|
Abstract:
|
The onset of several chronic diseases such as diabetes are asymptomatic can only be detected by diagnostic tests. In such settings, the true time to event of the event of interest is not observed exactly - when perfect diagnostic tests are administered sequentially, the time to event is only known up to the interval of time between the last negative and the first positive test. Such data are referred to as interval censored. In this paper, we describe semi-parametric and parametric approaches to calculate power and sample size in settings where the outcome is interval censored. We describe a semi-parametric approach in which the distribution of the time to event is left unspecified - we compare this strategy to parametric methods in which the time to event is assumed to follow a piece-wise exponential or Weibull distribution. We compare the small sample properties and robustness of the proposed methods under various settings. We illustrate the implementation of the proposed methods through a freely available R software package.
|
