ICH E17 Guideline ate statistical model. Covariate adjusted models may be especially important in this set- ting as a way to account for intrinsic and extrinsic factors. Methods using weighted av- erages of the overall effect estimate and the estimate using data from individual regions (shrinkage estimates) may be considered, particularly when regional sample sizes are small and outlying values may be overly influential. The choice of model should reflect an understanding of how intrinsic/extrinsic factors affect the regional estimates and be based on appropriate statistical methods. Sensitivity analyses should be planned that vary the assumptions required for the model.
Impact of Trial Quality on Analysis Differences in trial conduct across regions can negatively impact the power to detect an overall treatment effect as well as the ability to examine consistency of treatment effects at the analysis stage. Important factors impacting the quality of the trial, such as follow- up of study subjects, should be managed consistently across regions, and issues identified during the trial corrected as early as possible. During the conduct of an MRCT, trial monitoring and blinded data review may uncover various issues that require modifications to be made to the analysis plan for the trial.
For example, it might be necessary in an MRCT for better assessment to modify pooling strat- egies for regions or subpopulations, that were carefully defined during trial planning, after sufficient data have been accumulated on the baseline characteristics (e.g., intrinsic and extrinsic factors) of the multi-regional population. However, such changes should be justified, discussed with the relevant regulatory authorities, and carried out in a way that preserves trial integrity.
2.2.8 Selection of Comparators The choice of control groups should be considered in the context of the available standard therapies, the adequacy of the evidence to support the chosen design, and ethical consid- erations. The selection of comparators should be discussed and agreed with the relevant regulatory authorities. Comparators in MRCTs should in principle be the same in all participating regions. Due to the complexity in setting up MRCTs, some key points are addressed in the following paragraphs, focusing on practical and ethical issues associated with the use of comparators: 23