Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 338 - Semiparametric and Non-Parametric Methods in Survival Analysis
Type: Contributed
Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
Sponsor: Biometrics Section
Abstract #312454
Title: Kaplan-Meier Estimation for Correlated Data
Author(s): Bernard Rosner* and Camden Bay and Mei-Ling Ting Lee and Robert Glynn
Companies: Brigham and Women's Hospital and Brigham and Women's Hospital and University of Maryland and Brigham and Women's Hospital
Keywords: Kaplan-Meier; correlated data; clustered data; log-rank; survival analysis

The Kaplan-Meier (KM) estimator is widely used to estimate survival probabilities for time-to-event data. It has the advantage of being nonparametric and thus does not require specification of a survival distribution model. However, it does assume that the risk set at any time consists of independent observations. This assumption does not hold for data originating from paired organ systems such as in the fields of ophthalmology (eyes) or otolaryngology (ears). We have generalized the Greenwood variance estimator for the traditional KM estimator in the presence of correlated data by (a) assuming a beta-binomial probability distribution for the number of failures at unique failure times (b) constraining the odds ratio for failures between two subunits at the same time to be the same over all time points and (c) using a profile likelihood approach to estimate the odds ratio that maximizes the log likelihood. We have used this approach with a dataset of 956 eyes (478 patients) followed for progression of retinopathy in a clinical trial of diabetic retinopathy. Correlation-adjusted Kaplan-Meier curves and CI’s were calculated for each arm and compared with standard KM estimates.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program