Immortal time bias, and how to avoid building it into your study
A period during which the outcome could not have happened, assigned to one group by the way the study defines exposure. It reliably makes treatments look protective.
Immortal time is follow-up during which, by the design of the study, a participant could not have experienced the outcome — usually because experiencing it would have prevented them being classified as exposed.
How it appears
Take a study of whether a drug improves survival, where "treated" means having received a prescription during follow-up.
To receive the prescription, a patient had to survive until the day it was written. The time between study entry and that day is immortal: nobody in the treated group could have died during it, because dying would have put them in the untreated group.
Assigning that time to the treated group inflates their survival by construction. The apparent benefit can be large, and it appears in a study where nothing else is wrong.
Where it hides
It is not confined to prescriptions. Any exposure defined by something that takes time to happen carries the same risk.
- Surgery performed at some point during admission.
- Transplantation, in a cohort followed from listing.
- Achieving a treatment response.
- Adherence measured over a period, then used to define the group.
How to avoid it
- Define exposure at a fixed time and classify everyone at that moment. Start follow-up there for both groups.
- Where exposure genuinely occurs later, treat it as time-varying: participants contribute unexposed person-time until the exposure occurs and exposed person-time afterwards.
- Consider a landmark analysis: choose a point after entry, classify by exposure status at that point, and analyse only those still at risk then.
- Emulate the trial you would have run. Ask when a real trial would have randomised, and start follow-up there. Most immortal time disappears when that question is asked.
How to check a published study for it
Find the sentence defining the exposure groups, and the sentence defining when follow-up starts. If exposure is defined by something that happens after follow-up begins, ask where that time was assigned. The answer is often not in the paper.
Frequently asked questions
What is immortal time bias?
Follow-up during which, by the study's design, a participant could not have experienced the outcome — usually because experiencing it would have prevented them being classified as exposed. Assigning that time to the exposed group inflates their survival by construction.
How do you avoid immortal time bias?
Define exposure at a fixed time and start follow-up there for everyone; or treat exposure as time-varying so participants contribute unexposed person-time until it occurs; or use a landmark analysis. Emulating the trial you would have run removes most of it.
How do I spot immortal time bias in a published study?
Find the sentence defining the exposure groups and the sentence defining when follow-up starts. If exposure is defined by something happening after follow-up begins, ask where that time was assigned. The answer is often not in the paper.