Theory 98/Selection & Bias/Cohort studies
Time Before Treatment Counts
The claim
Classifying participants by treatment received after follow-up begins creates immortal time, biasing survival in favour of the treated group.
01Mechanism
Participants must survive to receive treatment; the period before treatment is counted as exposed, which no one who died could contribute.
02Prediction
Bias should shrink when treatment is treated as time-varying.
03Refutation
Stable estimates under time-varying analysis would remove the bias.