[Skip to Content]
[Skip to Content Landing]
Figure.  Definitions of Study Duration (for Response Rate, Progression-Free Survival, and Overall Survival and Patient-Reported Outcomes) When the Data Cutoff Date Is Available
Definitions of Study Duration (for Response Rate, Progression-Free Survival, and Overall Survival and Patient-Reported Outcomes) When the Data Cutoff Date Is Available

If a cutoff date was unavailable, we used (1) median time to response plus duration of response, (2) median progression-free survival, or (3) median overall survival.

Table 1.  Characteristics of 188 Drug Indications From 107 Cancer Drugs Approved by the FDA From 2006 to 2017
Characteristics of 188 Drug Indications From 107 Cancer Drugs Approved by the FDA From 2006 to 2017
Table 2.  Proportion of 172 Drugs Granted Accelerated Approval and Any Approval on the Basis of Each End-Point Type Across Lines of Therapy Settings
Proportion of 172 Drugs Granted Accelerated Approval and Any Approval on the Basis of Each End-Point Type Across Lines of Therapy Settings
Table 3.  Comparison of Study Duration for RR, PFS, and OS End Points Across Line of Therapy Using the Original Trial Data Set
Comparison of Study Duration for RR, PFS, and OS End Points Across Line of Therapy Using the Original Trial Data Set
Table 4.  Multivariate Linear Regression Examining the Association Between Study Duration (Main Outcome) and the Hypothesized 4 Variables (Exposures) Using the Original Trial Data Set
Multivariate Linear Regression Examining the Association Between Study Duration (Main Outcome) and the Hypothesized 4 Variables (Exposures) Using the Original Trial Data Set
1.
Baker  SG, Kramer  BS.  Surrogate endpoint analysis: an exercise in extrapolation.  J Natl Cancer Inst. 2013;105(5):316-320. doi:10.1093/jnci/djs527PubMedGoogle ScholarCrossref
2.
Gellad  WF, Kesselheim  AS.  Accelerated approval and expensive drugs—a challenging combination.  N Engl J Med. 2017;376(21):2001-2004. doi:10.1056/NEJMp1700446PubMedGoogle ScholarCrossref
3.
Gyawali  B, Kesselheim  AS.  Reinforcing the social compromise of accelerated approval.  Nat Rev Clin Oncol. 2018;15(10):596-597. doi:10.1038/s41571-018-0066-3PubMedGoogle ScholarCrossref
4.
Kim  C, Prasad  V.  Strength of validation for surrogate end points used in the US Food and Drug Administration’s approval of oncology drugs.  Mayo Clin Proc. 2016;91(6):713-725. doi:10.1016/j.mayocp.2016.02.012PubMedGoogle ScholarCrossref
5.
Carpenter  D, Kesselheim  AS, Joffe  S.  Reputation and precedent in the bevacizumab decision.  N Engl J Med. 2011;365(2):e3. doi:10.1056/NEJMp1107201PubMedGoogle ScholarCrossref
6.
Prasad  V, Kim  C, Burotto  M, Vandross  A.  The Strength of association between surrogate end points and survival in oncology: a systematic review of trial-level meta-analyses.  JAMA Intern Med. 2015;175(8):1389-1398. doi:10.1001/jamainternmed.2015.2829PubMedGoogle ScholarCrossref
7.
Haslam  A, Hey  SP, Gill  J, Prasad  V.  A systematic review of trial-level meta-analyses measuring the strength of association between surrogate end-points and overall survival in oncology.  Eur J Cancer. 2019;106:196-211. doi:10.1016/j.ejca.2018.11.012PubMedGoogle ScholarCrossref
8.
Prasad  V, Mailankody  S.  Research and development spending to bring a single cancer drug to market and revenues after approval.  JAMA Intern Med. 2017;177(11):1569-1575. doi:10.1001/jamainternmed.2017.3601PubMedGoogle ScholarCrossref
9.
US Food and Drug Administration. Hematology/oncology (cancer) approvals & safety notifications. https://www.fda.gov/Drugs/InformationOnDrugs/ApprovedDrugs/ucm279174.htm. Accessed February 27, 2019.
10.
Harris  PA, Taylor  R, Thielke  R, Payne  J, Gonzalez  N, Conde  JG.  Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support.  J Biomed Inform. 2009;42(2):377-381. doi:10.1016/j.jbi.2008.08.010PubMedGoogle ScholarCrossref
11.
Lakdawalla  DN, Chou  JW, Linthicum  MT, MacEwan  JP, Zhang  J, Goldman  DP.  Evaluating expected costs and benefits of granting access to new treatments on the basis of progression-free survival in non-small-cell lung cancer.  JAMA Oncol. 2015;1(2):196-202. doi:10.1001/jamaoncol.2015.0203PubMedGoogle ScholarCrossref
12.
Beaver  JA, Howie  LJ, Pelosof  L,  et al.  A 25-year experience of US Food and Drug Administration accelerated approval of malignant hematology and oncology drugs and biologics: a review.  JAMA Oncol. 2018;4(6):849-856. doi:10.1001/jamaoncol.2017.5618PubMedGoogle ScholarCrossref
13.
Zettler  M, Nabhan  C.  Fulfillment of postmarketing requirements to the FDA for therapies granted oncology indications between 2011 and 2016.  JAMA Oncol. 2018;4(7):993-994. doi:10.1001/jamaoncol.2018.0610PubMedGoogle ScholarCrossref
14.
Kim  C, Prasad  V.  Cancer drugs approved on the basis of a surrogate end point and subsequent overall survival: an analysis of 5 years of US Food and Drug Administration approvals.  JAMA Intern Med. 2015;175(12):1992-1994. doi:10.1001/jamainternmed.2015.5868PubMedGoogle ScholarCrossref
15.
Kemp  R, Prasad  V.  Surrogate endpoints in oncology: when are they acceptable for regulatory and clinical decisions, and are they currently overused?  BMC Med. 2017;15(1):134. doi:10.1186/s12916-017-0902-9PubMedGoogle ScholarCrossref
16.
Davis  C, Naci  H, Gurpinar  E, Poplavska  E, Pinto  A, Aggarwal  A.  Availability of evidence of benefits on overall survival and quality of life of cancer drugs approved by European Medicines Agency: retrospective cohort study of drug approvals 2009-13.  BMJ. 2017;359:j4530. doi:10.1136/bmj.j4530PubMedGoogle ScholarCrossref
17.
Qunaj  L, Jain  RH, Atoria  CL, Gennarelli  RL, Miller  JE, Bach  PB.  Delays in the publication of important clinical trial findings in oncology.  JAMA Oncol. 2018;4(7):e180264. doi:10.1001/jamaoncol.2018.0264PubMedGoogle ScholarCrossref
18.
Massey  PR, Wang  R, Prasad  V, Bates  SE, Fojo  T.  Assessing the eventual publication of clinical trial abstracts submitted to a large annual oncology meeting.  Oncologist. 2016;21(3):261-268. doi:10.1634/theoncologist.2015-0516PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    April 1, 2019

