Hazard Regression for Interval-Censored Outcome with Interval-Censored Covariate

• Activity Number: 349 - Lifetime Data Science Student Awards
• Type: Topic-Contributed
• Date/Time: Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Lifetime Data Science Section
• Author(s): Yue Song* and Wenbin Lu and Rui Wang
• Companies: Harvard T. H. Chan School of Public Health and North Carolina State University and Harvard Pilgrim Health Care Institute
• Keywords: HIV viral rebound; EM algorithm; Composite likelihood; Robust variance

Abstract

Motivated by the need to assess whether the time to viral suppression after ART initiation is predictive of the time to viral rebound after ART interruption, we investigate modeling approaches relating an interval-censored outcome (e.g., time to viral rebound) and an interval-censored covariate (e.g., time to viral suppression). To this end, we present a proportional hazards regression model and use an Expectation-Maximization algorithm for parameter estimation, where the observed data are augmented with Poisson random variables to facilitate computation. Moreover, some individuals experienced multiple episodes of treatment initiation and interruption. We extend the method to accommodate the clustering effect where estimation is based on the independent composite likelihood function and inference makes use of a robust variance estimator. We evaluate the finite-sample performance of the proposed procedure for both independent and clustered settings through simulations. We apply it to data from the Zurich Primary HIV Infection Cohort to assess the effect of time to HIV viral suppression after antiretroviral treatment initiation on time to viral rebound after treatment interruption.