Clinical Approval Success Rates for Investigational
`Cancer Drugs
`JA DiMasi1, JM Reichert1, L Feldman1 and A Malins1
`
`We examined development risks for new cancer drugs. For the full study period, the estimated clinical approval success
`rate for cancer compounds was 13.4% (9.9% for the first half of the study period, 19.8% for the second half). Small
`molecules had a somewhat higher clinical approval success rate than did large molecules (14.3 vs. 11.5%). Compounds
`studied solely in hematologic indications had markedly higher estimated clinical approval success rates than did
`compounds studied only in solid tumor indications (36.0 vs. 9.8%). The first, second, and third cancer indications pursued
`had estimated clinical approval success rates of 9.0, 8.2, and 6.9%, respectively. Success rates of second and third
`indications were found to be highly dependent on the success or failure of the first indication pursued (54.9 and 42.4%,
`respectively, for second and third indications if the first indication is a success, but 2.5 and 1.8%, respectively, if the first
`indication is a failure).
`
`The biopharmaceutical industry is increasingly focusing on the
`development of new targeted therapies for cancer, a number
`of which have been approved in the past few decades.1 Drugs
`of particular note in this category are small-molecule protein
`kinase inhibitors (e.g., pazopanib, crizotinib, ruxolitinib, vemu-
`rafenib, and axitinib) and monoclonal antibodies (e.g., cetuxi-
`mab, bevacizumab, panitumumab, ofatumumab, ipilimumab,
`and brentuximab vedotin). Development of these products
`depended on understanding the molecular characteristics of the
`various types of cancer and the modes of action of the drugs.
`Despite progress in this area, there remains substantial unmet
`medical need for new cancer drugs with superior efficacy and
`better safety profiles that can be administered using methods
`that provide greater patient convenience.
`A variety of innovative approaches to cancer drug develop-
`ment are being simultaneously explored by companies, includ-
`ing focusing on new targets within validated pathways and new
`pathways, design of novel drug formats, and improved clini-
`cal study design and protocols that may expedite the process.
`Development of new drugs, however, is lengthy, is very expen-
`sive, and has substantial technical risks.2–17 To foster a better
`understanding of the technical risks specifically associated with
`development of new cancer drugs, we examined the biopharma-
`ceutical industry pipeline of cancer drugs that entered clinical
`study from 1993 to 2004 and estimated clinical phase transi-
`tion rates and overall clinical approval rates. We analyzed these
`data to determine success rates for cancer drugs in general, and
`
`examined factors such as the composition of matter, the type of
`cancer investigated, and successive indications pursued clini-
`cally overall and conditional on the success or failure of the lead
`indication.
`We collected data relevant to the clinical development and
`approval of cancer drugs that first entered studies during 1993
`to 2004 from the public domain and the archives of the Tufts
`Center for the Study of Drug Development. Public sources
`accessed included company websites such as ClinicalTrials.
`gov and Drugs@FDA, the commercial pipeline databases such
`as IMS Health R&D Focus and Thomson Reuters Partnering,
`and the medical literature. Pipeline data were obtained for
` companies located worldwide and of all sizes.
`
`INCLUSION/EXCLUSION CRITERIA
`Clinical study was sponsored, at least in part, by a commercial
`firm. Candidates either originated at a company or were
`licensed from a commercial, government, or academic source.
`Candidates sponsored in clinical study exclusively by academic,
`government, or nonprofit organizations were excluded.
`Cooperative group or other noncommercial sponsorship of
` trials postapproval or after the commercial sponsor abandoned
`the compound were excluded.
`Clinical study was first initiated during the interval between
`1 January 1993 and 31 December 2004. This criterion allowed
`us to follow the development of a given group of investigational
`drugs over time, with a sufficient amount of time to have elapsed
`
`1Tufts Center for the Study of Drug Development, Tufts University, Boston, Massachusetts, USA. Correspondence: JA DiMasi (joseph.dimasi@tufts.edu)
`Received 6 May 2013; accepted 29 May 2013; advance online publication 17 July 2013. doi:10.1038/clpt.2013.117
`
`CliniCal pharmaCology & TherapeuTiCS | VOLUME 94 NUMBER 3 | SEPTEMBER 2013
`
`329
`
`nature publishing group
`
`state art
`
`Abraxis EX2081
`Cipla Ltd. v. Abraxis Bioscience, LLC
`IPR2018-00162; IPR2018-00163; IPR2018-00164
`
`

`

`for a substantial number of the compounds to have reached a
`final fate.
