The Effective Statistician - in association with PSI A novel approach for finding predictive biomarkers
May 12, 2025
Join Pavel Mozgunov, a group leader at the MRC Biostatistics Unit, and Julia Geronimi, a principal translational statistician at Servier, as they delve into innovative methodologies for biomarker analysis. They discuss how to effectively pre-specify analysis strategies, enhance predictive modeling, and tackle the challenges of small sample sizes. The conversation unpacks the complex distinctions between prognostic and predictive biomarkers while advocating for structured approaches to optimize clinical trial outcomes. Get ready for fresh insights into the future of biomarker-based trials!
AI Snips
Chapters
Transcript
Episode notes
Difference Between Biomarker Types
- Prognostic biomarkers predict clinical outcomes regardless of treatment.
- Predictive biomarkers identify who benefits from specific treatments based on interaction effects.
Challenges in Biomarker Assessment
- Distinguishing prognostic from predictive biomarkers requires control arm data.
- Continuous biomarkers pose challenges since dichotomizing sacrifices power and methods need innovation.
Separate Predictiveness from Cutoff
- First confirm the biomarker predictiveness signal before identifying cutoffs.
- Pre-specify analysis steps rigorously in protocols and SAPs for biomarker evaluation.
