Presentation (CTI)

OBJECTIVES: The Food & Drug Administration Accelerated Approval (AA) Program approves drugs based on surrogate endpoints. Some oncology drugs may not show statistically significant overall survival (OS) benefit when converted to Regular Approval (RA). This lack of statistical significance is often considered to be a failure to demonstrate survival benefit. Bayesian frameworks may guide clinical decision-making under uncertainty by providing the probability of an OS benefit.
METHODS: Oncology AA drug-indications converted to an RA between 2013 and 2023 on a non-OS endpoint were evaluated. Current OS data was extracted through 12/1/2024. The Bayesian analysis assumed normally distributed hazard ratios (HRs), non-informative priors, and HR thresholds of 0.8 and 1.0 to analyze OS.
RESULTS: There were 21 drug-indications converted to RA on a non-OS endpoint that were analyzed and had available OS HRs. OS HRs ranged from 0.4 to 0.995. The probability of at least a 20% lower risk of death in the active arm ranged from 2% to 99%. Of the 21 drug-indications, 11 had a 65% or greater probability of at least 20% lower risk of death and 10 had a 50% or lower probability of at least 20% lower risk of death A majority (18) of drug-indications had a 65% or greater probability of OS HR < 1.There were no cases of 50% or higher probability of OS HR > 1.
CONCLUSIONS: This analysis finds that whilst 5 drug-indications showed a statistically significant survival benefit by frequentists methods, our Bayesian analysis found most had > 65% probability of the true hazard ratio indicating benefit, which provides confidence in the AA system and important information for shared decision-making. Future work can include AA drug-indications pending RA and incorporate median survival differences to evaluate clinical benefit. Clinicians may use a Bayesian approach to discuss probability of OS benefit with patients.