Immortal time bias: the period when the outcome could not happen

Immortal time is a stretch of follow-up time during which a study's design makes it impossible for the outcome being measured, such as death, to be counted against a group. When that stretch gets misclassified or folded into the treated group's favor, it makes a treatment look more effective than it really is, an error called immortal time bias.

The bias was first identified in the 1970s in studies of heart transplant survival, where patients waiting for a donor heart were compared with those who had already received one. Anyone who died while waiting was counted in the untreated group, but the time a transplant recipient spent waiting before their operation was effectively erased from the untreated side and folded into the treated side's survival time, even though they could not have died from the transplant during the period before they received it.

Two traps follow. First, the bias tends to appear whenever a study defines a treated group using information that could only be known after the study's starting point, such as whether someone eventually filled a prescription. Second, the distortion grows with the length of the immortal period, so a study built from many years of records carries more risk than a short one, and the more impressive a drug's apparent benefit looks, the more worth checking how the treated group was defined.

When an observational study reports a large survival benefit for a drug or treatment, check how the treated and untreated groups were defined and from what starting point their follow-up time was counted. If people had to survive long enough to receive or continue the treatment before being counted as treated, some of that survival time may have been immortal by design rather than earned by the treatment. Pair a striking survival claim with a look at whether the study, or a later re-analysis, addressed immortal time directly.