    Estimation of Study Time Reduction Using Surrogate End Points Rather Than Overall Survival in Oncology Clinical Trials

    Author Affiliations
    • 1Division of Hematology Oncology, Knight Cancer Institute, Oregon Health & Science University, Portland
    • 2School of Medicine, Oregon Health & Science University, Portland
    • 3Department of Public Health and Preventive Medicine, Oregon Health & Science University, Portland
    • 4Center for Health Care Ethics, Oregon Health & Science University, Portland
    JAMA Intern Med. 2019;179(5):642-647. doi:10.1001/jamainternmed.2018.8351
    Key Points

    Question  How much time is saved in the 6 to 15 years of drug development by using surrogate end points rather than overall survival as the basis for US Food and Drug Administration approval in contemporary cancer drug trials?

    Findings  In this retrospective study of 107 oncology drugs with 188 indications, compared with the duration of studies using overall survival as the basis for approval, progression-free survival was associated with a savings of 11 months, and response rate with a savings of 19 months.

    Meaning  Using surrogate end points for oncology drug approval was associated with reduced drug development time by approximately 11 months compared with an overall survival end point.

    Abstract

    Importance  Surrogate end points in oncology trade the advantage of reducing the time needed to conduct clinical trials for the disadvantage of greater uncertainty regarding the treatment effect on patient-centered end points, such as overall survival (OS) and quality of life.

    Objective  To quantify the amount of time saved through the acceptance of surrogate end points, including response rate (RR) and progression-free survival (PFS).

