The Effective Statistician - in association with PSI Introduction to adaptive designs and ICH E20
10 snips
Oct 20, 2025 In this discussion, Kaspar Rufibach, an experienced biostatistician and adaptive clinical trial expert, delves into the practicalities of adaptive designs and the ICH E20 guideline. He explains why these designs save time and the types of meaningful adaptations available for drug development. Kaspar also addresses the operational barriers preventing wider use of adaptive trials and emphasizes the importance of understanding their implications on endpoints and inference. Lastly, he outlines the key points of the new ICH E20 guideline that statisticians should note.
AI Snips
Chapters
Transcript
Episode notes
Group-Sequential Is Adaptive By Design
- Group-sequential designs let you look early and preserve type I error by reallocating alpha across looks.
- They are conceptually adaptive because you can stop early or continue based on interim results.
Adaptations Create Multiplicity Risks
- Mid-trial observations can justify enriching populations or dropping arms, but these changes introduce multiplicity concerns.
- Advanced statistical methods were developed to maintain type I error after such adaptations.
Preplan Adaptations And Adjust Inference
- Pre-specify adaptations like arm dropping, enrichment, and sample-size reassessment at design stage to keep inference valid.
- Do not compare unadjusted post-adaptation data as if no interim decisions occurred.
