Estimation and Testing for Clustered Bivariate Survival Databased on an Interval-Censored Version of the Clayton-Oakes Model

• Activity Number: 136 - Recent Advances in Clustered Time-to-Event Data
• Type: Topic Contributed
• Date/Time: Monday, August 8, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Lifetime Data Science Section
• Author(s): Bernard Rosner* and Camden Bay and Robert Glynn and Gui-Shuang Ying and Maureen Maguire and Mei-Ling Ting Lee
• Companies: Harvard Medical School and Harvard Medical School and Harvard Medical School and University of Pennsylvania and University of Pennsylvania and University of Maryland
• Keywords: survival analysis; clustered data; goodness-of-fit test; Kaplan-Meier estimator; Clayton-Oakes model; statistical methods in ophthalmology

Abstract

The Kaplan-Meier (KM) estimator is ubiquitously used to estimate survival probabilities for time-to-event data. However, it assumes that the risk set at any time t consists of independent observations. This assumption does not hold for data from paired organ systems such as occur in ophthalmology (eyes), otolaryngology (ears), or other types of clustered data. In this talk, we estimate survival probabilities in the setting of clustered data, and provide confidence limits for this estimator with intracluster correlation accounted for by an interval-censored version of the Clayton-Oakes model.We also present a goodness-of-fit test applicable to any bivariate interval-censored model. In addition, we propose an unconditional likelihood ratio test to compare survival distributions between two groups in the setting of clustered data, and compare our test to the weighted log rank (WLR) test proposed by Gregg, Datta and Lorenz and the test based on the marginal Cox proportional Hazards model with robust standard errors proposed by Lin and Wei (LW); our test has appropriate type I error and higher power than the other methods. It is demonstrated in two real examples from ophthalmology.