Consider a Phase 3 group sequential trial where the primary endpoint is overall survival. Suppose that longitudinal data are observed on a biomarker which is assumed to be predictive of survival, and we believe that there could be a large benefit from using this biomarker information to inform early stopping decisions. For example, we consider a trial for the treatment of metastatic breast cancer where repeated ctDNA measurements are available.
We present a joint model for survival and longitudinal data and a method which establishes the distribution of successive estimates of parameters in the joint model across interim analyses. Then, we are equipped to use the estimates to define both efficacy and futility stopping rules.
Extending upon this idea, we create an adaptive design which incorporates subgroup selection based on this joint model. With the methodology in place, by simulation we can assess the potential benefits of including biomarker information, how this affects interim decisions and ultimately alters the trial. We demonstrate that by including the longitudinal data in the analysis, the required sample size is dramatically reduced.
|