and
`risk. Development times increased substantially from the 1960s
`through the 1980s but overall remained relatively stable during
`the 1990s.7,8 Thus, development times did not directly contrib-
`ute much to the rapid increase in pharmaceutical R&D costs in
`the past two decades. However, if clinical trials become larger
`and more complex, and the costs of inputs to the develop ment
`process increase faster than inflation, the “time costs” associated
`with the investment of resources in new drug develop ment will
`increase in absolute terms, even if development times remain
`the same. Indeed, there is evidence that the clinical trial proc-
`ess has become more extensive and complex in the past few
`decades.4,9 The situation is similar for drug development risks.
`By development risk, we mean the likelihood that development
`of a drug will be terminated owing to efficacy, safety, or commer-
`cial concerns. High drug failure rates contribute substantially
`to R&D costs, whether or not these costs are otherwise increas-
`ing. Thus, the rate at which pharmaceutical firms successfully
`develop investigational compounds for marketing approval by
`
`regulatory agencies is an important indicator of the effectiveness
`of the drug development process. Processes and technological
`innovations that can improve the predictability of outcomes
`for new compounds can therefore significantly increase the
` productivity of new drug innovation.10
`The historical literature focusing specifically on the quan-
`tification of drug development risks is fairly robust.11–20 The
`aforementioned research on drug development costs includes
`estimates of drug development risks. Early research on devel-
`opment risks suggested that clinical approval rates for self-
`originated drugs in the 1960s were in the neighborhood of one
`in eight.11 Subsequent studies indicated that development risks
`fell in the 1970s, with approval rates averaging approximately
`one in five; the risk levels pertaining to the 1970s remained
`fairly stable to the mid-1990s.1,3,14,15
`This study provides updated clinical approval success rates
`and clinical phase transition analyses for the investigational
`compounds that entered clinical testing between the mid-1990s
`and the early 2000s from the 50 largest pharmaceutical firms
`(as determined by sales). We analyze approval success rates and
`phase transition rate trends within this period for new com-
`pounds as a whole and by therapeutic class. The data are also
`stratified by product type (large molecule vs. small molecule).
`
`1Tufts center for the Study of Drug Development, Tufts University, Boston, Massachusetts, USa. correspondence: Ja DiMasi (joseph.dimasi@tufts.edu)
`
`Received 9 December 2009; accepted 9 December 2009; advance online publication 3 Febraury 2010. doi:10.1038/clpt.2009.295
`
`272
`
`
`
`
`
`
`
`VOLUME 87 NUMBER 3 | MaRch 2010 | www.nature.com/cpt
`
`Abraxis EX2085
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`
`
`state artstate art
`
`The results relating to phase transition rates (or their converse,
`phase attrition rates) allow us to examine whether pharmaceu-
`tical firms are “failing” drugs earlier in the development proc-
`ess and thereby (other factors assumed to be equal) potentially
`reducing overall development costs.
`We examined the investigational drug pipelines of the
`50 largest pharmaceutical firms as determined on the basis of
`sales in 2006. Several data sources were consulted, but the core
`source for the compound list was the IMS R&D Focus investi-
`gational drug pipeline database. We supplemented that database
`with information from two other commercial pipeline databases
`(iDdb3 and Pharmaprojects), as well as from Tufts CSDD inves-
`tigational drug, approved drug, and investigational biopharma-
`ceutical databases that were derived, in part, from confidential
`company surveys, published regulatory agency documents,
`online company pipeline lists, and Internet searches.
`
`Inclusion criteria
`The resulting database contains information on nearly 4,000
`drugs and biologics. For the purpose of simplifying the discus-
`sion, we refer to all the compounds analyzed as “new drugs.”