    Design, Setting, and Participants  This retrospective study of US Food and Drug Administration (FDA) oncology approvals and their drug registration trials based on actual publication analyzed the original and updated clinical trials data that led to FDA-approved drug indications in oncology from 2006 to 2017 by using existing publications, conference abstracts, and package inserts from the FDA. Data related to cancer type, line of therapy (first-line, second-line, and third- or later-line treatment of advanced or metastatic disease), FDA approval type, end point basis for approval (RR, PFS, or OS/quality of life), sample size, accrual rate, and drug RR were extracted by March 23, 2018. All data were analyzed by July 13, 2018.

    Main Outcomes and Measures  The main outcome was the study duration needed to complete the primary end point analysis used for each drug indication approval. This was estimated from reported enrollment dates, analysis cutoff dates, time to response, median duration of response, median PFS, and median OS.

    Results  In total, 188 distinct indications among 107 cancer drugs were identified. The RR was more often used for FDA approval in subsequent lines of therapy (17 of 71 drug indications [24%] in first-line therapy vs 34 of 77 drug indications [44%] in second-line therapy vs 19 of 24 drug indications [79%] in third- or later-line therapy, P < .001). Study duration for PFS (median, 31 [range, 10-104] months) was similar to that for OS (median, 33 [range, 12-117] months; P = .31), whereas study duration for RR (median, 25 [range, 11-54] months) was shorter than that for OS (P = .001). In multivariate analysis, compared with using OS, use of PFS as the end point was associated with study durations that were shorter by a mean of 11 months (95% CI, 5-17 months), and the use of RR as the end point was associated with study durations that were shorter by a mean of 19 months (95% CI, 13-25 months).

    Conclusions and Relevance  From the findings of this study, an estimated 11 months appeared to be needed (ie, approximately 12% longer in the drug development cycle) to assess the OS benefit of a cancer drug. This study’s findings suggest that this must be weighed against the downside of increased uncertainty of clinical benefit arising from using surrogate end points.

    Introduction

    Overall survival (OS) and validated patient-reported outcomes (PROs) are clinical end points in oncology clinical trials that are intrinsically meaningful to patients. The US Food and Drug Administration (FDA) has the ability to approve cancer treatments on the basis of surrogate end points, such as response rate (RR) or progression-free survival (PFS). These end points may arrive sooner than OS, thereby speeding approval, which is a benefit for ultimately effective drugs. At the same time, surrogate end points carry the disadvantage of greater uncertainty for clinical efficacy,1-4 potentially allowing drugs that do not improve clinically meaningful end points to enter the market. In addition to high-profile examples in which surrogate end points failed to determine survival benefit, such as bevacizumab in metastatic breast cancer,5 an umbrella meta-analysis of oncology clinical trials demonstrated a weak association between commonly used surrogate end points and OS in most cancer types.6,7

    Although it is widely asserted that RR and PFS reduce development times, we are unaware of any empirical analysis that sought to estimate the extent to which study duration is reduced. One underdiscussed aspect of surrogate end points is that the FDA often relies on both RR and duration of response prior to approval (to show responses are enduring, not fleeting), and this additional follow-up time may decrease the time saved from using RR end points. Prior work has suggested that development times—from early clinical trials to drug approval—can range from an average of 5.8 to 15.2 years,8 and quantifying the benefit of surrogates in this context would be valuable. However, to date, no such attempt has been made, nor method of estimation proposed, across all contemporary oncology drug registration trials to better inform policymaking in the approval process. Therefore, in the present study, we sought to estimate and compare the study durations needed to assess common cancer end points (RR, PFS, and OS) among all recent drugs approved by the FDA and to provide a framework for estimating the additional time needed to establish the OS benefit of a cancer drug.

    Methods
    Study Design

    This study is a retrospective analysis of all original and updated clinical trials that led to FDA accelerated or regular approval of cancer-specific drugs from 2006 to 2017 providing details on the disease setting, basis for approval, sample size and accrual rate, study duration (with respect to RR, PFS, and OS end points), and results of important end points. This study obtained data directly from existing publications with no protected health information and therefore was not submitted to the institutional review board.

