[How I explore ... a risk difference in the occurrence of an event in clinical trials]

Evidence-based medicine often requires the comparison of two therapeutic interventions in controlled clinical trials with the demonstration of a superiority (versus a placebo or an active comparator) or at least a non-inferiority (versus an active reference) concerning a primary endpoint that has been defined a priori (occurrence of a major clinical event, for instance). The difference in the occurrence of such an event between two treatments may be statistically analyzed by absolute risk reduction, relative risk reduction, hazard ratio or odds ratio. The present article discusses the nuances, sometimes of importance, concerning the significance of these various indices and analyses the cautions to be taken and the pitfalls to be avoided in their interpretation and use in practice. The clinician is, indeed, increasingly confronted to results of clinical trials, but is generally poorly informed regarding the nuances of these various statistical analyses.