

Ep. 99: A Conversation with Cremieux - How to Tell Good Science from Junk, and What’s Next in US Biotech & Deregulation
7 snips Sep 23, 2025
Cremieux sheds light on the common pitfalls in scientific research, revealing how published effect sizes often inflate reality. His insights reveal gene therapy's major hurdle: it's all about delivery. A discussion on regulatory challenges shows how policies can stifle innovation, and he argues for necessary reforms in trials. He highlights the surprisingly undervalued potential of siRNA in biotech, while unpacking the implications of staggering vaccine hesitancy. Prepare to rethink what you know about science and biotechnology!
AI Snips
Chapters
Books
Transcript
Episode notes
How A Debate Sparked A Scientific Career
- As a teenager Cremieux researched the death penalty to win an argument and learned statistics deeply.
- That project launched his habit of dissecting papers and led to his public science commentary.
P-Value Clustering Flags Bad Studies
- Clusters of p-values just below significance usually indicate p-hacking and poor replicability.
- Cremieux dismisses papers with bundled p≈0.05 results unless the hypothesis is clear and pre-registered.
Beware Deep Subgroup Fishing
- Unchecked subgroup interactions often produce spurious findings when authors slice samples repeatedly.
- Cremieux tests omitted interaction checks and often finds no real significant difference versus the full sample.