The FDA’s Shifting Stance on Cancer Clinical Trials

Last month, the US Food and Drug Administration issued a draft guidance that could dramatically reshape how we conduct cancer clinical trials. Will preclinical drug developers need to adapt, too? 

A report published last year in the Journal of Clinical Oncology found that one in five people with cancer in the US participated in a clinical trial, a much higher number than experts expected¹. Data from these studies help to inform the future use of hundreds of cancer drugs, so it was illuminating when the US Food and Drug Administration announced in August 2025 that it was taking a closer look at how drugs are assessed, and whether the current endpoints in clinical trials should be more stringent. The guidance, titled “Approaches to Assessment of Overall Survival in Oncology Clinical Trials" is seeking public input (the deadline for responses is October 2025). While the guideline’s current scope and recommendations are squarely on the clinical trials conducted in humans, it could ripple back into preclinical drug development, too, particularly concerning animal study designs and New Approach Methodologies (NAMs) discussions.

What are some of the key points in the current FDA draft? 

The guidance recommends prospective study sponsors use overall survival (OS) in randomized oncology clinical trials, especially as a prespecified safety endpoint. Essentially, OS as a primary endpoint should be prioritized over progression-free survival when feasible, given its objectivity and clinical relevance. Both overall survival and progression-free survival help assessing how effective a treatment is, but they measure different outcomes. While overall survival looks at the length of time from treatment start or randomization until death from any cause, progression-free survival is a surrogate endpoint that measures the length of time during and after treatment that a patient lives without the disease worsening. When OS cannot be a primary endpoint, trials should still be designed for collecting and assessing OS data with fit-for-purpose analysis plans prespecified to evaluate potential harm, even if not statistically powered for efficacy².

However, factors such as crossover, subsequent therapies, and unequal randomization can complicate OS interpretation and must be carefully considered and the importance of adequate follow-up duration, tailored by disease context and survival expectations, is stressed. Prespecified analyses need to be done as well as cautions around post-hoc evaluations, subgroup analyses, and integrating early or limited OS data into benefit–risk assessments.

How could this have implications for preclinical animal studies and NAMs? 

While the current guidance is tailored toward clinical trial design, it may have important downstream effects for preclinical development phases, particularly in how translational modeling, animal studies, and advanced NAMs are conducted and interpreted. Key implications are summarized below:

Currently, animal model studies in oncology frequently rely on endpoints such as tumor volume reduction, time to progression, or surrogate markers, not typically OS. New Approach Methodologies or NAMs, such as cell-based systems, organoids, microphysiological systems (MPS), or computational models, also rarely include survival analyses. With the FDA underscoring the primacy of OS for safety and benefit–risk evaluations, there may be increased demand for preclinical models that better approximate or predict survival-related outcomes. This could shift the focus to greater development and validation of animal models designed to track survival endpoints rather than merely tumor burden.

Likewise, NAMs that simulate longer-term effects relevant to survival (e.g. models of metastatic progression, resistance development, or late toxicities) would be prioritized in development. The guidance highlights how crossover and subsequent therapies can cloud OS interpretation in clinical trials. For preclinical phases, NAMs would need to be validated so that they could ideally model adaptive treatment responses, allowing mechanistic insights into how resistance or combination sequences might influence survival, particularly for translating clinical scenarios where patients receive multiple lines of therapy.

With OS being both an efficacy and safety metric to recognize therapy-driven toxicities that can shorten survival, preclinical studies in turn may need to aggregate toxicological and efficacy readouts. NAMs could support this by incorporating multi-endpoint assessments, for example, by combining indicators of organ-level toxicities with measures of tumor cell kinetics or immune interactions. The emphasis on adequate follow-up and data maturity for OS in clinical trials suggests that regulators may scrutinize the durational adequacy of preclinical studies. These may need to simulate chronic exposure, late-emerging effects, or long-term viability to better inform risk–benefit prior to clinical translation.

The draft guidance advocates for prespecified harm measures, analytical rigor, and consideration of complex OS-related statistical challenges. In preclinical validation of NAMs, there may be growing expectations for a) setting analytical frameworks, for example by defining thresholds for detrimental survival signals in models and b) for providing quantitative translation strategies, such as linking NAM-derived endpoints to survival predictions using joint models or multistate approaches (e.g. joint modeling of biomarker kinetics with survival).

The centrality of overall survival for benefit-risk evaluation in oncology trials, either as a safety metric or an efficacy signal, is an evolving area for those developing alternative preclinical models. These systems and models not only must show tumor efficacy but also yield predictive survival relevance and safety insight.

References:

1.    Unger et al. National estimates of the participation of patients with cancer in clinical research studies based on commission on cancer accreditation data. JOC, 2024, 42:2139-2148. https://doi.org/10.1200/JCO.23.01030
2.    Rodrigues et al. CCR Perspectives in regulatory science and policy: Improving collection and analysis of overall survival data. Clin Cancer Res, 2024, 30: 3974-3982. https://doi.org/10.1158/1078-0432.CCR-24-0919.