    Data Set

    We identified FDA-approved drug indications for any malignant neoplasm from February 1, 2006, to December 31, 2017, through the FDA hematology/oncology (cancer) approvals and safety notifications website.9 We counted every approved drug, its approved indications, and its expansion of prior indications as separate entries if they led to unique labeling supported by distinct clinical trials. For example, nivolumab was first approved for first-line treatment of metastatic melanoma, followed by second-line treatment of metastatic lung squamous cell carcinoma, and then finally for second-line treatment of metastatic lung adenocarcinoma. We counted each clinical trial separately because each new indication was linked to a respective, unique clinical trial for a different disease setting. Minor modifications were not counted again as separate entries. Approvals for dose schedule changes, supportive medications, pediatric cancers, genetic syndromes, and nonneoplastic hematologic disorders were excluded from this investigation. See the eMethods in the Supplement for details of the inclusion and exclusion criteria, with explicit examples. The main comparison analyses excluded indications in the adjuvant and maintenance settings. All data were analyzed by July 13, 2018.

    We identified all clinical trials relevant to each approval indication through the corresponding package insert (specifically section 14, which gives details on clinical studies). We defined the original trial as the first clinical trial leading to an approval indication. We defined the updated trial as the currently available clinical trial with the largest sample size and longest follow-up published before March 23, 2018. The original trial data set was used for primary analysis, whereas the updated trial data set was used for sensitivity analysis. See the eMethods in the Supplement for the detailed criteria used for each data set.

    Data Extraction

    For each published original trial and updated trial corresponding to every distinct indication label of a cancer drug, we recorded the following: cancer type, line of therapy, FDA approval process, end point basis used for initial approval, enrollment start and end dates, sample size, analysis cutoff dates for each end point, and end-point results (including RR, PFS, and OS). We collected PRO results only for drugs in which the FDA used PROs as the basis for approval. See the eMethods in the Supplement for details of the data extraction. All data were managed using REDCap10 (research electronic data capture) tools hosted at Oregon Health & Sciences University in Portland.

    Main Outcomes

    The Figure and eFigure 1 in the Supplement graphically describe the definition of the main outcome: study duration. We defined study duration based on end point type (RR, PFS, or OS and PRO combined) by the sum of enrollment duration (from the first patient enrolled to the last patient enrolled) and the time from the last patient enrolled to the analysis cutoff date of the end point used for FDA approval. When cutoff dates were not available, the study duration with an RR end point was defined by the sum of the enrollment duration, the median time to response, and the median duration of response. When a cutoff date was not available for PFS, the study duration with a PFS end point was defined by the sum of enrollment duration plus the median PFS of the drug. Likewise, if a cutoff date was not available for OS, the study duration with an OS end point was defined by the sum of enrollment period and the median OS of the drug. The few studies that used PROs all had cutoff dates. They were categorized with OS studies because both are considered definite end points by the FDA. The main outcome used these definitions to estimate the time required to reach an intended sample size and complete an end-point analysis.

    We defined the study duration in the original trial data set by the end point that formed the basis for FDA approval. With regard to the updated trial data set used for sensitivity analysis, all 3 possible study durations (RR, PFS, and OS and PRO combined) were recorded for each clinical trial, if available. Two other definitions of study duration adjusting the enrollment duration were recorded using the updated trial data set for additional sensitivity analysis. All 3 outcome definitions are explained in detail in the eMethods in the Supplement.

    Statistical Analysis

    We used Kaplan-Meier analyses to estimate the time to conversion of accelerated to regular approval, with a cutoff date of March 23, 2018, to provide additional context for our data set. The main univariate and multivariate analyses focused on primary systemic treatment of advanced or metastatic cancers and excluded indications in the adjuvant and maintenance settings, and we completed all analyses using both the original trial and updated trial data sets.

    For univariate analyses, we compared the proportion of drug indications initially granted accelerated approval by line of therapy (first-line, second-line, and third- or later-line treatment of advanced or metastatic cancers) using Fisher exact tests. In addition, we compared the proportion of drug indications approved on the basis of RR, PFS, and OS by line of therapy using Fisher exact tests. Finally, we compared the study duration by each end-point type. Specifically, we compared study duration per RR and PFS end points to study duration per the OS end point (the reference) using Mann-Whitney tests (RR vs OS, and PFS vs OS) and Kruskal-Wallis tests (RR vs PFS vs OS).