`Our analyses are restricted to the new drugs for which the start-
`ing dates for phase I testing were available and for which this
`phase I testing was initiated anywhere in the world from 1993
`through 2004. The dataset used for the analysis contains infor-
`mation on the development histories of 1,738 new drugs. For the
`purposes of this study, the dataset’s key elements include infor-
`mation on the drug’s therapeutic class (identified by the major
`indication pursued), the drug type (small molecule, including
`synthetic peptides and oligonucleotides, or large molecule,
`including monoclonal antibodies, recombinant proteins, and
`other biologics), the clinical phases in which the drug has been
`tested, whether the drug has been approved for marketing in
`the United States, the latest phase (clinical or regulatory) that
`the compound had entered (if research on the drug has been
`terminated), the sponsor company, and the source of the drug
`(self-originated, licensed-in, or licensed-out). The bulk of the
`licensed-in compounds were licensed from firms outside the
`top 50. A compound was considered licensed-out only if it had
`been licensed from one of the top 50 firms to a firm outside the
`top 50. We excluded from analysis diagnostics, vaccines, and new
`formulations and indications for already-approved drugs. We
`placed drugs in therapeutic categories according to their clas-
`sification in the IMS R&D Focus database. The database uses the
`Anatomical Therapeutic Chemical classification system estab-
`lished by the World Health Organization Collaborating Centre
`for Drug Statistics Methodology for classifying indications.
`Clinical approval success rates are defined in terms of US
`regulatory approval for marketing. Current success rates for
`the compounds were examined through June 2009. Analyses
`were conducted for the entire study period (1993–2004) and also
`separately for two subperiods (1993–1998 and 1999–2004). Data
`on more recent investigational drugs were available, but, given
`the length of the new drug development process, we judged
`them too recent to be included in a comprehensive analysis of
`success rates.
`
`calculation of success-rate estimates
`The dataset used contains information on the latest phase
`( development or regulatory) of the abandoned drugs at the time
`they were terminated. These data allow us to estimate the likeli-
`hood that an investigational drug will proceed from one clinical
`phase to the next as well as the distribution of research termi-
`nations by phase. They also, in aggregate, permit us to estimate
`the probability of approval for new drugs that enter the clinical
`pipeline. Specifically, we estimate the proportion of new drugs
`that transition from phase i to phase i + 1 as the ratio:
`No. of new drugs that proceeded to phase i + 1/total
`no. of new drugs that entered phase i
`The denominator in the ratio includes only drugs that
`either proceeded thereafter to phase i + 1 or were
` terminated in phase i.
`We estimate the clinical approval success rate as the product
`of the individual phase transition probabilities. These transition
`probability estimates will be unbiased estimates of the population
`transition probabilities if the drugs that are still active in a phase
`are, on average, no different (in terms of the likelihood of pro-
`ceeding to the next phase) from the set of drugs that either have
`been terminated in the phase or have moved on to the next phase.
`There are likely to be variable time lags as to when new informa-
`tion on the status of a drug is available in a database. However,
`if a database firm has not been able to obtain an update on the
`status of a drug over a set period of time (e.g., 18 months for
`R&D Focus), it will show that no development activity has been
`reported for the drug. For purposes of analysis, we assumed that
`the drug was discontinued in the latest phase that it had entered
`if no development activity was subsequently reported. Therefore
`our transition probability estimates may be underestimated; how-
`ever, even if this is so, the downward bias is probably small.
`As noted above, we utilized information from more than
`half a dozen databases and other sources. We recognized that,
`among the databases (pipeline-based or survey-based) and
`other sources that we used, no single source would have the
`most recent information for all drugs. For our study, we took
`the earliest date recorded for the start of phase I testing as the
`date on which clinical testing of the drug began, and the latest
`available development or regulatory phase as its current status.
`For example, if one database had information to the effect that
`a drug has entered phase III while other databases and sources
`showed its status at phase II, we assumed that the drug has pro-
`ceeded to phase III. We thus made use of the most recent infor-
`mation available from the multiple sources regarding the status
`of an investigational drug.
`For the entire study period, 70% of the new drugs in our data-
`set were self-originated (Table 1). We found that the proportion
`of all new drugs that were licensed out to firms outside of the top
`50 pharmaceutical companies was small. These shares were simi-
`lar for the 1993–1998 subperiod. For the full study period, we
`determined a final outcome (success or failure) for 76% of all the
`drugs analyzed; for self-originated drugs, this figure was 81%.
