Former FDA Officials Highlight Evidentiary Challenges for Real-World Data

Share

Former FDA officials and a regulatory evidence expert say real-world evidence could play an increasingly important role in cancer drug development, particularly when randomized clinical trials are impractical, but sponsors must address significant methodological challenges before relying on such data to support regulatory decisions.

In an 11/16 FDAWatch podcast hosted by Wayne Pines, former FDA Chief Medical Officer Hilary Marston, former FDA Oncology Real-World Evidence Program leader Donna Rivera, and Analysis Group epidemiologist Mei Sheng Duh discussed the use of external control arms, combination therapy studies and innovative clinical trial designs.

Rivera said a central consideration is whether real-world data are fit for their intended regulatory purpose. Sponsors must establish that data are relevant and reliable and that study designs adequately address potential bias and confounding.

Duh emphasized that statistical methods cannot compensate for missing critical prognostic information. Although techniques such as target-trial emulation, propensity score matching and weighting can improve observational comparisons, unmeasured confounding remains a significant limitation.

The panel also discussed FDA's concerns about demonstrating the contribution of individual components in combination cancer therapies.

Duh cited the agency's complete response letter for Replimune's investigational RP1 therapy in combination with nivolumab for melanoma. She said the case illustrates the difficulty of determining whether an investigational drug adds meaningful benefit when a single-arm study also includes an established active treatment.

Rivera noted that FDA guidance addresses approaches for evaluating the contribution of individual components, including factorial study designs and, potentially, external data sources. Marston emphasized that oncology combination studies must assess not only the efficacy of individual components but also their contributions to toxicity and the overall benefit-risk profile.

The discussion also addressed the Advanced Research Projects Agency for Health's SURPASS initiative, which seeks to accelerate clinical development through innovative research infrastructure and trial methodologies.

Marston said potential approaches include adaptive designs and more efficient use of shared control data across development programs. Panelists also identified AI-assisted extraction of information from electronic health records as a potential way to improve the quality of oncology real-world data.

The participants emphasized that early communication with FDA reviewers remains important when sponsors propose unconventional evidence-generation strategies.

Read more