Impact of pain definition on the analysis of clinical trials

Several clinical trials in herpes tester assessing the efficacy of antivira l agents on pain resolution have been published. Definition of pain vary fr om one trial to another and statistical analysis is often biased or inappro priate. Time between randomisation and cessation of pain should be the majo r outcome measure. Comparison between treatment groups should be made by us ing a logrank test or a Cox's proportional hazards model. The analysis shou ld be performed on all patients included in the trial, If the difference be tween treatment groups is significant, secondary separate analyses could be performed on different periods of time, for example, for the first three m onths and for the following months, The periods of time in the secondary an alyses should be determined a priori and the sample size has to be large en ough to ensure an adequate power in the analysis of the later period.