`The candidate’s activity was primarily directed against cancer-
`ous cells or it functioned secondarily to affect cancerous cells.
`Candidates studied for supportive care use (e.g., nausea and
`pain drugs) or as adjunct treatments (e.g., erythropoietin) were
`excluded, as were candidates that improved the efficacy of cancer
`therapeutics but had no inherent anticancer activity (e.g., radio
`and chemosensitizers, detoxifying agents, multidrug resistance
`gene/protein inhibitors).
`Candidates included an active ingredient that had not been
`previously approved for any indication. Covalently modified
`therapeutics (e.g., pegylated molecules) were considered new
`relative to the parent molecule. New formulations (e.g., liposome
`encapsulation) of candidates or previously approved products
`were excluded.
`The majority of studies carried out during clinical develop-
`ment were for cancer indications. Candidates in clinical study
`primarily for noncancer indications were excluded even if
`some cancer studies were performed. Products first marketed
`for noncancer indications but later studied for cancer (e.g., tha-
`lidomide) were excluded. Indications were defined at the level
`of the organ system affected. Changes in the line of therapy for
`the same organ system (e.g., first-line therapy for breast cancer
`after prior approval as second-line therapy for breast cancer) or
`combination therapy after approval as monotherapy in the same
`organ system were not considered to be separate indications for
`purposes of this analysis.
`Candidates studied for precancerous conditions (e.g., myelo-
`dysplastic syndrome) were included, but candidates for condi-
`tions involving noncancerous cellular proliferation (e.g., actinic
`keratosis and benign prostatic hyperplasia) were excluded.
`Composition of matter was assigned on the basis of the drug’s
`molecular structure. The cancer drugs were classified as small
`molecule, natural product, peptide, oligonucleotide, monoclonal
`antibody, recombinant protein, or biologic (i.e., matter derived
`from a natural source). For analysis, the small-molecule drug
`(SMD) category was composed of synthesized chemicals, pep-
`tides, and oligonucleotides, as well as natural products. The bio-
`logics drug category was composed of monoclonal antibodies,
`recombinant proteins, and biologics.
`Data available for the complete clinical development pro-
`grams, including study start dates and specific indications
`studied, were collected for all candidates. Indication data for
`candidates that were first evaluated in exploratory studies of
`patients with a variety of solid tumors were collected, but these
`exploratory studies were not included during assignment of
`lead, secondary, or follow-on indications. The lead, second-
`ary, and follow-on indication categories were assigned on the
`basis of the start dates for defined tumor types, that is, the lead
`indication was the first specific indication studied, the second-
`ary indication was the second specific indication studied, and
`follow-on indications comprised all other indications studied.
`The specific tumor types defined for this study were biliary
`tract, bladder, brain, breast, cervical, central nervous system,
`colorectal, esophageal, gastrointestinal, head and neck, kidney,
`
`leukemia, liver, lung, non–small cell lung, small-cell lung, lym-
`phoma, melanoma, mesothelioma, myeloma, neuroendocrine,
`ovarian, pancreatic, peritoneal, prostate, sarcoma, stomach, thy-
`roid, and urinary tract cancers. The more general categories of
`hematological malignancies, cancers of the female reproductive
`organs, and solid tumors were assigned in cases when patients
`with multiple relevant tumor types were included in the studies.
`For example, studies defined as for female reproductive organ
`cancers included patients with ovarian, fallopian tube, or peri-
`toneal cancer.
`
`CALCULATION OF SUCCESS-RATE ESTIMATES
`Candidates were considered terminated if no clinical studies
`were active or recently concluded. The clinical development sta-
`tus of the candidates was assigned on the basis of data available
`through mid-2012. Candidates were categorized as in phase I,
`phase II, phase III, US regulatory review, approved in the United
`States, approved outside the United States, or discontinued.
`Clinical approval success-rate calculations were determined as
`the product of estimated clinical phase transition probabilities.
`Percent completion was defined as the percentage of products
`with a known fate (US approval or worldwide discontinuation).