`As expected, the percentage of drugs for which a final outcome
`was available was higher for the earlier period. For example,
`final outcomes were reported for 88% of all drugs and 92% of
`
`CliniCal pharmaCology & TherapeUTiCS | VOLUME 87 NUMBER 3 | MaRch 2010
`
`
`
`
`
`273
`
`Abraxis EX2085
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`
`
`state artstate art
`
`table 1 current and maximum-possible success rates by source of molecule for compounds first tested in humans from 1993 to 2004
`percentage
`Current
`maximum‑possible
`approved
`open
`moleculesa
`completed (%)a
`success rate (%)a
`success rate (%)b
`molecules
`
`Source
`
`n
`
`1993–2004
`
` Self-originated
`
` Licensed-in
`
` Licensed-out
`
` all
`
`1993–1998
`
` Self-originated
`
` Licensed-in
`
` Licensed-out
`
`1,225
`
`412
`
`101
`
`1,738
`
`584
`
`180
`
`57
`
`87
`
`41
`
`10
`
`138
`
`64
`
`32
`
`9
`
`239
`
`141
`
`42
`
`422
`
`48
`
`30
`
`21
`
`80.5
`
`65.8
`
`58.4
`
`75.7
`
`91.8
`
`83.3
`
`63.2
`
`7.1
`
`10.0
`
`9.9
`
`7.9
`
`11.0
`
`17.8
`
`15.8
`
`26.6
`
`44.2
`
`51.5
`
`32.2
`
`19.2
`
`34.4
`
`52.6
`
`24.8
`
`105
`821
` all
`aThrough June 2009. bassumes that all open compounds will eventually be approved.
`
`99
`
`87.9
`
`12.8
`
`27%
`
`19%
`
`16%
`
`93%
`
`93%
`
`93%
`
`64%
`
`64%
`
`64%
`
`56%
`
`45%
`
`40%
`
`82%
`
`71%
`
`65%
`
`Transition probability
`
`16%
`
`16%
`
`100%
`
`90%
`
`67%
`
`64%
`
`63%
`
`66%
`
`41%
`
`39%
`
`Transition probability
`
`Phase I−II
`
`Phase II−III
`
`Phase III−
`NDA/BLA Sub
`
`NDA/BLA Sub−
`NDA/BLA App
`
`Phase I−
`NDA/BLA App
`
`Phase I−II
`
`Phase II−III
`
`Phase III−
`NDA/BLA Sub
`
`NDA/BLA Sub−
`NDA/BLA App
`
`Phase I−
`NDA/BLA App
`
`1993−1998
`
`1999−2004
`
`Self-originated
`
`Licensed-in
`
`All
`
`Figure 1 Phase transition probabilities and clinical approval success
`probabilities for self-originated compounds by period of first-in-human
`testing. BLa, biologics license application; NDa, new drug application.
`
`self-originated drugs that commenced clinical trials during the
`1993–1998 subperiod. Given that the data are censored (some
`drugs are still active), we show both the current and maximum-
`possible US clinical approval success rates. These rates were
`higher for licensed-in than for self-originated drugs.
`
`Success-rate trends
`Figure 1 shows estimated phase transition probabilities and
`the overall clinical approval success rates for the 1993–1998
`and the 1999–2004 subperiods. The results do not suggest any
`trend in the overall clinical approval success rates for new drugs
`over this period; estimates showed that approximately one in
`six new drugs that entered clinical testing during each of these
` subperiods was eventually approved for marketing. However,
`there were small differences between the two subperiods with
`respect to the estimated clinical phase transition rates. The
`results suggest that the failures occurred somewhat earlier in
`the clinical trial process (phases I and II) for drugs initiated into
`clinical trials during the later subperiod.
`There are at least two good reasons for the generally higher clin-
`ical approval success rates for licensed-in compounds. First, these
`compounds have generally undergone some screening or testing
`
`Figure 2 Phase transition probabilities and clinical approval success
`probabilities by source of compound, for compounds first tested in
`humans from 1993 to 2004. BLa, biologics license application; NDa, new
`drug application.
`
`prior to licensing and have been shown to be promising candidates
`for marketing approval. Thus, there may be a screening effect for
`new drugs that are licensed-in. Second, it is likely that many of
`these licensed-in drugs were acquired after some clinical testing
`had been done on them. Although drugs may be licensed-in at
`any point during the development process, including during the
`preclinical period, later clinical phases are associated with higher
`approval rates. We do not have data on when in the development
`process each of the licensed-in drugs was acquired, but if, for
`example, the average licensed-in drug was acquired at phase II,
`then we would expect higher clinical approval success rates for
`the licensed-in group for that reason alone.
`Figure 2 shows estimated phase transition probabilities and
`clinical approval success rates by source of the compound.
