Back to: Statistical Considerations for Premarketing Risk Assessment
🔔 Note: Stop playback at 1:07:43 to complete this lesson.
Presenter
Mat Soukup, Ph.D.
Deputy Director
Division of Biometrics VII
Office of Biostatistics (OB)
Office of Translational Science (OTS)
Center for Drug Evaluation and Research (CDER)
US Food and Drug Administration (FDA)
Abstract
This presentation focuses on Statistical Considerations for Premarketing Risk Assessment, specifically addressing how to analyze adverse event data from clinical trials. The analysis of safety outcomes centers on estimating risks and uncertainty for comparing two or more treatments, rather than solely relying on hypothesis testing. The evaluation on estimating risks and uncertainty for comparing two or more treatments, rather than solely relying on hypothesis testing. The evaluation must involve identifying an appropriate summary measure of risk (e.g., cumulative incidence or incidence rate) and ensuring the statistical methods align with the trial design. Crucially, the presentation illustrates the potential for bias and confounding when using crude proportions from on-treatment analyses and emphasizes the importance of conducting on-study analyses which include all events regardless of treatment discontinuation, preserving the integrity of randomization. The session highlights that relative metrics (ratios) and absolute difference metrics (risk difference) are both essential for comparative assessments, with absolute differences being most meaningful for evaluating public health impact and benefit-risk assessments. Furthermore, for integrated analyses based on multiple trials, it is critical to account for trial and stratify analyses to avoid misinterpretation due to confounding, such as Simpson’s paradox, and ensure reliable risk information. Finally, the presentation underscores that reliable safety assessment requires collaboration among statisticians, clinicians, and data scientists to interpret statistically appropriate analysis results.