Auxiliary Covariate Augmentation with Correlated Joint Time Toevent and Longitudinal Outcomes

Hannah Pervin, The University of Texas School of Public Health

Abstract

The main outcome of a clinical trial is often a time-to-event outcome based on treatment group assignment. Including additional baseline and longitudinal covariates may improve the predictive model for the outcome and the estimate of the treatment effect. Using semiparametric methods, this analysis approach allows for more efficient estimation of the treatment on a time-to-event outcome and a secondary, correlated longitudinal marker. Simulations compare this augmented joint model to other existing analysis approaches under various conditions. ^ The augmented joint method was also applied to real data. Stroke is associated with high blood pressure. Antihypertensive medications confer protection from stroke as well as achieving lower blood pressure. Among 33357 participants in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT), 392 fatal strokes occurred during the follow-up, and 1525 participants had at least one fatal or non-fatal stroke. Chlorthalidone was associated with lower systolic blood pressure (SBP) compared to amlodipine or lisinopril, and, after accounting for follow-up SBP through a joint model augmented by auxiliary baseline covariates, was still associated with lower risk of stroke than lisinopril. This effect was especially present among Blacks. ^

Subject Area

Biostatistics|Epidemiology

Recommended Citation

Pervin, Hannah, "Auxiliary Covariate Augmentation with Correlated Joint Time Toevent and Longitudinal Outcomes" (2017). Texas Medical Center Dissertations (via ProQuest). AAI10617279.
http://digitalcommons.library.tmc.edu/dissertations/AAI10617279

Share

COinS