`As expected, the estimated overall clinical approval success
`rate is substantially higher for the licensed-in drugs than for
`self-originated drugs (27 vs. 16%). However, the estimated
`transition probabilities for phase III and regulatory review
`are identical for licensed-in and self-originated drugs. The
`higher estimated clinical approval success rate for licensed-in
`drugs derives from higher transition probabilities at phases I
`
`274
`
`
`
`
`
`
`
`VOLUME 87 NUMBER 3 | MaRch 2010 | www.nature.com/cpt
`
`Abraxis EX2085
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`
`
`state artstate art
`
`and II. This suggests that many of the licensed-in drugs were
`acquired after phase I or phase II testing had already been
`conducted by the licensor.
`
`Success rates by therapeutic class
`Prior research has shown that success rates for new drugs
`vary by therapeutic class.3,5,14–16 Table 2 shows current and
` maximum-possible success rates and the percentage of self-
`originated drugs that have had a reported final outcome by
`therapeutic class. Given that the number of compounds avail-
`able for analysis is greatly reduced when the data are stratified
`into therapeutic categories, the entire study period (1993–2004)
`is used. Explicit results are reported for the seven therapeutic
`classes with the most new drugs taken into clinical testing over
`the study period (≥80 compounds). These seven classes account
`for 85% of all self-originated drugs that were included for analy-
`sis. The proportion of drugs in these classes that have reached a
`final outcome varied from 71% for antineoplastic/immunologic
`drugs to 89% for systemic anti-infectives.
`Table 3 shows the estimated phase transition and clini-
`cal approval success probabilities for the seven therapeutic
`classes and one miscellaneous category. There was substantial
`variability by class for both the phase transition probabilities
`
`and the clinical approval success rates. More than 70% of the
` self-originated drugs in the antineoplastic, musculoskeletal, and
`respiratory categories moved from phase I testing to phase II
`testing, whereas fewer than 60% of the self-originated drugs in
`the systemic anti-infective and central nervous system (CNS)
`categories did so. One-third or fewer of the self-originated
`drugs in the respiratory, cardiovascular, and CNS categories
`proceeded from phase II to phase III testing, but nearly half
`of the antineoplastic/immunologic drugs moved from phase II
`trials to much more expensive phase III testing. However, once
`antineoplastic/immunologic drugs reached phase III, they had
`a relatively low estimated probability (55%) of having an appli-
`cation for marketing approval submitted to the US Food and
`Drug Administration. Similarly, only 50% of gastrointestinal/
`metabolism drugs and 46% of CNS drugs moved from phase
`III to regulatory review. In contrast, the systemic anti-infective,
`musculoskeletal, and respiratory drug categories had relatively
`high estimated probabilities of getting to regulatory review after
`they had entered phase III (79% or higher).
`The estimated clinical approval success rates for self- originated
`drugs varied substantially by therapeutic class. The CNS (8%),
`cardio vascular (9%), gastrointestinal/metabolism (9%), and res-
`piratory (10%) categories had relatively low estimated approval
`
`table 2 current and maximum-possible success rates by therapeutic class for self-originated compounds first tested in humans from
`1993 to 2004
`
`Therapeutic class
`
`antineoplastic/immunologic
`
`cardiovascular
`
`cNS
`
`GI/metabolism
`
`Musculoskeletal
`
`Respiratory
`
`Systemic anti-infective
`
`n
`
`254
`
`134
`
`235
`
`120
`
`88
`
`83
`
`122
`
`approved
`molecules
`
`18
`
`4
`
`9
`
`4
`
`8
`
`4
`
`19
`
`21
`
`open
`moleculesa
`75
`
`percentage
`completed (%)a
`70.5
`
`Current
`success rate (%)a
`7.1
`
`maximum‑possible
`success rate (%)b
`36.6
`
`24
`
`40
`
`28
`
`18
`
`15
`
`14
`
`25
`
`82.1
`
`83.0
`
`76.7
`
`79.5
`
`81.9
`
`88.5
`
`86.8
`
`3.0
`
`3.8
`
`3.3
`
`9.1
`
`4.8
`
`15.6
`
`11.1
`
`20.9
`
`20.9
`
`26.7
`
`29.5
`
`22.9
`
`27.0
`
`24.3
`
`Miscellaneous
`
`189
`
`cNS, central nervous system; GI, gastrointestinal.
