Analyzing Length-Biased Right-Censored Data with Uncertain Onset Time

• Activity Number: 596
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Jun Liu*+ and Jing Ning and Yu Shen
• Companies: Univ of Texas MD Anderson Cancer Center and The University of Texas MD Anderson Cancer Center and UT M D Anderson Cancer Center
• Keywords: Length-biased ; right-censored ; uncertain onset ; parametric

Abstract

Right-censored time-to-event data observed from a prospective prevalent cohort are often subject to length-biased sampling. Because the sampling mechanism leads to the informative censored data, the standard survival analysis tools are not applicable to the observed length-biased data. Moreover, the onset time of initial event is sometimes not observed but in an interval, which increases challenge in analyzing such data. We consider a class of flexible parametric model to estimate the survival function of population failure time and to assess the covariance effects on the population failure time by modeling the length-biased outcome. The unobserved truncation time is integrated out over the observed intervals. We construct the full likelihood, given a biased sample with unobservable onset of disease. The parameters are estimated through the maximization of the constructed likelihood by adjusting the selection bias and interval-censored onset time. The proposed method is evaluated by simulations and is also illustrated by a study for a prevalent cohort of HIV positive patients.