    Multivariate linear regression analysis to estimate the time from first patient enrollment to the end point (study duration) was performed with the following variables: basis for approval (RR, PFS, or OS), accrual rate, line of therapy, and drug activity (ie, RR or change in RR). In other words, we asked how much time is saved by the acceptance of PFS or RR in clinical trials, adjusting for the accrual rate (ie, clinical trials that recruit more rapidly will get their results sooner), line of therapy (ie, end points are reached sooner in latter lines of therapy owing to greater disease severity and higher event rates), and drug activity (ie, more potent compounds yield more rapid results). The accrual rate is a product of prevalence, number of options, the desirability of the clinical trial, and the ability of investigators to recruit. Thus, this multivariate model, which is explained in detail in the eMethods in the Supplement, tests the question of to what degree surrogates are associated with faster drug approval over OS when holding constant the activity, accrual, and cancer setting. We used SAS software, version 9.4 (SAS Institute Inc) for statistical analyses but edited the figures using Microsoft PowerPoint. A 2-sided P < .05 was considered statistically significant.

    Results

    In total, 188 distinct indications among 107 drugs approved for treatment of solid tumors and hematologic malignant neoplasms were identified from 2006 to 2017. The median study duration was 1 month longer for studies with a PFS outcome compared with studies with OS and PRO (combined) end points (Table 1). The study duration was not available for 5 clinical trials in the original trial data set but was available in the updated trial data set. The full list of these drugs and their indications are given in eTable 1 in the Supplement. With regard to surrogate end points, 71 of 188 drug indications (38%) used RR end points, and 65 of 188 (34%) used PFS end points. The OS and PRO end points accounted for 52 of 188 indications (28%). Only 3 drug indications (2%) used PRO end points as the basis for FDA approval.

    Our analysis also indicated that 56 drug indications (30%) were initially granted accelerated approval, of which 30 (54%) had been converted to regular approval by March 23, 2018. The median time to conversion was 42 months (interquartile range, 26-57 months), as noted in eFigure 2 in the Supplement. Accelerated approval as opposed to regular approval was more likely to be sought for subsequent lines of therapy (12 of 71 drug indications [17%] for first-line, 26 of 77 drug indications [34%] for second-line, and 16 of 24 drug indications [67%] for third-or later-line; P < .001), (Table 2). Compared with other end-point types, RR was used more frequently in subsequent lines of therapy (17 of 71 drug indications [24%] for first-line vs 34 of 77 drug indications [44%] for second-line vs 19 of 24 drug indications [79%] for third- or later-line; P < .001). Surrogate end points (RR and PFS combined) showed a trend toward being used in the third- or later-line setting (49 of 71 drug indications [69%] for first-line vs 53 of 77 drug indications [69%] for second-line vs 21 of 24 drug indications [87%] for third- or later-line; P = .17), but they were used in all lines of therapy (Table 2).

    The study duration for PFS (median, 31 months; range, 10-104 months) was similar to that for OS and PRO combined (median, 33; range, 12-117 months, P = .31). The study duration for RR was shorter than that for OS and PRO combined (median, 25 months; range, 11-54 months vs 33 months; range, 12-117 months; P = .001). Descriptively, the median study duration for RR was 6 months shorter than that for OS and PRO combined, and the study duration for PFS was 2 months shorter than that for OS and PRO combined. Table 3 gives the study duration (per RR, PFS, and OS and PRO combined) for each line of therapy in the original trial data set. The difference in study duration among these 3 end-point types was less apparent in the third- or later-line setting.

    The study durations for RR and PFS were shorter than that for OS by no more than 13 months using the updated trial data set in a sensitivity analysis (eTable 2 in the Supplement). Additional sensitivity analysis using the other 2 definitions of study duration, which adjusted the enrollment duration, yielded similar results (eTable 3 and eTable 4 in the Supplement). The accrual rate was slower for studies with an RR end point (median, 9 [range, 1-66] persons/month) compared with studies having an OS end point (median, 32 [range, 2-153] persons/month), and the sample size was a median of 136 (range, 12-1052) for RR and a median of 676 (range, 79-1725) for OS (eTable 5 in the Supplement).

