Bayesian Methods for Sample Size Re-estimation

This project will develop and evaluate new Bayesian approaches for re-estimating the sample size needed for a clinical trial as it progresses.

There are always assumptions made to determine the sample size of a trial, such as the variability or size of treatment effect. These assumptions are used to decide the number of participants needed for a specified probability of success (PoS) for the definitive trial. PoS is used directly by pharmaceutical companies to make optimal decisions. When design assumptions are incorrect then the sample size may not be sufficient. Sample size reestimation approaches introduce an interim analysis to assess assumptions and alter the sample size using actual trial data.

Bayesian Sample Size Reestimation (SSR) can allow incorporating other sources of data, in conjunction with the trial data, to improve the re-estimation.

Areas of particular interest are:
1) Can Bayesian SSR handle blinded reestimation approaches (i.e. where arm assignment is not used)?
2) Would a Bayesian SSR approach improve the probability of a successful phase 3 study as
compared to frequentist approaches?
3) How do methods extend to other endpoint types like binary and time-to-event?
4) How can we adapt approaches to consider more complex designs such as platform, umbrella
and basket?