Interim analyses of survival data in cancer clinical trials

It is common practice to undertake interim analyses of accumulating surviva l data in a cancer clinical trial while patient entry and/or follow-up is s till in progress. The purpose of making an interim analysis is to stop the trial if a convincing treatment difference can be demonstrated. The main pr oblem with repeated significance tests is that the more often one analyses accumulating data, the greater the probability of eventually obtaining a si gnificant result, p < 0.05 say, even when there is in reality no treatment difference. To allow for repeated testing one must therefore choose a more stringent nominal significance level as a criterion for stopping the trial. Repeated significance tests are normally applied at equally spaced interva ls, and the maximum number of teals is decided in advance. This paper prese nts and compares the properties of two simple and commonly applied strategi es for undertaking interim analyses of accumulating survival data.