PS33: Statistical Review and Inspection of Real-world Evidence in Regulatory Submissions: Challenges, Risks, and Emerging Solutions

As regulatory agencies increasingly rely on real-world evidence (RWE) to support drug approvals, label expansions, and postmarketing decisions, the statistical review and inspection of regulatory submissions based on real-world data (RWD) present challenges that differ fundamentally from those encountered with traditional randomized clinical trials. While well-established statistical and operational frameworks exist for reviewing trial data collected under controlled protocols, RWD-derived study data introduce additional sources of uncertainty that directly affect bias, estimand derivation, reproducibility, and inferential validity. This panel will examine the practical and statistical challenges that arise when reviewing and inspecting regulatory submissions derived from RWD sources such as electronic health records and administrative claims. Panelists will discuss how heterogeneity across data sources, complex data transformations, outcome and exposure misclassification, and limited traceability to original source data complicate the assessment of statistical assumptions underlying RWE analyses. These issues raise critical questions for reviewers, including how study variables and estimands can be transparently derived from source data, whether those derivations can be independently verified, and how errors introduced during data curation and transformation propagate to effect estimates and measures of precision. Drawing contrasts with the statistical review and inspection of randomized trials, the panel will explore emerging approaches to improve the inspectability and reliability of RWE analyses. Topics will include transparent and independently verifiable derivation of study estimands from real-world data, sensitivity analyses, and quantitative bias analysis to assess the impact of misclassification and other sources of bias. The panel will also discuss ongoing work with FDA and multiple pharma on establishing approaches and infrastructure to support the transformation of source RWD into reliable study data in data standards such as CDISC and FHIR and the review and inspection of RWD in FDA submissions. Perspectives from regulatory statisticians, industry statisticians, and methodological experts will highlight where existing guidance is insufficient and where new statistical and operational best practices are needed. Attendees will leave with a clearer understanding of how statistical review and inspection of RWE differ from traditional trial review, what statistical risks are most salient in RWD-based regulatory submissions, and how RWE analyses can be designed to be more transparent, reproducible, and robust under regulatory scrutiny.