Fig. 6From: A DAG-based comparison of interventional effect underestimation between composite endpoint and multi-state analysis in cardiovascular trialsMean hazard ratio estimates in the simulation model under λ 1=0.14,λ 2=1.17, ρ=1, a lower treatment effect on fatal than on non-fatal outcomes (log(0.92)=β 1>β 2= log(0.75)) and varying influence of an unobserved variable Z (having variance θ). Cox proportional hazards analysis for the composite and the single components, respectively (1st events), Andersen-Gill analysis (all events, composite outcome) and multi-state analysis (all events, fatal and non-fatal outcomes)Back to article page