Correcting for multiple-testing in multi-arm trials: is it necessary and is it done?

Abstract

BackgroundMulti-arm trials enable the evaluation of multiple treatments within a single trial. They provide a way of substantially increasing the efficiency of the clinical development process. However, since multi-arm trials test multiple hypotheses, some regulators require that a statistical correction be made to control the chance of making a type-1 error (false-positive). Several conflicting viewpoints are expressed in the literature regarding the circumstances in which a multiple-testing correction should be used. In this article we discuss these conflicting viewpoints and review the frequency with which correction methods are currently used in practice.MethodsWe identified all multi-arm clinical trials published in 2012 by four major medical journals. Summary data on several aspects of the trial design were extracted, including whether the trial was exploratory or confirmatory, whether a multiple-testing correction was applied and, if one was used, what type it was.ResultsWe found that almost half (49%) of published multi-arm trials report using a multiple-testing correction. The percentage that corrected was higher for trials in which the experimental arms included multiple doses or regimens of the same treatments (67%). The percentage that corrected was higher in exploratory than confirmatory trials, although this is explained by a greater proportion of exploratory trials testing multiple doses and regimens of the same treatment.ConclusionsA sizeable proportion of published multi-arm trials do not correct for multiple-testing. Clearer guidance about whether multiple-testing correction is needed for multi-arm trials that test separate treatments against a common control group is required.Electronic supplementary materialThe online version of this article (doi:10.1186/1745-6215-15-364) contains supplementary material, which is available to authorized users.