BAGEL: A Bayesian Graphical Model for Inferring Drug Effect Longitudinally on Depression in People with HIV

TL;DR

BAGEL is a Bayesian graphical model designed to analyze the longitudinal effects of ART drugs on depression symptoms in HIV patients, accounting for individual heterogeneity and clinical factors, thereby aiding personalized treatment strategies.

Contribution

This paper introduces BAGEL, a novel Bayesian nonparametric graphical model that assesses drug effects on depression over time in HIV patients, incorporating population heterogeneity.

Findings

BAGEL effectively identifies ART drugs associated with depression symptoms.

Application to HIV study data yields clinically interpretable results.

Simulation studies demonstrate BAGEL's accuracy and robustness.

Abstract

Access and adherence to antiretroviral therapy (ART) has transformed the face of HIV infection from a fatal to a chronic disease. However, ART is also known for its side effects. Studies have reported that ART is associated with depressive symptomatology. Large-scale HIV clinical databases with individuals' longitudinal depression records, ART medications, and clinical characteristics offer researchers unprecedented opportunities to study the effects of ART drugs on depression over time. We develop BAGEL, a Bayesian graphical model to investigate longitudinal effects of ART drugs on a range of depressive symptoms while adjusting for participants' demographic, behavior, and clinical characteristics, and taking into account the heterogeneous population through a Bayesian nonparametric prior. We evaluate BAGEL through simulation studies. Application to a dataset from the Women's Interagency HIV Study yields interpretable and clinically useful results. BAGEL not only can improve our understanding of ART drugs effects on disparate depression symptoms, but also has clinical utility in guiding informed and effective treatment selection to facilitate precision medicine in HIV.