Remission duration: an example of interval-censored observations.

Remission duration data from two observational studies, the German ALL/AUL study on acute lymphoblastic and acute undifferentiated leukaemia and the Kiel Lymphoma Study, are used to demonstrate the adequacy of an interval-censoring approach. The Turnbull estimator is contrasted with the conventionally used Kaplan-Meier estimator. In addition, a parametric model is used for estimation or simulation of the delay times of complete remission diagnosis and relapse diagnosis. Two possible consequences of the conventional approach are discussed: biased estimation (for example, overestimation of remission duration), and underestimation of the true error variance, which may lead to false positive results. On the other hand, the applicability of the interval-censoring approach is doubtful when the censoring mechanism (examination pattern, patient's behaviour) is confounded with the endpoints.