`aThrough June 2009. bassumes that all open compounds will eventually be approved.
`
`table 3 Phase transition and clinical approval probabilities by therapeutic class for self-originated compounds first tested in humans
`from 1993 to 2004
`
`Therapeutic class
`
`phase i−ii (%)
`
`phase ii−iii (%)
`
`phase iii−rr (%)
`
`rr−approval (%)
`
`Clinical approval
`success rate (%)
`
`antineoplastic/immunologic
`
`cardiovascular
`
`cNS
`
`GI/metabolism
`
`Musculoskeletal
`
`Respiratory
`
`Systemic anti-infective
`
`Miscellaneous
`
`71.8
`
`62.9
`
`59.6
`
`67.5
`
`72.4
`
`72.5
`
`58.2
`
`62.8
`
`Through June 2009.
`cNS, central nervous system; GI, gastrointestinal; RR, regulatory review.
`
`49.0
`
`32.4
`
`33.0
`
`34.9
`
`35.2
`
`20.0
`
`52.2
`
`48.7
`
`55.3
`
`64.3
`
`46.4
`
`50.0
`
`80.0
`
`85.7
`
`78.6
`
`69.8
`
`100
`
`66.7
`
`90.0
`
`80.0
`
`100
`
`80.0
`
`100
`
`91.3
`
`19.4
`
`8.7
`
`8.2
`
`9.4
`
`20.4
`
`9.9
`
`23.9
`
`19.5
`
`CliniCal pharmaCology & TherapeUTiCS | VOLUME 87 NUMBER 3 | MaRch 2010
`
`
`
`
`
`275
`
`Abraxis EX2085
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`
`
`state artstate art
`
`in their estimated phase transition probabilities. Specifically,
`recombinant proteins had higher phase transition rates for the
`early clinical phases but a lower estimated phase transition prob-
`ability for phase III to regulatory review (66% for recombinant
`proteins and 87% for monoclonal antibodies).
`
`Summary
`We estimated phase transition probabilities and clinical approval
`success rates for drugs in the pipelines of the 50 largest pharma-
`ceutical firms by sales. These firms are likely to represent very
`large proportions of the total number of investigational drugs
`and of aggregate industry R&D expenditures. For self-originated
`new drugs that first entered clinical testing in 1993–2004 and
`were observed through mid-2009, the results indicated that
`approximately one in six drugs that enter the clinical testing
`pipeline will eventually obtain approval for marketing in the
`United States. The data did not support the hypothesis of a
`within-period trend, but the overall estimated clinical approval
`success rate is lower than it has been for prior periods.1,4,11–15
`Although the overall success rate was fairly constant over the
`study period, we did find that the failures occurred somewhat
`earlier in the clinical process for the latter half of the study
`period. This has implications for the average cost of new drug
`development.10 However, the reduction in cost because of a rela-
`tively modest improvement in the speed at which firms identify
`failures may easily be more than offset by increases over time
`in the out-of-pocket costs of conducting clinical trials. There is
`evidence to show that clinical trials have become more complex,
`and therefore probably costlier, in recent years.9 In addition,
`when viewed against the background of reported costs of new
`drug development in earlier periods, the increasing complexity
`of clinical trials and the overall drop in clinical approval success
`rates strongly suggest that new drug R&D costs have continued
`to increase at a high rate in recent years.
`We also found, as we have in the past, that clinical approval
`success rates differ by therapeutic class in any given period. Our
`analysis of self-originated drugs found estimated clinical approval
`success rates that varied from 8% for CNS drugs to 24% for sys-
`temic anti-infectives. This variability in success rates by thera-
`peutic class might be explained, at least partially, by differences in
`the uncertainty (inherent in the differing scientific objectives and
`underlying science knowledge base) about the regulatory stand-
`ards that must be satisfied for different drug classes. For example,
`efficacy end points for antibiotics are often clearly defined and can
`be assessed in a relatively straightforward way. In contrast, it can
`often be difficult to prove the efficacy of psychotropic compounds,
`or to establish causal links between these drugs and side effects.
