Bayesian Semiparametric Joint Models to Study Weight and Islet Autoimmunity in Subjects at High Risk for Type 1 Diabetes

• Activity Number: 46 - Statistical Challenges and Breakthroughs in Diabetes and Obesity Research in the Big Data Era
• Type: Topic-Contributed
• Date/Time: Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Xiang Liu* and Roy Tamura and Kendra Vehik and Jeffrey Krischer
• Companies: University of South Florida and University of South Florida and University of South Florida and University of South Florida
• Keywords: Semiparametric joint model; Survival data analysis; Longitudinal data analysis; Type 1 diabetes; Weight

Abstract

In the pathogenesis of Type 1 diabetes (T1D), the accelerator and overload hypotheses postulate that rapid growth and high weight speed up both beta cell insufficiency and insulin resistance. Islet autoimmunity (IA) precedes clinical T1D. In order to study the effect of weight on the risk of IA, joint modeling of longitudinal and time-to-event data is necessary to consider the growth process and the development of IA, while accounting for their interrelationship. A child’s weight trajectory in early life is nonlinear. In addition, longitudinal data collected in large observational studies introduces burdens in computation. In order to address the nonlinearity and also the computational challenge, we introduce a Bayesian semiparametric joint model, in which a partial linear mixed sub-model was used to model the weight trajectories. Splines were used to approximate the nonlinear weight trajectories and the joint model was estimated within a Bayesian framework. We used data from the Environmental Determinants of Diabetes in the Young (TEDDY) study to illustrate its use and showed that weight is associated with the risk of IA.