Practical side of multistage clinical trials for screening new agents.

Coordinating center operations are recommended for carrying out multistage phase II clinical trials in oncology. The relevant statistical issues in designing phase II trials are summarized. The predictive probability approach to multistage design is briefly described, along with an interactive computer program that facilitates design selection. A multistage trial monitoring form is described, along with data manager-biostatistician responsibilities for monitoring the trial for possible early termination. Procedures are recommended for dealing with patients who cannot be evaluated by response, for minimizing or eliminating selection bias introduced by clinician knowledge of interim results, and for minimizing clinician tendency to report responding patients sooner than those not responding.