The Effective Statistician - in association with PSI Top 5: The analysis of adverse events done right
Aug 25, 2025
Kaspar Rufibach, an expert in survival analysis and member of the SAVVY collaboration, and Jan Beyersmann, a professor of biostatistics at Ulm University, discuss the complexities of analyzing adverse events in clinical trials. They highlight how varying follow-up times can bias results and advocate for using the Aalen–Johansen estimator as a standard practice. The conversation emphasizes the successful collaboration between pharma and academia, revealing how real-world data can change the perception of treatment risks and enhance benefit-risk evaluations.
AI Snips
Chapters
Transcript
Episode notes
Follow-Up Time Distorts AE Counts
- Counting adverse events as simple proportions gives misleading comparisons when follow-up differs between arms.
- Longer survival inflates raw AE counts even if true AE rate per time is unchanged.
Safety And Efficacy Methods Are Inconsistent
- Safety and efficacy analyses often use different mindsets despite studying the same trial.
- That mismatch can create absurd conclusions, like a drug appearing less safe solely because it prolongs survival.
Define Your Safety Objective First
- Decide whether you want rough signal detection or accurate AE probability estimates before choosing methods.
- For precise probability estimates, define endpoints, estimands and adjust data collection accordingly.
