Keywords: Immuno-oncology, Probability of Study Success, Composit endpoints, Go/no-go decision
IO drug(s), due to its mechanism of action, may demonstrate marginal benefit in Progression Free Survival (PFS) but meaningful improvement in Overall Survival (OS). To mitigate the risk, sponsors often consider PFS and OS as co-primary in phase III development. This unique feature brings challenges to constructing a go/no-go decision boundary in phase 2 study. To accelerate the development due to the highly competitive landscape, sponsors sometimes would like to trigger a phase 3 trial based on strong phase 2 interim PFS and overall response rate (ORR) results while continuing the follow-up of OS.
To accommodate this risk, we propose a Bayesian joint modeling approach that utilizes composite endpoints (e.g. PFS and OS) at early stage for phase II to evaluate PrSS of PhIII study. The model assumes the hazard ratio of PFS and OS are bivariate log-normally distributed with certain correlation. The correlation can be evaluated based on two distinctive approaches, a) retrieving from historical data, and b) conducting simulations. PrSS of phase 3 study based on composite endpoints (e.g. HR_PFS