Impact of sex-representation imbalance on effect sizes in cardiovascular randomized controlled trials
X Zhong, Q Xu, L Tao, H ShangAbstract
Background
Women are often underrepresented in cardiovascular randomized controlled trials (RCTs), raising concerns about the generalizability of reported treatment effects. However, the extent to which sex-representation imbalance alters overall effect estimates remains unclear.
Purpose
To evaluate whether sex-representation imbalance in cardiovascular RCTs associated with changes in overall reported effect sizes after post-stratification weighting based on real-world sex prevalence.
Methods
We conducted a systematic review of cardiovascular RCTs published in 2023 that reported sex-stratified treatment effects. For each trial, we extracted the proportion of female participants, sex-specific effect sizes, and the overall reported effect estimate. The participation-to-prevalence ratio (PPR) was calculated as the proportion of women among trial participants divided by the proportion of women in the corresponding real-world disease population. Real-world sex prevalence data were obtained from contemporary epidemiological sources. To assess the influence of sex-representation imbalance, we recalculated overall effect sizes using post-stratification weighting based on the real-world sex distribution. Sex-specific risks were weighted according to female and male disease prevalence to derive a sex-standardized overall effect size. The relative difference between the sex-standardized and originally reported effect sizes was calculated to quantify the magnitude of change attributable to sex-representation imbalance.
Results
Among 137 trials involving 148,204 participants, the overall mean PPR was 0.79 (SD 0.34), indicating moderate underrepresentation of women on average. Notably, 20% of trials demonstrated marked underrepresentation (PPR ≤0.5), whereas 12% showed overrepresentation (PPR >1.2); only 31% of trials fell within the balanced range (0.8–1.2). Fifty-two trials reported sex-stratified subgroup analyses, of which six identified statistically significant sex-based effect modification. Among 37 trials with extractable sex-specific effect estimates, reweighting according to real-world female disease prevalence resulted in two reversals of statistical significance (both from significant to non-significant). Overall, 13 of 37 trials (35.1%) exhibited a relative shift in effect size greater than 5%, and 12 trials (32.4%) showed a shift greater than 10%, with a median absolute change of 2.09%.
Conclusion
Sex imbalance was prevalent across contemporary cardiovascular RCTs. Post-stratification weighting revealed that although overall shifts were generally modest, effect estimates were meaningfully altered in a subset of studies. These findings highlight the potential sensitivity of reported treatment effects to sex distribution. The analysis was constrained by aggregate-level data and assumptions inherent to prevalence-based weighting.Table 1Characteristics of included trialsFigure 1Impact on effect estimates