The Design Of Target Clinical Trials Based On Phase Ⅱ/Ⅲ Seamless Trial

Objective:The research and development of target drugs is getting more and more attention. The exact definition and effect of a target population is often ambiguous during the initial period. Although it can be identified by post-hoc analysis, an additional confirmatory trial must be undertaken to evaluate the safety and efficacy. Therefore, this may lead to an inflation in development time and cost. Our proposed design adds extra flexibility by allowing the trial to take into account the effect of both the target population and the full population as co-primary analysis data set in a phase Ⅱ/Ⅲ seamless trial, and then to evaluate the statistical characteristics. Besides, an improved understanding of biomarkers has stimulated the population selection in phase Ⅲ clinical trials. And this could potentially result in more target studies and reducing the number of subjects to recruit. In the second part, we compared the sample sizes of three different trial designs which select the population by biomarkers. It is expected to provide a framework for the calinical practice.Methods:Based on the phase Ⅱ/Ⅲ seamless design, both surrogate endpoints and primary endpoints are allowed for the interim analysis when the first stage trial is finished and then the population for the second stage trial are decided. Final analysis is conducted through combing the primary endpoints from the two stages after finishing the whole trial. Next we compare the sample sizes of three selection methods, which are based on the screening by prospective biomarker, prognostic biomarker and the short-term change of biomarkers.Results:It is shown that the type-Ⅰ-error rate is well controlled and is independent of the correlation between OS and PFS. Simulation studies also demonstrate that the probability of a right choice is high and correct conclusions are reached sufficiently often in various scenarios. In most scenarios, the sample size of the enrichment design and the active run in design are less than those of traditional RCT.Conclusions:In oncology trials, if there is a priori hypothesis on increased efficacy in a prospectively defined target population and this target population can be well characterized,our design can save the time and reduce the cost. For the comparison of sample sizes, if the short-term change of biomarkers is a reliable and efficient predictor to the therapy, the active run in design can reduce the sample size substantially. And ED is more efficient than RCT but often less than ARD.