ICH E17 Guideline further post-hoc investigation to either identify plausible reasons for the differ- ences or to better understand the observed heterogeneity. In some cases, addi- tional data, including data from other clinical trials, or supportive evidence from other sources, may be needed to understand the regional differences observed. These eventualities should be carefully considered at the planning stage.
Planned Subgroup Analyses In addition to analyses intended to investigate any regional or pooled subpopulation dif- ferences in treatment effects that may be observed in an MRCT, other subgroup analyses will usually also be of interest, just as they are for any clinical trial (e.g., analyses to investigate differential treatment effects by sex and age) and should be planned. If sub- group differences in overall treatment effects are observed, then exploring whether the subgroup differences are consistent across regions or pooled regions may be informative. In summary, the assessment of consistency of treatment effects should be done with dili- gence to inform regulatory decision-making.
The credibility of regional, pooled subpop- ulation, or subgroup findings should take into consideration biological plausibility, inter- nal consistency (e.g., similar patterns of regional variability observed for other secondary endpoints) and/or external consistency (e.g., similar patterns observed in another clinical trial of the same drug), the strength of evidence, the clinical relevance, and the statistical uncertainty. The more of the aforementioned considerations support a potential finding, the greater the likelihood the finding is not false. The evaluation of consistency will ben- efit from joint clinical and statistical perspectives.
Estimation of Regional Treatment Effects The statistical analysis section of the protocol should describe appropriate statistical methods for estimating and reporting treatment effects and measures of their uncertainty for individual regions. This plan should include a determination of the adequacy of sam- ple sizes to support robust estimation within each region or pooled region for which re- porting of treatment effect is of interest.
If the sample size in a region is so small that the estimates of effect will likely be unreliable, the use of other methods should be consid- ered, including the search for options for additional pooling of regions based on common- alities, or borrowing information from other regions or pooled regions using an appropri- 22