Subgroups, treatment effects, and baseline risks: Some lessons from major cardiovascular trials

Background The objective of this study was to determine how subgroup analys es are performed in large randomized trials of cardiovascular pharmacothera py. Methods and Results We reviewed 67 randomized, double-blind, controlled tri als involving pharmacotherapy in at least 1000 patients with unstable angin a, myocardial infarction, left ventricular dysfunction, or heart failure wi th clinical outcomes as primary end points, published between 1980 and 1997 . Nine had no subgroup analyses but 43 reported on 5 or more subgroups and 31 reported subgroups without formal statistical tests for treatment-subgro up interactions. In most trials, a rationale for subgroup selection was mis sing. All but 6 focused on single-factor subgroups. Conclusions Trial subgroups should ideally be defined a priori on 2 bases: single-factor subgroups with a strong rationale for biological response mod ification and multifactorial prognostic subgroups defined from baseline ris ks. However, single-factor subgroup analyses are often reported without a s upporting rationale or formal statistical tests for interactions. We sugges t that clinicians should interpret published subgroup-specific variations i n treatment effects skeptically unless there is a prespecified rationale an d a significant treatment-subgroup interaction.