A simulated sequential analysis based on data from two MRC trials.

Abstract

The motivation for proposing sequential methods for cancer clinical trials is presented, and the methodology examined by re-analysing two completed phase III cancer trials of the Lung Cancer Working Party of the British Medical Research Council. The reanalysis proceeds as if the trials had been designed with a planned series of interim analyses governing stopping. Specifically, the triangular and double-triangular tests were applied. The sequential reanalysis gave a substantial reduction in the number of patient required, and deaths observed, for conclusions to be reached in comparison with the completed studies. In each case, the sequential analysis was stratified for baseline prognostic factors which were seen to be important at the first interim analysis.