Fig. 3From: Interval-cohort designs and bias in the estimation of per-protocol effects: a simulation studyIntention-to-treat survival estimates by treatment armBack to article page