`Finally, we did find substantial differences in clinical approval
`success rates by product type (large vs. small molecules). The
`success rate for large molecules (nearly one-third) is consist-
`ent with the findings from a study of biopharmaceutical R&D
`costs covering a somewhat earlier period.6 We also found higher
`phase transition rates at all phases for large molecules. Although
`R&D costs should be much lower for large molecules given that
`success rates in this category are substantially higher, other
`factors may offset that impact. This appears to be the case for
`
`success rates. In contrast, systemic anti-infectives had a relatively
`high clinical approval success rate (24%). Although the sample sizes
`are much smaller, the rankings of approval success rates by thera-
`peutic class were generally similar for the two study subperiods.
`
`Success rates by product type
`We also analyzed phase transition probabilities and clinical
`approval success rates by product type. Specifically, we exam-
`ined outcomes by grouping drugs into small- and large-molecule
`categories. Large-molecule compounds comprise a minority of
`the compounds in the pipelines of the 50 largest pharmaceutical
`firms, but their number is still significant. For all compounds
`and for the entire study period, large-molecule compounds con-
`stituted 15% of the total number of drugs. There was a slight
`downward trend in that percentage over time, from 17% for the
`1993–1998 period to 13% for the 1999–2004 period. Given that
`large pharmaceutical firms often seek licensing candidates from
`small biopharmaceutical firms, the percentage of large-molecule
`compounds was lower (but not much lower) for self-originated
`drugs. Of the self-originated drugs over the entire study period,
`12% were large-molecule compounds (14% for 1993–1998 and
`11% for 1999–2004). The large-molecule category is dominated
`by monoclonal antibodies and recombinant proteins. For self-
`originated drugs during the entire study period, 47% of the large
`molecules were monoclonal antibodies, 43% were recombinant
`proteins, and 10% were other biologics.
`Figure 3 shows our results for estimated transition and clinical
`approval success probabilities by product type. Estimated transi-
`tion probabilities for all phases were higher for large molecules.
`The estimated clinical approval success rate for large molecules
`(32%) was much higher than for small molecules (13%). Studies
`have indicated that success rates differ within the monoclonal
`antibody class by type of antibody (murine, chimeric, human, or
`humanized).20 However, overall, the estimated clinical approval
`success rates for recombinant proteins and monoclonal anti-
`bodies did not differ by much (34% for recombinant proteins
`and 36% for monoclonal antibodies for self-originated drugs).
`The large-molecule subtypes, however, did vary somewhat
`
`32%
`
`13%
`
`96%
`
`91%
`
`74%
`
`61%
`
`53%
`
`38%
`
`84%
`
`63%
`
`Transition probability
`
`Phase I−II
`
`Phase II−III
`
`Phase III−
`NDA/BLA Sub
`
`NDA/BLA Sub−
`NDA/BLA App
`
`Phase I−
`NDA/BLA App
`
`Small molecule
`
`Large molecule
`
`Figure 3 Phase transition probabilities and clinical approval success
`probabilities by type of compound, for self-originated compounds first tested
`in humans from 1993 to 2004. BLa, biologics license application; NDa, new
`drug application.
`
`276
`
`
`
`
`
`
`
`VOLUME 87 NUMBER 3 | MaRch 2010 | www.nature.com/cpt
`
`Abraxis EX2085
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`

`

`
`
`state artstate art
`
`large-molecule development; the overall projected cost per new
`small-molecule drug was found to be similar to the reported cost
`per large-molecule drug.6
`
`acknowledgmentS
`This research was supported, in part, by a grant from the Pharmaceutical
`Research and Manufacturers of america. We thank Louis cabanilla,
`Laura Faden, Stephanie Rochon, and Julia Wenger, who worked on the
`development of an early version of the database used for this study.
`
`conFlIct oF IntereSt
`The Tufts center for the Study of Drug Development is partially funded
`by unrestricted grants from pharmaceutical and biopharmaceutical
`companies, contract research organizations, trade associations, niche
`providers, and other corporate entities. The principal investigator, J.a.D., has
`consulted for the pharmaceutical industry and served as an expert witness
`in litigation involving pharmaceutical firms.
`
`© 2010 american Society for clinical Pharmacology and Therapeutics
`
`1. DiMasi, J.a., hansen, R.W., Grabowski, h.G. & Lasagna, L. cost of innovation
`in the pharmaceutical industry. J. Health Econ. 10, 107–142 (1991).
