Aside from statistical power considerations, we would like to comment on other methodologic aspects that deserve attention. Indeed, the use of hospitalization rather than hospitalization or death would result in very similar power because of the much higher hospitalization rates compared with mortality rates. In addition, the use of composite end points remains a matter of controversy. For instance, Montori et al2 suggested certain criteria required for their appropriate use, including the clinical relevance of individual components, the frequency of each component, as well as treatment effects on each component. Hence, why not use hospitalizations rather than composite end point, which would result in very similar power and would facilitate conclusions regarding treatment effect?