[Analyzing clinical trials: subgroup, stratified, and intermediate analyses, and composite outcome criteria]

Subgroup analyses examine the effect of a treatment regimen on different subgroups in the overall population. If the subgroups are not comparable, differences between them cannot be attributed to the treatment alone, and analyses can only be considered as exploratory. When, however, strata are defined before randomization and subjects are randomly allocated to treatment groups within each stratum, analysis can be conducted within these subgroups and differences between them can be shown to be caused by the treatment. Intermediate analyses, by multiplying the number of statistical tests, increase the risk of rejecting the null hypothesis when it is in fact true. Accordingly, specific statistical methods should be used to compensate for this risk. Composite outcome criteria combine several different elements, all of which must be clinically relevant. Unequal treatments effects on the various criteria may complicate interpretation of the results.