`2. Office of Technology assessment, US congress. Pharmaceutical R&D: Costs,
`Risks, and Rewards (Government Printing Office, Washington, Dc, 1993).
`3. DiMasi, J.a., hansen, R.W., Grabowski, h.G. & Lasagna, L. Research and
`development costs for new drugs by therapeutic category. a study of the US
`pharmaceutical industry. Pharmacoeconomics 7, 152–169 (1995).
`4. DiMasi, J.a., hansen, R.W. & Grabowski, h.G. The price of innovation: new
`estimates of drug development costs. J. Health Econ. 22, 151–185 (2003).
`5. DiMasi, J.a., Grabowski, h.G. & Vernon, J. R&D costs and returns by therapeutic
`category. Drug Inf. J. 38, 211–223 (2004).
`6. DiMasi, J.a. & Grabowski, h.G. The cost of biopharmaceutical R&D: is biotech
`different? Manag. Decis. Econ. 28, 285–291 (2007).
`7. DiMasi, J.a. New drug development in the United States from 1963 to 1999.
`Clin. Pharmacol. Ther. 69, 286–296 (2001).
`
`8. Kaitin, K.I. & cairns, c. The new drug approvals of 1999, 2000, and 2001: drug
`development trends a decade after passage of the Prescription Drug User Fee
`act of 1992. Drug Inf. J. 37, 357–371 (2003).
`9. Getz, K.a., Wenger, J., campo, R.a., Seguine, E.S. & Kaitin, K.I. assessing the
`impact of protocol design changes on clinical trial performance. Am. J. Ther.
`15, 450–457 (2008).
`10. DiMasi, J.a. The value of improving the productivity of the drug development
`process: faster times and better decisions. Pharmacoeconomics 20 (suppl. 3),
`1–10 (2002).
`11. cox, c. a statistical analysis of the success rates and residence times for the
`IND, NDa and combined phases. In: Technological Innovation and Government
`Regulation of Pharmaceuticals in the U.S. and Great Britain (eds. Lasagna, L.,
`Wardell, W. & hansen, R.W.), report submitted to the National Science
`Foundation, august 1978.
`12. Sheck, L., cox, c., Davis, h.T., Trimble, a.G., Wardell, W.M. & hansen, R. Success
`rates in the United States drug development system. Clin. Pharmacol. Ther. 36,
`574–583 (1984).
`13. Tucker, S.a., Blozan, c. & coppinger, P. The Outcome of Research on New
`Molecular Entities Commencing Clinical Research in the Years 1976–79
`(OPE Study 77). (Office of Planning and Evaluation, US Food and Drug
`administration, Rockville, MD, 1988).
`14. DiMasi, J.a. Success rates for new drugs entering clinical testing in the
`United States. Clin. Pharmacol. Ther. 58, 1–14 (1995).
`15. DiMasi, J.a. Risks in new drug development: approval success rates for
`investigational drugs. Clin. Pharmacol. Ther. 69, 297–307 (2001).
`16. Kola, I. & Landis, J. can the pharmaceutical industry reduce attrition rates?
`Nat. Rev. Drug Discov. 3, 711–715 (2004).
`17. Bienz-Tadmor, B., Dicerbo, P.a., Tadmor, G. & Lasagna, L. Biopharmaceuticals
`and conventional drugs: clinical success rates. Biotechnology (N.Y.) 10,
`521–525 (1992).
`18. Struck, M.M. Biopharmaceutical R&D success rates and development times.
`a new analysis provides benchmarks for the future. Biotechnology (N.Y.) 12,
`674–677 (1994).
`19. Gosse, M.E., DiMasi, J.a. & Nelson, T.F. Recombinant protein and therapeutic
`monoclonal antibody drug development in the United States from 1980 to
`1994. Clin. Pharmacol. Ther. 60, 608–618 (1996).
`20. Reichert, J.M. Monoclonal antibodies as innovative therapeutics. Curr. Pharm.
`Biotechnol. 9, 423–430 (2008).
`
`CliniCal pharmaCology & TherapeUTiCS | VOLUME 87 NUMBER 3 | MaRch 2010
`
`
`
`
`
`277
`
`Abraxis EX2085
`Apotex Inc. and Apotex Corp. v. Abraxis Bioscience, LLC
`IPR2018-00151; IPR2018-00152; IPR2018-00153
`
`