Table 1.

Overview of the four strategies of dealing with treatment initation after baseline

Strategy	Estimand	Example	Estimators	Key assumptions
Ignore treatment	Risk of the event, regardless of treatment	Risk of cardiovascular events where some patients will initiate statins according to routine-care prescriptions	Survival model for T, do not censor at V	Treatment assignment policy in application setting similar to development data
Composite	Risk of the event or treatment initiation	Risk of a composite of cardiovascular death, myocardial infarction and treatment with revascularisation (PCI or CABG)	Survival model for min(T, V)	Treatment assignment policy in application setting similar to development data
While untreated	Risk of the event occurring before treatment is started	Risk of dying while on the waiting list for a liver transplant	Competing risks methods	Treatment assignment policy in application setting similar to development data
Hypothetical	Risk of the event if treatment were never started	Risk of a natural pregnancy without IVF treatment	Survival model for T, censor at V or include treatment as time-dependent covariate in the model and set to 0 when predicting	Exchangeability, consistency and positivity

T time to event of interest; V time to start of treatment; PCI percutaneous coronary intervention; CABG coronary artery bypass grafting; IVF in vitro fertilization