Cumulative/Dynamic Time-Dependent ROC Analysis for Left-Truncated and Right-Censored Data

Abstract

Time-dependent Receiver Operating Characteristics (ROC) analysis is a standard method to evaluate the ability of a biomarker or predictive score to discriminate between cases and non-cases for time-to-event outcomes. Extensions of this useful method to left-truncated right-censored data have been understudied, with the exception of Li 2016 under limited settings. In this paper, we first developed several extended estimators based on Li 2016 for settings with independent or covariate-induced dependent left truncation and right censoring. We further introduced novel inverse probability weighting estimators of sensitivity, specificity, and area under the ROC curve (AUC), where the weights simultaneously account for left truncation and right censoring, in various scenarios. We evaluated the proposed AUC estimators in comprehensive simulation studies and performed the proposed time-dependent ROC analysis to evaluate the discriminatory performance of a mortality prediction model in elderly end-stage renal disease patients using the Organ Procurement and Transplantation Network data.

arXiv comment49 pages, 5 figures