Two-stage screened selection designs for randomized phase II trials with time-to-event endpoints

Biom J. 2022 Oct;64(7):1207-1218. doi: 10.1002/bimj.202100305. Epub 2022 Jun 5.


Phase II exploratory multiarm studies that randomize among new treatments are found to be broadly useful and appear to be of value both scientifically and logistically, especially in the areas of unmet needs, for example, pediatric cancer. This multiarm design also has a faster recruitment rate because it provides patients with more treatment choices than traditional two-arm randomized controlled trials do. In contrast to direct formal comparisons in multiarm multistage designs, for example, umbrella or platform designs, the screened selection design (SSD) recommends using a promising treatment arm by ranking according to the effect size, which often needs lesser sample sizes than the former. In this paper, the usefulness of the phase II SSD design is exemplified by three real trials. However, the existing SSD methods can only deal with binary endpoints. Motivated by the real trials in the authors' respective institutions, we propose using the two-stage SSD and its variant for randomized phase II trials with the time-to-event endpoint. The proposed methods not only provide a high probability of selecting a superior treatment arm but also control the type I error rate for testing the efficacy of each treatment arm versus a common external control. Sample size calculations have been derived and simulation studies demonstrate desirable operating characteristics. The proposed design has been used for designing three real trials. An R package frequentistSSD has been developed and is freely accessible for practitioners.

Keywords: one-sample log-rank test; randomized phase II trial; screened selection design; time-to-event.

Publication types

  • Clinical Trial, Phase II
  • Research Support, N.I.H., Extramural
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Child
  • Computer Simulation
  • Humans
  • Neoplasms* / drug therapy
  • Randomized Controlled Trials as Topic
  • Research Design*
  • Sample Size