    In the multivariate analysis, the use of PFS instead of OS as the regulatory end point was associated with a study duration shorter by a mean of 11 months (95% CI, 5-17 months). The use of RR instead of OS as the regulatory end point was associated with a study duration shorter by a mean of 19 months (95% CI, 13-25 months). Table 4 gives the multivariate linear regression analysis results between study durations to achieve the end point used for FDA approval (main outcome) and the type of end point chosen as the basis for approval (exposure of interest) while adjusting for accrual rate, line of therapy, and drug RR, per the original trial data set. The intercept of 46 months (95% CI, 41-50 months) was the estimated study duration for an oncology drug across our data set when OS was used as the basis for approval. A subsequent line of therapy and an increase in accrual rate were both associated with a shorter study duration. There was no significant association with drug RR, and this variable was excluded in the final model. This same analysis was also conducted using the updated trial data set, yielding similar results (eTable 6 in the Supplement).

    Discussion

    Surrogate end points are increasingly used in oncology clinical trials with the intention of expediting drug development. Although RR serves as an important preliminary end point to assess the activity of cancer drugs, it is not a direct measure of patient benefit and often poorly correlates with survival.6 Our study showed that study durations for RR and PFS end points were shorter than that for the OS end point only in the first-line and second-line setting across both original and updated trial data sets. Overall the difference was less than 1 year on unadjusted analysis and was associated with duration shortening by an estimated 19 months (RR) and 11 months (PFS) when adjusting for accrual rate and line of therapy. Although a time-to-event end point, PFS is a surrogate, and several clinical trials have demonstrated PFS benefits that fail to translate to improvements in OS.11

    The median time to develop a modern cancer drug has been estimated at 7.3 years,8 and our results showed that 38% of the current US cancer drugs were approved on the basis of RR, 34% on the basis of PFS, and 28% on the basis of OS and PRO. If only meaningful clinical end points like OS were used, we would estimate 38% of clinical trials to be 19 months slower, 34% to be 11 months slower, and 28% unchanged. Combining these calculations, the median time to develop a modern cancer drug using only definite end points would therefore be associated with an increased duration of 11 months, extending the previously observed 7.3 years8 to 8.2 years, or 12% longer.

    Researchers have found that 5 cancer drugs initially approved through the accelerated program in the last decade were withdrawn owing to a lack of OS benefit yet remained on the market for 3.4 to 11.5 years.12 There are at least 3 cancer drugs currently approved through the accelerated program that remain on the market despite having failed to show OS benefit in subsequent clinical trials.13 Subsequent clinical trials designed to fulfill postmarketing requirements can use the same end points with longer follow-up and larger sample sizes, other surrogate end points, or patient-centered end points (OS or quality of life), with the choice dependent on the specific discretion of FDA commissioners. An analysis of 36 cancer drugs approved on the basis of a surrogate end point found that only 5 later improved OS, with a median of 4.5 years of follow-up.14 Because manufacturers have an incentive to bring existing drugs to earlier lines of therapy to reach more customers and increase market shares,15 regulators must institute valid end points for confirming benefit and expanding subsequent indications. Notably, many approved cancer drugs have been found to offer additional survival of a few months (1.0-5.8 months).16

    Limitations

    Our study has 6 limitations. First, not all FDA label updates are reported on the FDA website; however, we made a comprehensive effort to review all versions of every FDA-approved package insert included here. Nevertheless, there may have been inadvertent exclusions.

    Second, not all study enrollment start and end dates as well as data cutoff dates were reported. Thus, we used the available end points and performed sensitivity analyses using a modified definition of study durations as well as a direct comparison of all 3 end points. This sensitivity analysis did not materially change the conclusions.

    Third, we recognize there are relevant clinical trials with negative results that have not led to FDA approval and that further review of those studies would be important. However, such studies are not readily available on the FDA website; indeed, negative results may be delayed or never published.17,18 We encourage other researchers to perform broader studies.

    Fourth, study design differences (randomized and controlled vs single arm) and the actual primary end point of each study (rather than the basis for FDA approval) could contribute to the study duration in each indication. Each clinical trial was counted as a single observation and not weighted by the number of participants in each clinical trial or by differences in study design. The sample size of each study was not directly associated with study duration in our analysis.

