Tom Coates on AI For Pharma Growth: why interim analysis is due an upgrade
Our CEO, Tom Coates, joined Dr Andree Bates on AI For Pharma Growth to talk about interim analysis, why so few trials use it well, and what it would take to catch futility and success years sooner.
The episode starts from an uncomfortable fact: a great many commercial phase 2 and phase 3 trials still carry no pre-planned interim analysis at all. The consequence is that sponsors wait far longer than they need to before finding out a study is failing and are slower to act when it is succeeding. Tom's argument is that this is not only a limitation of the statistics. It is a gap between what is methodologically possible and what is the default operationally.
Much of the conversation is about what has to be true before a sponsor, a data monitoring committee or a regulator will act on an early readout. A high performing prediction algorithm is necessary but not sufficient. A deeper requirement is an architecture that can handle unblinded data without compromising trial integrity, audit trails that stand up to inspection, and methods a DMC can actually interrogate. Tom and Andree also get into where AI belongs in this picture and where it does not; for a critical stop-or-go decision, explainability and reproducibility matter considerably more than a potentially modest increase in speed, efficiency or other characteristics. AI model -based methods, synthetic data and subgroup analysis have real value in understanding which patients benefit and designing better trials around that.
Topics covered include:
- Why interim analysis is underused
- Stopping early for futility or success
- Protecting blinding and trial integrity
- Secure handling of unblinded data
- What data monitoring committees need to see
- Where AI fits, and where it does not
- Subgrouping and individual treatment effects
- Synthetic data and trial simulation
- Regulatory confidence and audit trails
- The future of continuous trial monitoring
The episode runs 28 minutes.
Listen to episode 232, "The Early Readout: Upgrading the Interim Analysis to Catch Futility and Success Years Sooner", on Spotify.
AI For Pharma Growth is Dr Andree Bates' podcast on how AI-based technologies are used across pharma, biotech and healthcare.