The Cancer Drug Development Forum (CDDF) held a live webinar on Bayesian Approaches in Drug Development on 11 July 2024. In this webinar, Rob Hemmings (Consilium) gave an insightful talk followed by a Q&A discussion moderated by Axel Glasmacher (CDDF) and Eva Skovlund (CDDF). Here are key takeaways from the webinar:
- Bayesian statistics differs to Frequentist statistics by making probability statements about drug effects, rather than about data.
- Current evidence for a question of interest (e.g., Probability that drug effect > 0) is synthesised into a “Prior” distribution, to which data from a prospective experiment (e.g., another RCT) can be added to generate an updated “Posterior” distribution for the question of interest.
- Either Bayesian or Frequentist approaches can be used in principle for many applications in drug development. Bayesian statistics are already in widespread use in clinical drug development, though rarely as primary analysis for demonstration of efficacy.
- “Informative” prior distributions can reflect existing, favourable evidence that a drug effect exists with the consequent potential to inform the extent of evidence that needs to be generated in order to meet an evidentiary standard (e.g., for regulatory approval). This might be attractive for various use cases (e.g., extending use of a product to a new), but related target population or extrapolation to paediatrics.
- Operational issues exist to define evidentiary standards in the Bayesian framework, and to agree on methodological and procedural aspects relating to derivation of the Prior distribution.
- Regulators including FDA are committed to generating methodological guidance to inform use of Bayesian approaches.
🔶To watch a key takeaways video, click the link: https://youtu.be/zU72SXxtQ3U
🔶To watch a full recording of the webinar, click the link: https://youtu.be/81ZQsLaIv6Y