borderline significancep = 0.036
Sensitivity analyses addressing missing smoking data confirmed the robustness of this effect: HRs for smokers versus non-smokers remained around 1.9–2.0 with borderline significance ( p = 0.036–0.052), and although attenuated when “unknowns” were treated as smokers (HR = 1.53; p = 0.089), the direction of association persisted.