Optimal two-stage design of single arm Phase II clinical trials based on median event time test

Abstract

The Phase II clinical trials aim to assess the therapeutic efficacy of a new drug. The therapeutic efficacy has been often quantified by response rate such as overall response rate or survival probability in the Phase II setting. However, there is a strong desire to use survival time, which is the gold standard endpoint for the confirmatory Phase III study, when investigators set the primary objective of the Phase II study and test hypotheses based on the median survivals. We propose a method for median event time test to provide the sample size calculation and decision rule of testing. The decision rule is simple and straightforward in that it compares the observed median event time to the identified threshold. Moreover, it is extended to optimal two-stage design for practice, which extends the idea of Simon’s optimal two-stage design for survival endpoint. We investigate the performance of the proposed methods through simulation studies. The proposed methods are applied to redesign a trial based on median event time for trial illustration, and practical strategies are given for application of proposed methods.

Sample size determination and evaluation for two-stage adaptive designs of single arm clinical trials based on median event time test

“…The median event time test proposed by Park [15] provides a valuable tool for study developers to determine the required sample size in order to achieve sufficient power (at least ) at a significance level of for hypothesized median survivals. This approach is appealing due to its simplicity in decision-making and the ease of understanding the drug’s efficacy based on the median event time.…”

“…Park [15] determines the decision rule of METT by using the empirical search, which requires a lot of computation time. In this work, to save the computational cost, we use the true survival distribution to determine the decision rule for the two-stage design and show that the proposed decision rule is not sensitive to censoring information, provided the median survival time is observed at interim.…”

“…The computing time was measured with Apple M1 Ultra and 128 GB memory. Using the proposed method, the decision rules in Table 1 were obtained very quickly with less than 0.2 s. However, using the numeric search proposed in Park [15] , it took several hours to obtain the decision rules in Table 2 of Park [15] , which was computationally expensive. Table 2 shows that the proposed method improves the computation to determine the decision rules for the two-stage design.…”

“…Our previous work uses the parametric survival distributions to develop the median event time test for single-arm Phase II trials whose primary endpoint is time-to-event [15] . It extends the idea of Simon’s two-stage design [16] for a time-to-event endpoint based on the median event time test.…”

“…It tests the improvement of median survival and can be easily implemented by comparing the observed median survival time for the new drug with a prespecified threshold. Park [15] determines sample size and threshold of clinical trial design based on median event time test through the numerical search for the given target type I and II error rates.…”