Sample size considerations to assess sex-related treatment effects

Abstract

Randomized controlled trials (RCTs) represent the gold standard for establishing the efficacy of new interventions. Yet, RCTs are often conducted in selected populations, not representative of those at risk for or living with the disease. This lack of representativeness has raised concerns about the generalizability of RCT results to those populations not (adequately) represented in RCTs, in particular in the absence of evidence on potential differences in treatment effects across population subgroups. In the field of cardiovascular disease, women are often underrepresented in RCTs relative to disease burden. The present paper seeks to define representative participation of women in RCTs. Trial data from two cardiovascular trials, the ADVANCE trial and the LoDoCo2 trial, are used to illustrate sample size requirements for (i) achieving representative participation, (ii) achieving representative contribution to knowledge gained, and (iii) assessing whether treatment effects are the same or different in men and women (i.e. sex * treatment interactions). Achieving representative participation and representative contribution to knowledge gained requires relatively small increases in sample size and is mainly driven by sex differences in disease rates. However, sample size requirements for RCTs powered to assess sex differences in treatment effects are unfeasibly large. Hence, while efforts to enhance representative trial participation should continue, as RCTs provide the cornerstone for guideline recommendations and clinical decision-making, these should be complemented by mechanistic studies to provide insights in possible treatment interactions and inform optimal treatment for both women and men.