Fig. 2: Power to detect association by env-MR-MEGA and MR-MEGA under different ancestral heterogeneity settings with similar smoking proportions between male and female cohorts.

Across six heterogeneity scenarios involving 16 sex-stratified cohorts, where there are minor random reductions or increases in smoking proportions between male and female cohorts (mixed direction), env-MR-MEGA exhibits higher power in detecting association compared to MR-MEGA, especially in the ancestrally homogeneous and non-ancestral Africa scenarios. “env_MR_MEGA_assoc” and “MR_MEGA_assoc” refer to the power to detect association obtained from env-MR-MEGA and MR-MEGA. Power was assessed at P < 5 × 10−8 and based on 1000 replications with unequal sample sizes (≥3000 in each cohort).