Estimation, testing and adjustment of transient healthy entrant effects

At recruitment, cohorts often selectively exclude non-healthy individuals with symtoms prodromal to the response. After a certain time from study entry, the health status of participants regresses towards the underlying population. Methods developed here estimate, test and adjust out this transient healthy entrant effect from survival function estimation. In the simplest case, each participant's time is divided into: Ω₁i - immediately after study entry while these transient healthy entrant effects are strong, and Ω₂ - later, after they have vanished or greatly diminished. Truncated Kaplan-Meier models fit separately to risk sets in Ω₁ and Ω₂ quantify and remove the bias from transient healthy entrant effects and test or statistical significance. If sufficient data are available, these methods can be expanded to dividing the subjects' time into multiple subperiods or fitting specific models of attenuation for the transient healthy entrant effect. Examination of nickel refinery workers and an HIV-1 infected cohort suggest that, without the adjustments proposed here, transient healthy entrant effects cause moderate overestimations of survival probability.