`Clinical phase transition rates were calculated as follows: the
`number of candidates that completed a given phase and entered
`the next was divided by the difference between the number of
`candidates that entered the phase and those that were still in
`the phase at the time of the calculation. Transitions occurring
`between phases of clinical studies conducted worldwide were
`included. Phase transition and clinical approval success rates
`were calculated at both the molecule level and the indication
`level. Estimates at the molecule level were determined regardless
`of indication pursued. That is, a molecule was taken as having
`progressed from one clinical phase to the next if testing was
`initiated in at least one indication, even if other indications were
`abandoned at the earlier phase. A success at the molecule level
`is defined as US regulatory approval for marketing in at least
`one indication. Analyses performed at the indication level were
`calculated in the same way as at the molecule level, except that
`phase progression and success are defined for particular indica-
`tions. The indication level analyses reported here are for the lead,
`second, and third indications.
`
`BASE DATA SET CHARACTERISTICS
`The data set of investigational cancer drugs that fulfilled the
`inclusion/exclusion criteria was composed of 625 candidates. Of
`these, 449 (72%) were SMDs and 176 (28%) were biologics. The
`annual shares of investigational cancer compounds varied from
`64 to 80% for SMDs. Overall, the number of investigational can-
`cer drugs entering clinical study per year increased 50% between
`the first half of the study period and the second half (from 250
`to 375). Increases were observed for both the SMD and biolog-
`ics categories, with a somewhat greater increase observed for
`biologics (59%) than for SMDs (47%). A final outcome (success
`or failure) across all indications studied was known for 72% of
`the compounds. Final outcomes were known for 84% of the
`compounds in the first half of the study period (first clinical
`
`330
`
`VOLUME 94 NUMBER 3 | SEPTEMBER 2013 | www.nature.com/cpt
`
`state art
`
`Abraxis EX2081
`Cipla Ltd. v. Abraxis Bioscience, LLC
`IPR2018-00162; IPR2018-00163; IPR2018-00164
`
`

`

`testing during 1993 to 1998) and 65% of the compounds in the
`second half of the study period (first clinical testing during 1999
`to 2004).
`
`ONCOLOGY SUCCESS-RATE TRENDS
`For the entire study period, we estimated that 13% of the inves-
`tigational cancer molecules would be approved for marketing
`in the United States (Figure 1). This compares to an estimated
`overall approval success rate of 16% for all compounds origi-
`nated by top 50 firms found in a prior study.1 Although three-
`quarters of the oncology compounds that entered clinical
`testing progressed from phase I to phase II, less than half of
`the compounds that entered phase II moved on to phase III,
`and of those compounds that underwent phase III testing, less
`than half of those made it to submission of an application for
`marketing approval to the US Food and Drug Administration.
`The pattern was, in part, different for compounds originated
`by top 50 firms and studied clinically during the same period.
`For the top 50 firms, the estimated transition rate was some-
`what lower (approximately two-thirds) than that for the oncol-
`ogy compounds analyzed here, whereas the phase II to phase
`III transition rates were similar at ~40%. However, we found a
`dramatic difference in phase III success rates (transitions from
`phase III to New Drug Application/Biologic License Application
` submission). Although ~65% of the compounds originated by
`the top 50 firms across all therapeutic areas that entered phase
`III were estimated to progress to regulatory review, we estimated
`that only 47% of oncology compounds will do so.
`Figure 1 also shows estimated phase transition and overall
`clinical approval success rates for oncology compounds when
`the study sample is divided into two equal periods. The results
`indicate a nearly doubling of the clinical approval success rate
`from ~10% for cancer compounds that first entered clinical test-
`ing during 1993 to 1998, to ~20% for those that first entered
`clinical testing during 1999 to 2004. Across the two periods,
`the transition rates for early-stage clinical testing were similar
`
`and actually slightly lower for the later period. However, the
`transition rates for late-stage clinical testing and regulatory
`review were notably higher for the later period. Whereas only
`slightly more than one in three of the compounds from the ear-
`lier period that entered phase III testing progressed to regulatory
`review, two in three of the compounds from the later period did
`so. The success rate for regulatory submissions was also higher
`for compounds tested in the later period. The results from our
`previous study on compounds originated by top 50 firms across
`all therapeutic categories show no appreciable change in the
`overall clinical approval success rate for these two study sub-
`periods. The results were qualitatively similar to those for all
`oncology compounds in that phase transition rates were lower
`for the later period for early-stage clinical testing but higher for
`late-stage clinical testing and regulatory review. However, unlike
`the case for all oncology compounds, for the top 50 companies
`and all therapeutic categories combined these changes in phase
`transition rates approximately offset each other so that the over-
`all clinical approval success rate was stable.
