Common criticisms of the accelerated approvals program: A context and data-driven assessment of overall survival in oncology indications.

Abstract

e23035 Background: The accelerated approvals (AA) program uses surrogate endpoints to enable early access to promising therapies. The AA program receives scrutiny over surrogate endpoints with critics suggesting the program approves ineffective therapies without significant overall survival (OS) data. This analysis focused on aspects of trial design and efficacy to comprehensively evaluate OS benefit among oncology drug-indications approved on a non-OS endpoint. Methods: FDA AA oncology drug-indications from 2013 to 2023 converted to a regular approval (RA) were categorized by the endpoint supporting the conversion: OS vs. non-OS endpoints. Data sources included FDA communications, clinical trial publications and protocols through 12/1/2024. Trial design details and results were extracted (e.g., cross-over, lines of therapy, and efficacy data, including OS when available). Results: The FDA AA program approved 129 drug-indications from 2013 - 2023: 18 were withdrawn, 63 were not yet converted to RA, and 48 received RA. Of the 48 RA conversions, 29 drug-indications were converted on a non-OS endpoint (e.g., progression-free survival). Six were converted with single-arm trial data. Among the 23 drug-indications with comparative data, 21 reported OS hazard ratios (HR), and 2 did not have available OS HR (one has not met the pre-specified number of events for analysis, and one did not include OS HR in statistical analysis). OS HRs ranged from 0.4 to 0.995. Five (22%) had significant OS HR in later analyses of confirmatory trials and 2 (9%) had significant OS HR in separate trials for the same indication. For drug-indications not reaching statistical significance, point estimates for OS HR were < 0.9 in 10 (43%), of which 6 had OS HR < 0.8. Eight of the 23 drug-indications had immature data where median OS was not yet reached. Of the 23 drug-indications, the majority (57%) had confirmatory trials in an earlier line of therapy, most commonly in first line. Most trials allowed cross-over (35%, n = 8) or reported patients in the control arm who received active arm or similar therapies in subsequent treatment (39%, n = 9). Conclusions: With immature data pointing to an OS benefit across trials, there was no evidence of survival disadvantage, even when statistical significance was not reached. Cross-over and confirmation in earlier lines may contribute to a lag in showing a significant OS HR, and along with the positive trends seen, highlight the danger in concluding no OS benefit at arbitrary time points. In some cases, OS HR was not feasible due to use of single-arm trials in rare diseases or the disease’s nature (e.g., chronic myeloid leukemia). Delaying approvals to wait for full OS data delays patient access to innovative therapy. The AA program balances the risk of waiting on OS data through stringent safeguards to address the unmet needs of patients unable to wait for certainty on OS benefit.