To lump or to split? Statistical evidence of subgroup findings in confirmatory clinical trials and the regulatory perspective.
Randomised controlled trials (RCTs) are considered the gold standard for the identification and estimation of causal effects and thus for establishing causal conclusions. Most RCTs are designed based on the implicit assumption that patients recruited to the trial will be homogeneous and will respond (relatively) homogeneously to the intervention. As subgroups may well be associated with different treatment effects within a clinical trial, even with strict inclusion criteria, this assumption of homogeneous treatment effect should be challenged on a regular basis. To do so, subgroups should be investigated to validate the interpretation, hence the internal consistency, of the overall treatment effect. However, investigation of subgroups suffers from methodological, statistical and interpretational complications, reasons why subgroup analyses constitute a major challenge in the assessment of results from clinical trials. In this thesis, a new framework for regulatory decision making based on four purposes for subgroup analyses has been proposed: (1) Investigate the consistency of treatment effects across subgroups of clinical importance, (2) Explore the treatment effect across different subgroups within an overall non-significant trial, (3) Evaluate safety profiles limited to one or a few subgroup(s), (4) Establish efficacy in the targeted subgroup(s) when included in a confirmatory testing strategy of a single trial. These four categories, deemed relevant for practice, share common aspects but also require specific considerations to improve regulatory decision making. First a literature review was conducted to get insight into existing methods as well as into which areas research is needed as regard to the proposed ”purpose-based” framework. Secondly, exploring subgroups in trials for which the primary outcome showed no significant treatment effect in the full population (purpose 2) is very likely to cause problems as in principle no firm confirmatory conclusions can be made regarding subgroups. The evidence based on such findings is generally weak. The best available solution is therefore to conduct an appropriately designed replication trial. Finally, the impact of subgroups from a regulatory decision-making perspective was evaluated. By assessing a) whether subgroups evaluations differ between approved and non-approved products, b) whether subgroup assessment led to the approved indication being different from the indication as proposed by the applicant, and c) whether the impact of subgroups depends on an orphan status of the medicine, it appears that subgroup analyses are an integral part of the analyses the applicant presents and the analyses the regulators ask for during assessment of a marketing authorisation application. Therefore, subgroup analyses are essential in terms of influencing the medicine licensing, labelling, reimbursement, and prescribing decisions. In general, subgroup analyses constitute a fundamental step in the assessment of evidence from confirmatory (Phase III) clinical trials. This thesis emphasises that the context in which subgroup analyses are performed is of prime importance to have an appropriate and thorough clinical trial results assessment.