A non-PH accelerated hazard model for analyzing clinical trial data

In longitudinal studies with a set of continuous or ordinal repeated response variables it may be convenient to summarize the outcome as a threshold event. Then, the time to this event becomes of interest. In this paper we obtain the general likelihood for the unknown parameters when the underlying survival model is parametric and the survival times are interval-censored. We investigate the use of a member of the Generalized Time Dependent Logistic family of survival distributions (MacKenzie, 1996) which is a non- PH Accelerated Hazard Model and has a logistic baseline hazard function. We use simulation to investigate how inference on the treatment parameter is compromised by using the mis-specified likelihood, which treats the interval-censored survival times as if they were exact.