Bayesian Joint Modeling and Selection Among Many Biomarkers Measured Longitudinally

• Activity Number: 557 - New Directions in Bayesian Methods for Longitudinal and Graph Data
• Type: Contributed
• Date/Time: Thursday, August 11, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Soumya Sahu* and Sanjib Basu and Jiehuan Sun and Joelle Hallak
• Companies: Biostatistics, University of Illinois Chicago and Biostatistics, University of Illinois Chicago and Biostatistics, University of Illinois Chicago and Ophthalmology, Illinois Eye and Ear Infirmary, University of Illinois Chicago; AbbVie
• Keywords: Joint Modeling; Variable Selection; Non-parametric; Multivariate; Macular degeneration; Ophthalmology

Abstract

The literature on joint modeling is diverse; however, current approaches can typically jointly analyze one or a few longitudinal processes and a time-to-event outcome. This work is motivated by imaging features of the eye, measured longitudinally at multiple visits of patients with early-stage age-related macular degeneration (AMD). A primary scientific question in this context is selection of a panel of features that can prognosticate conversion to neovascular AMD. We develop a Bayesian nonparametric joint model that (1) flexibly models the longitudinal trajectories, (2) provides flexibility in modeling the association between the longitudinal processes and time-to-event outcome, and (3) addresses selection among the multiple longitudinally measured features. We compare performance of the proposed approach with other existing models and machine-learning models used in scientific literature. Our analysis of all imaging features simultaneously in the proposed model highlights unique features associated in multiple ways with prognostication of conversion to neovascular AMD that are distinct from findings based on marginal joint modeling involving one longitudinal feature at a time.