Keywords: subgroup, regulatory, extrapolation, interaction
The aim of subgroup analysis is often to identify groups of patients where the efficacy is different from that presented for the whole population. Subgroup analysis represents a major statistical challenge and it is very hard to identify in a single clinical trial what is a true difference in effects as opposed to a false positive finding, because of the number of potential subgroups available for analysis and the large potential for effects to differ among levels of a covariate by chance.
Sponsors often assume consistent effects of treatment across the trial population and seek to explain any different effects in subgroups as due to chance. On the other hand, regulators may expect effects to vary across subgroups and can ask the sponsor to establish that the risk-benefit is acceptable for each subgroup. However, consistency of effect is difficult to define. Interaction tests are of limited value and any requirement for each subgroup to show a given level of effect is problematic. Modelling approaches can provide additional insight over simple separate subgroup analysis, particularly for variables measured on a continuous scale.
The need for subgroup analysis is related to the diversity of the overall patient population enrolled as defined by the inclusion/exclusion criteria. The more homogeneous the population studied, the fewer requirements there should be for subgroup analyses. In order to limit the extent of post-hoc subgroup analysis, there is a need for increased focus and discussion on subgroups at the design stage. Pre-agreement with regulatory authorities on important subgroups may be helpful.
The talk will be illustrated with two examples of different aspects of subgroup analysis. The first example will discuss modelling analysis to investigate the relationship between baseline values of a biomarker and efficacy. The second will discuss a Bayesian partial extrapolation approach to assessing efficacy in an adolescent population based on combining adult and adolescent data.
Disclaimer: The views expressed above are personal and do not necessarily represent those of GlaxoSmithKline or of the Pharmaceutical Industry in general.