Charts for the early stopping of pilot studies

Abstract

Cooperative oncology groups usually run pilot studies of new agents or combinations concurrently with their major randomized clinical trials. A primary objective of these studies is to determine whether the new regimen should be tested further in a group-wide clinical trial. The accrual goals of such pilot studies are typically fixed in advance at between 30 and 40 patients, on the grounds that this number provides a reasonably tight confidence interval on the true response rate. Nevertheless early termination of pilot studies is often desirable either because the regimen appears inactive or because early results indicate extreme activity and justify immediate testing in a randomized study. Statistical charts are provided for early termination in both these situations. The charts are read by specifying the number of evaluable patients already accrued, the number of responses observed and the minimum true response rate, theta 0, at which the regimen would be considered active. The charts provide the posterior probability that the true response rate exceeds theta 0, that is, that the regimen is active. An additional chart that computes a 90% probability interval for the true response rate, based on the observed rate and sample size, is also provided. The use of the chart is illustrated with two examples from the Eastern Cooperative Oncology Group.