Sponsors should have a pre-specified plan to rule out adverse impacts on overall survival in every potential registrational study in oncology, the US Food and Drug Administration said in a new draft guidance.
“If the trial is intended to support a regulatory submission, the SAP [statistical analysis plan] should specify that at the time of the primary analysis of the primary endpoint, if the primary endpoint is not overall survival, an analysis of overall survival will be conducted to assess the potential for harm,” the FDA said in an Aug. 18 draft guidance, “Approaches to Assessment of Overall Survival in Oncology Clinical Trials.”
The draft builds on a 2023 discussion of the role of OS analyses in oncology trials when using survival as a primary or secondary efficacy endpoint may not be feasible that was held as part of a workshop sponsored by the FDA, American Association for Cancer Research and the American Statistical Association.
The draft guidance does not set explicit thresholds for harm as the agency once did for diabetes drug developers and cardiovascular outcomes. Instead, the FDA states the plan to rule out harm will depend heavily on the specific context for the investigational therapy.
“When overall survival is not an alpha-controlled primary or secondary efficacy endpoint with a formal testing plan in a randomized oncology clinical trial, sponsors should include a pre-specified plan to assess overall survival, as a safety endpoint, with an aim to evaluate for potential harm due to the therapeutic intervention,” the agency said in the guidance.
“To adequately inform regulatory decision-making, any trial should be designed a priori to adequately capture the number of events needed to rule out harm based on a specified threshold(s),” the FDA wrote. “Sponsors should specify and justify in the SAP a threshold or a range of thresholds to indicate the potential for harm for the overall survival summary measure.”
To set a threshold for harm and the “proposed degree of precision to rule out harm,” the FDA advised sponsors to consider many factors, including “disease setting, known safety profile of the product or drug class, input from patients and physicians, expected number of events, length of expected survival, duration and severity of expected adverse events, use of subsequent therapies or crossover, potential for non-proportional hazards, evidence of benefit from other endpoints, expected benefit of the control, acceptable level of uncertainty, feasibility of follow-up, and other available data.”
In some cases, the overall survival data will be immature or extremely limited, leading to a high degree of uncertainty about any interpretation, the FDA states in the guidance. The agency makes clear that it is not comfortable ignoring OS data in those cases, but will be open to creative approaches to modeling potential outcomes.
“Overall survival data may not be mature at the time of an early interim analysis, but an assessment of the available overall survival data is important in the determination of the benefit-risk profile,” the FDA wrote. “Sponsors should contextualize the available overall survival information and feasibility of obtaining additional overall survival data to inform regulatory decision-making.”
“If the sponsor anticipates that overall survival data will be immature or there will be high uncertainty in the overall survival summary measures at the time of the proposed final analysis, sponsors should conduct additional calculations or simulations to assess if it is likely that harm may be ruled out with additional follow-up time, which may or may not be feasible,” the FDA said.
“Various methodologies can be used to calculate the probability of ruling out harm based on observing hypothetical future data,” the agency wrote in the guidance. “Sponsors should specify in the SAP approaches to rule out harm under a variety of scenarios, justification of methods, and justification of assumptions, including the assumption of observing additional events over a specified time period. Sponsors should include a variety of assumed scenarios for future data.”
“However, the uncertainty in these methods generally increases for longer looks into the future and evaluation of harm cannot solely be based on hypothetical future data,” the FDA added.
The Friends of Cancer Research has been spearheading work to develop modeling and simulation approaches to provide more rigorous safety assessments in the context of extremely limited OS data. An FOCR working group addressed the topic in 2024.
FOCR also announced Aug. 19 that it was collaborating with clinical research organization MMS to further advance the project.