`
`ONCOLOGY SUCCESS RATES BY MOLECULE AND
` CANCER TYPE
`We examined phase transition rates and overall clinical approval
`success rates by molecule type. Specifically, we estimated success
`rates separately for SMDs and biologics (Figure 2). The results
`for the overall clinical approval success rate were similar (11.5%
`for biologics vs. 14.3% for SMDs), but the patterns of phase tran-
`sitions were different. SMDs had higher transition rates in early
`to mid-stage clinical testing, particularly for phase II to phase III
`transitions (12% higher). However, biologics were more success-
`ful in transitioning from phase III to regulatory review and from
`regulatory review to approval. The higher later-stage transition
`rates were not large enough to completely offset the lower early-
`stage transition results.
`The biologics category was dominated by monoclonal anti-
`bodies. Of the 176 biologics, 121 (69%) were monoclonal
`
`100%
`
`91.7%
`
`84.2%
`
`66.7%
`
`46.7%
`
`42.4%
`
`41.0%
`
`41.6%
`
`35.1%
`
`78.8%
`
`75.0%
`
`72.4%
`
`Transition probability
`
`Phase I–II
`
`Phase II–III
`
`Phase III to
`NDA/BLA submission
`
`NDA/BLA submission to
`NDA/BLA approved
`
`Phase I to
`NDA/BLA approved
`
`1993–1998
`
`1999–2004
`
`1993–2004
`
`19.8%
`
`13.4%
`
`9.9%
`
`Figure 1 Phase transition probabilities for cancer compounds by period of first clinical testing. NDA/BLA, New Drug Application/Biologic License Application.
`
`CliniCal pharmaCology & TherapeuTiCS | VOLUME 94 NUMBER 3 | SEPTEMBER 2013
`
`331
`
`state art
`
`Abraxis EX2081
`Cipla Ltd. v. Abraxis Bioscience, LLC
`IPR2018-00162; IPR2018-00163; IPR2018-00164
`
`

`

`100%
`
`90.0%
`
`100%
`
`90.5%
`
`87.5%
`
`37.9%
`
`41.1%
`
`36.0%
`
`75.0%
`
`69.7%
`
`54.8%
`
`Transition probability
`
`11.5% 14.3%
`
`9.8%
`
`44.8%
`
`50.0%
`
`46.0%
`
`33.0%
`
`77.1%
`69.5%
`
`Transition probability
`
`Phase I–II
`
`Phase II–III
`
`Phase III to
`NDA/BLA
`submission
`
`NDA/BLA
`submission to
`NDA/BLA approved
`
`Phase I to
`NDA/BLA
`approved
`
`Biologics
`
`SMD
`
`Figure 2 Phase transition probabilities for cancer compounds by molecule
`type (first clinical testing, 1993–2004). NDA/BLA, New Drug Application/
`Biologic License Application; SMD, small-molecule drug.
`
`antibodies. This was the only biologics subcategory with a suffi-
`cient number of observations to conduct a separate success-rate
`analysis. The overall clinical approval success rate for mono-
`clonal antibodies was similar to that for biologics as a whole
`(12.6 vs. 11.5%, respectively). However, the early-phase tran-
`sition rates were somewhat lower for monoclonal antibodies
`(65.6% for phase I to phase II and 30.2% for phase II to phase
`III), whereas the phase III to regulatory submission transition
`rate was somewhat higher (63.6 vs. 50.0%).
`The results for SMDs and biologics contrast sharply with
`those of our previous study for all compounds originated by
`the top 50 firms studied clinically during the same period. The
`early-stage transition rates for the full set of oncology SMDs
`were higher than those for compounds originated by the top
`50 firms across all therapeutic areas (77 vs. 63% for phase I to
`phase II, and 45 vs. 38% for phase II to phase III). However,
`a much higher percentage of the oncology compounds that
`reached phase III failed in that phase than was the case for all
`compounds originated by the top 50 firms. The phase III to
`regulatory submission transition rate was 61% for the SMDs of
`the top 50 firms as compared with 46% for the oncology com-
`pounds. The regulatory submission to approval transition rate
`was similar (91% for top 50 firm SMDs vs. 90% for all oncology
`compounds). However, on net the overall clinical approval suc-
`cess rate for oncology SMDs was slightly higher than that for
`top 50 firm SMDs (14.3 vs. 13.0%).
