Study Design Issues and Analysis of Error-Prone Time-to-Event Data Using Progression-Free Survival as an Example

• Activity Number: 71
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: ENAR
• Author(s): Sally Hunsberger*+ and Paul Albert and Lori Dodd
• Companies: NCI and NICHD and National Institutes of Health
• Keywords: conditional independence ; Gaussian random effects ; Sensitivity ; Specificity

Abstract

In this talk we explore the estimation and testing of covariate effects on time-to-event data where event data is obtained at pre-specified discrete time points. In this setting, disease status can be assessed with two diagnostic tests: one being a test that has no measurement error and one that has error but may be less expensive, less invasive or less difficult to obtain. When data using the test with measurement error are analyzed, inferences about covariate effects may be invalid. A sampling strategy is proposed where data are collected on a subset of subjects using the test that has no error and on all subjects using the test that has error. The strategy is designed to maintain valid inferences while requiring the more expensive or difficult test on a small proportion of patients. We explore the effect of different diagnostic test properties on inference via simulation and illustrate the methodology with a renal cancer example. The simulations show inference is correct when a subset of measurements without error are collected.