    Fifth, intrinsic characteristics of specific cancer conditions may not be adequately captured by our multivariate model. For instance, the lethality of a cancer could influence the event rate and the speed at which it could be assessed in a clinical trial; in addition, the prevalence of a cancer type and its advanced or metastatic subgroup could affect the enrollment duration. The efficacy of a drug will also affect clinical trial time. In general, efficacious drugs being used for treatment of lethal cancers will reach survival end points quicker than marginal drugs for treatment of indolent cancers. We included RR and change in RR as surrogates for drug activity, yet our results found them to be noncontributing variables. We could not calculate cancer lethality because the natural history of a specific cancer is heterogeneous and inconstant, and the natural history of a specific molecular alteration may be unknown or not reported. We were able to adjust for line of therapy. Nevertheless, we invite other investigators to refine the estimates produced here and view our estimates as a starting point for discussion.

    Finally, we recognize that the vastly different financial resources behind every clinical trial could also affect the enrollment duration, effect size from the study cost, and the overall study duration, but this limitation could be addressed only if the expenditure of every licensing clinical trial were reported publicly.

    Conclusions

    In short, our findings suggest that, among clinical trials leading to FDA approval, the use of RR was associated with a median study duration of 25 months (range, 11-54 months), PFS with a median study duration of 31 months (range, 10-104 months), and OS with a median study duration of 33 months (range, 12-117 months). The surrogate end points RR and PFS appeared to be associated with greater uncertainty regarding the actual benefit and also appeared to be associated with development times of approximately 19 and 11 months (RR and PFS, respectively) shorter, when adjusting for line of therapy, RR, and accrual rate. When contrasted against the 7.3 years of development time for the current cancer drugs, surrogates seemed to be associated with reduced development times by only 11 months, or 12% of development time. Whether this tradeoff is beneficial or detrimental to patients deserves further scrutiny.

    Back to top
    Article Information

    Accepted for Publication: December 9, 2018.

    Corresponding Author: Emerson Y. Chen, MD, Division of Hematology Oncology, Knight Cancer Institute, Oregon Health & Science University, 3181 SW Sam Jackson Park Rd, Ste L586, Portland, OR 97239 (cheem@ohsu.edu).

    Published Online: April 1, 2019. doi:10.1001/jamainternmed.2018.8351

    Author Contributions: Drs Chen and Prasad had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Chen, Joshi, Prasad.

    Acquisition, analysis, or interpretation of data: Chen, Joshi, Tran.

    Drafting of the manuscript: Chen, Prasad.

    Critical revision of the manuscript for important intellectual content: Chen, Joshi, Tran.

    Statistical analysis: Chen.

    Administrative, technical, or material support: Chen, Joshi, Tran.

    Supervision: Prasad.

    Conflict of Interest Disclosures: Dr Prasad reported receiving royalties from his book Ending Medical Reversal; funding from the Laura and John Arnold Foundation; honoraria for grand rounds or lectures from several universities, medical centers, and professional societies; and payments for contributions to Medscape. Dr Prasad hosts the podcast Plenary Session, which has Patreon backers. No other disclosures were reported.

    Funding/Support: This work was supported by grant 1 UL1 RR024140 01 from the Oregon Clinical and Translational Research Institute.

    Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    References
    1.
    Baker  SG, Kramer  BS.  Surrogate endpoint analysis: an exercise in extrapolation.  J Natl Cancer Inst. 2013;105(5):316-320. doi:10.1093/jnci/djs527PubMedGoogle ScholarCrossref
    2.
    Gellad  WF, Kesselheim  AS.  Accelerated approval and expensive drugs—a challenging combination.  N Engl J Med. 2017;376(21):2001-2004. doi:10.1056/NEJMp1700446PubMedGoogle ScholarCrossref
    3.
    Gyawali  B, Kesselheim  AS.  Reinforcing the social compromise of accelerated approval.  Nat Rev Clin Oncol. 2018;15(10):596-597. doi:10.1038/s41571-018-0066-3PubMedGoogle ScholarCrossref
    4.
    Kim  C, Prasad  V.  Strength of validation for surrogate end points used in the US Food and Drug Administration’s approval of oncology drugs.  Mayo Clin Proc. 2016;91(6):713-725. doi:10.1016/j.mayocp.2016.02.012PubMedGoogle ScholarCrossref
    5.
    Carpenter  D, Kesselheim  AS, Joffe  S.  Reputation and precedent in the bevacizumab decision.  N Engl J Med. 2011;365(2):e3. doi:10.1056/NEJMp1107201PubMedGoogle ScholarCrossref
    6.
    Prasad  V, Kim  C, Burotto  M, Vandross  A.  The Strength of association between surrogate end points and survival in oncology: a systematic review of trial-level meta-analyses.  JAMA Intern Med. 2015;175(8):1389-1398. doi:10.1001/jamainternmed.2015.2829PubMedGoogle ScholarCrossref
    7.
    Haslam  A, Hey  SP, Gill  J, Prasad  V.  A systematic review of trial-level meta-analyses measuring the strength of association between surrogate end-points and overall survival in oncology.  Eur J Cancer. 2019;106:196-211. doi:10.1016/j.ejca.2018.11.012PubMedGoogle ScholarCrossref
    8.
    Prasad  V, Mailankody  S.  Research and development spending to bring a single cancer drug to market and revenues after approval.  JAMA Intern Med. 2017;177(11):1569-1575. doi:10.1001/jamainternmed.2017.3601PubMedGoogle ScholarCrossref
    9.
    US Food and Drug Administration. Hematology/oncology (cancer) approvals & safety notifications. https://www.fda.gov/Drugs/InformationOnDrugs/ApprovedDrugs/ucm279174.htm. Accessed February 27, 2019.
    10.
    Harris  PA, Taylor  R, Thielke  R, Payne  J, Gonzalez  N, Conde  JG.  Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support.  J Biomed Inform. 2009;42(2):377-381. doi:10.1016/j.jbi.2008.08.010PubMedGoogle ScholarCrossref
    11.
    Lakdawalla  DN, Chou  JW, Linthicum  MT, MacEwan  JP, Zhang  J, Goldman  DP.  Evaluating expected costs and benefits of granting access to new treatments on the basis of progression-free survival in non-small-cell lung cancer.  JAMA Oncol. 2015;1(2):196-202. doi:10.1001/jamaoncol.2015.0203PubMedGoogle ScholarCrossref
    12.
    Beaver  JA, Howie  LJ, Pelosof  L,  et al.  A 25-year experience of US Food and Drug Administration accelerated approval of malignant hematology and oncology drugs and biologics: a review.  JAMA Oncol. 2018;4(6):849-856. doi:10.1001/jamaoncol.2017.5618PubMedGoogle ScholarCrossref
    13.
    Zettler  M, Nabhan  C.  Fulfillment of postmarketing requirements to the FDA for therapies granted oncology indications between 2011 and 2016.  JAMA Oncol. 2018;4(7):993-994. doi:10.1001/jamaoncol.2018.0610PubMedGoogle ScholarCrossref
    14.
    Kim  C, Prasad  V.  Cancer drugs approved on the basis of a surrogate end point and subsequent overall survival: an analysis of 5 years of US Food and Drug Administration approvals.  JAMA Intern Med. 2015;175(12):1992-1994. doi:10.1001/jamainternmed.2015.5868PubMedGoogle ScholarCrossref
    15.
    Kemp  R, Prasad  V.  Surrogate endpoints in oncology: when are they acceptable for regulatory and clinical decisions, and are they currently overused?  BMC Med. 2017;15(1):134. doi:10.1186/s12916-017-0902-9PubMedGoogle ScholarCrossref
    16.
    Davis  C, Naci  H, Gurpinar  E, Poplavska  E, Pinto  A, Aggarwal  A.  Availability of evidence of benefits on overall survival and quality of life of cancer drugs approved by European Medicines Agency: retrospective cohort study of drug approvals 2009-13.  BMJ. 2017;359:j4530. doi:10.1136/bmj.j4530PubMedGoogle ScholarCrossref
    17.
    Qunaj  L, Jain  RH, Atoria  CL, Gennarelli  RL, Miller  JE, Bach  PB.  Delays in the publication of important clinical trial findings in oncology.  JAMA Oncol. 2018;4(7):e180264. doi:10.1001/jamaoncol.2018.0264PubMedGoogle ScholarCrossref
    18.
    Massey  PR, Wang  R, Prasad  V, Bates  SE, Fojo  T.  Assessing the eventual publication of clinical trial abstracts submitted to a large annual oncology meeting.  Oncologist. 2016;21(3):261-268. doi:10.1634/theoncologist.2015-0516PubMedGoogle ScholarCrossref
    ×