`The results for biologics were notably different for all oncology
`biologics as compared with biologics originated by top 50 firms
`in general. We found that the estimated clinical phase transition
`rates for all oncology biologics were consistently and substantially
`lower than those for the top 50 firm biologics. Phase I to phase II
`transition rates were 84% for top 50 firm biologics as compared
`with 70% for all oncology biologics. The difference in the phase
`II to phase III transition rate was much more pronounced, with
`53% of the top 50 firm biologics progressing from phase II to
`phase III whereas only 33% of the oncology compounds did so.
`The difference in successful phase III transitions was greater still;
`74% of the top 50 firm biologics transitioned from phase III to a
`
`Phase I–II
`
`Phase II–III
`
`Phase III to
`NDA/BLA
`submission
`
`NDA/BLA
`submission to
`NDA/BLA approved
`
`Phase I to
`NDA/BLA
`approved
`
`Hematologic only
`
`Solid tumor only
`
`Figure 3 Phase transition probabilities for cancer compounds by cancer type
`(first clinical testing, 1993–2004) . NDA/BLA, New Drug Application/Biologic
`License Application.
`
`regulatory submission, as compared with only 50% for all oncol-
`ogy biologics. Overall, the clinical approval success rate was more
`than twice as high for the top 50 firm biologics than that for the
`set of all oncology biologics (32 vs. 12%).
`The data set for this study was large enough to distinguish
`between compounds that were developed solely to treat solid
`tumors and those developed only for hematologic indications.
`Of the 625 compounds in the data set, 72% were developed
`only for solid tumor indications, 8% were developed only for
`hematologic cancers, and 20% were developed for both solid
`tumor and hematology indications. There were some differ-
`ences in the distribution of compounds by cancer type across
`molecule type. SMDs were more likely to be studied in both
`solid tumor and hematologic indications than were biologics.
`Whereas 10% of the biologics were studied in both cancer types,
`25% of the SMDs were tested in both. For biologics, 76% of the
`compounds were studied in solid tumors only, and 14% were
`studied in hematologic indications only. In the case of SMDs,
`70% were tested only in solid tumors, and just 5% were tested
`only in hematologic cancers.
`The success-rate results for hematologic-only and solid tumor-
`only compounds were pronounced (Figure 3). All of the transi-
`tion rates were higher for hematologic-only compounds than for
`solid tumor–only compounds. However, the distinctions were
`greatest for mid-to-late-stage clinical transitions. Whereas only
`38% of the solid tumor–only molecules progressed from phase
`II to phase III, 55% of the hematologic molecules did so. The
`differences in outcomes were especially pronounced for phase
`III transitions. The estimated successful transition rate from
`phase III to a regulatory submission was 41% for solid tumor–
`only compounds but a remarkable 88% for hematologic-only
`compounds. As a result, the estimated overall clinical approval
`success rate for hematologic-only compounds was more than
`three times higher than that for solid tumor–only compounds
`(36 vs. 10%).
`
`DEVELOPMENTAL SETTING
`The safety and efficacy of candidate cancer drugs in humans is
`not known before study. Therefore, because of ethical concerns,
`
`332
`
`VOLUME 94 NUMBER 3 | SEPTEMBER 2013 | www.nature.com/cpt
`
`state art
`
`Abraxis EX2081
`Cipla Ltd. v. Abraxis Bioscience, LLC
`IPR2018-00162; IPR2018-00163; IPR2018-00164
`
`

`

`these drugs are commonly evaluated in late-stage cancer
`patients who may have been heavily pretreated with approved
`products. Our data indicate that most phase III studies of can-
`cer drugs are performed in patients with metastatic disease or
`who have inoperable, locally advanced tumors. We examined
`the details of phase III clinical studies for cancer drugs in the
`data set that were either currently in phase III studies or had
`been terminated at phase III to determine what percentage were
`studied in an adjuvant setting, i.e., administered to patients
`after surgery to remove solid tumors. We found few instances
`of studies performed in an adjuvant setting. Of the 77 cancer
`drugs evaluated in phase III studies of any solid tumor type, we
`identified 12 (16%) drugs that had been studied specifically in
`an adjuvant setting. In these cases, the most frequent indica-
`tions studied were brain, ovarian, and breast cancers. The num-
`ber of drugs studied in an adjuvant setting that transitioned to
`regulatory review or from review to approval was insufficient
`for analysis.
