Treatment effects by baseline hematocrit level on the primary outcome and each component. The treatment effect based on baseline hematocrit levels on the primary composite outcome in males (A) and females (B), on MI in males (C) and females (D), on stroke in males (E) and females (F), and on any thromboembolism in males (G) and females (H). In males, the treatment effect on the primary outcome and MI varied according to baseline hematocrit levels (P interaction < .05), showing a benefit in anemic patients but a detrimental effect in those with erythrocytosis. By contrast, no effect modifications were observed in females. In addition, a sex-specific effect modification by hematocrit level was identified for the primary outcome (baseline hematocrit by treatment by sex interaction, P = .02). The black line indicates the point estimate, whereas the gray area represents the 95% CI. The graphs have truncated the lowest and highest 0.5 percentiles.