`Of the cancer drugs studied at phase III as treatments for
` metastatic cancer, the clinical phase transitions were simi-
`lar regardless of whether the drug was studied exclusively in
`a metastatic setting or in a combination of metastatic, locally
`advanced/inoperable, or adjuvant settings (50.0% for all meta-
`static vs. 51.6% for metastatic only for phase III to regulatory
`submission; 47.2% for all metastatic vs. 51.5% for metastatic
`only for phase III to regulatory approval).
`
`ONCOLOGY SUCCESS RATES FOR LEAD AND SECONDARY
`INDICATIONS
`We obtained information on cancer indications studied in the
`clinic and their development status for the 625 compounds
`included in the study data set. We ordered the indications
`according to when they were first studied. Each compound has
`a first, or what we refer to as a lead, indication, and many had
`additional indications studied clinically. In total, the data set
`contains information on 2,055 cancer indications for the 625
`compounds, or an average of 3.3 indications per compound. The
`number of indications for individual compounds ranged from
`
`1 to 23. Sixty percent of the molecules were studied clinically in
`more than one indication.
`The average number of indications per molecule was 3.5 for
`SMDs and 2.7 for biologics. Examining molecules by cancer
`type, we found that the average number of indications per mol-
`ecule was 2.8 for molecules studied only in solid tumors, 3.9 for
`molecules studied only in hematologic indications, and 4.8 for
`molecules studied in both solid tumor and hematologic indica-
`tions. A final outcome was available for 78% of the first indica-
`tions, 65% of the second indications for those compounds that
`had a second indication pursued, and 55% of the third indica-
`tions for those compounds that had a third indication pursued.
`The transition and success-rate estimates shown above are all
`for analyses performed at a molecule level. That is, they are esti-
`mates of the likelihood that a molecule will proceed from one
`phase to the next or be approved if the molecule enters clinical
`testing for some indication. A molecule that obtains regula-
`tory marketing approval for any indication is taken above to
`be a success, even if the compound fails in a number of other
`indications. The molecule is counted as a failure only if it has
`failed in all indications pursued. With the exception of first indi-
`cations, success rates at an indication level, in theory, can be
`lower, higher, or the same as success rates at the molecule level.
`The qualitative relationship cannot be determined a priori. The
`relationship, therefore, is an empirical, not theoretical, issue.
`Because all compounds in the data set have a first cancer indica-
`tion, for first indications, the indication success rate can be no
`higher than the molecule success rate.
`Figure 4 shows estimated phase transition rates and overall
`clinical approval success rates for first, second, and third indica-
`tions pursued, for those compounds that had such indications.
`Note that the success rate for the lead, or first, indication need
`not be the same as the molecule success rate. This is because
`although every compound in the data set had a first indication,
`the first cancer indication pursued could fail, whereas a later
`indication could succeed. The results show this to be the case
`because the estimated clinical approval success rate for the lead,
`
`100%
`
`85.2%
`
`54.9%
`
`41.7%
`
`34.0%
`30.0%
`
`62.9%
`
`50.9%
`
`Transition probability
`
`92.3%
`
`87.5%
`
`66.7%
`
`52.4%
`
`33.0%
`30.3%
`25.7%
`
`88.9%
`35.7%
`
`95.1%
`
`88.9%
`
`59.4%
`
`Transition probability
`
`Phase I–II
`
`Phase II–III
`
`Phase III to
`NDA/BLA
`submission
`
`NDA/BLA
`submission to
`NDA/BLA approved
`
`Phase I to
`NDA/BLA
`approved
`
`Lead indication
`
`Second indication
`
`Third indication
`
`9.0% 8.2% 6.9%
`
`Phase I–II
`
`Phase II–III
`
`10.0%
`
`6.4%
`
`Phase III to
`NDA/BLA
`submission
`
`NDA/BLA
`submission to
`NDA/BLA approved
`
`Phase I to
`NDA/BLA
`approved
`
`Biologics
`
`SMD
`
`Figure 4 Phase transition probabilities by cancer indication number for
`cancer compounds first entering clinical testing (1993–2004). NDA/BLA, New
`Drug Application/Biologic License Application.
`
`Figure 5 Phase transition probabilities for lead indications of cancer
`compounds first entering clinical testing during 1993–2004 by molecule type.
`NDA/BLA, New Drug Application/Biologic License Application; SMD, small-
`molecule drug.
`
`CliniCal pharmaCology & TherapeuTiCS | VOLUME 94 NUMBER 3 | SEPTEMBER 2013
`
`333
`
`state art
`
`Abraxis EX2081
`Cipla Ltd. v. Abraxis Bioscience, LLC
`IPR2018-00162; IPR2018-00163; IPR2018-00164
`
`

`

`the study period, failure rates for indications that reach phase
`III testing were higher for secondary than for lead indications.
`The success or failure of secondary indications may be correlated
`with the success or failure of a lead indication. We investigated this
`hypothesis by examining success rates for second and third indi-
`cations conditional on lead indication success or failure. Figure 6
`suggests that success in reaching the marketplace for second indi-
`cations pursued was highly correlated with whether the first indi-
`cation pursued reached the marketplace. If a second indication
`was pursued and the first indication was a success, then marketing
`approval of a second indication was more likely than not. However,
`if the lead indication failed, then the likelihood of success for a
`second indication was only 2.5%. The results are qualitatively simi-
`lar for third indications. If a third indication was pursued, then
`likelihood of success for that third indication falls from 42 to <2%,
`dependent on whether the first indication was a success.
`
`SUMMARY
`We developed comprehensive estimates of phase transition
`and clinical approval success rates for cancer drugs pursued
`in the clinic by the biopharmaceutical industry. Firms of all
`sizes were included. As compared with the results of a previ-
`ous study for all compounds originated by the top 50 firms,16
`we found cancer drug success rates at the molecule level to be
`somewhat lower than for drugs in general. The results did vary
`substantially across cancer types, with the approval success rate
`for compounds studied only for hematologic indications more
`than triple that for compounds studied only in solid tumor indi-
`cations. We found large-molecule cancer compounds to have
`somewhat lower success rates as compared with small-molecule
`cancer drugs, in contrast to the much higher success rates found
`for large molecules for compounds in general.16
`To our knowledge, this is the first published study that devel-
`ops success-rate estimates by indication order and the first to
`examine the issue of whether success in obtaining regulatory
`approval for marketing is correlated across indications. This is
`an especially important consideration for cancer drug devel-
`opment because many investigational cancer compounds are
`studied in multiple cancers. We found clinical approval success
`rates by indication order to be lower than at the molecule level.
`The first, second, and third indications pursued had success rates
`of only 9, 8, and 7%, respectively. However, the success rates for
`second and third indications appear to be highly correlated with
`lead indication success. Although we estimated that 55% of the
`second indications pursued and 42% of the third indications
`pursued obtain marketing approval if the lead indication was a
`success, only ~2% of the second and third indications succeeded
`if the lead indication was a failure.
`The only other study of which we are aware that addresses
`cancer indication success rates was described in a conference
`presentation.17 The study used a different methodology and
`combined all second and later indications. The methodol-
`ogy that was used appears to be better suited for picking up
`success-rate trends early, than for determining success-rate
`levels because the phase transitions involve different sets of
`compounds (i.e., a given set of compounds is not followed over
`
`95.7%
`
`84.7%
`
`76.5%
`
`75.0%
`
`100% 100%
`
`15.3%
`
`19.1%
`
`54.9%
`
`2.5%
`
`Transition probability
`
`Phase I–II
`
`Phase II–III
`
`Phase III to
`NDA/BLA
`submission
`
`NDA/BLA
`submission to
`NDA/BLA approved
`
`Phase I to
`NDA/BLA
`approved
`
`Lead indication success
`
`Lead indication failure
`
`Figure 6 Phase transition probabilities for second indications of cancer
`compounds first entering clinical testing (1993–2004), conditional on lead
`indication success or failure. NDA/BLA, New Drug Application/Biologic
`License Application.
`
`or first, indication is lower than the molecule approval success
`rate (9.0 vs. 13.4%).
`Although this need not be the case in general, here the success
`rates for second (8.2%) and third indications (6.9%) are much
`lower. Very high percentages of second and third indications make
`it to phase II testing in comparison to first indications (89 and 95
`vs. 59%, respectively). The high transition rates shown here for
`second and third indications can occur if phase I testing shows that
`the compound is safe enough to proceed to phase II, but there is
`insufficient evidence of efficacy for the lead indication. Firms may
`then proceed to phase II testing in secondary indications. Once
`testing in the lead and secondary indications proceeds to phase
`II, the transition rates to phase III are more similar. Where the
`second and third indications falter relative to lead indications is in
`the