A semicompeting risks model with an application to UK Biobank data to identify risk factors for diabetes onset and progression.

Abstract

Type 2 diabetes (T2D) is a major health concern worldwide with multiple disease stages, including onset, progression to complications, and death. Understanding the roles of genetic and nongenetic factors at different disease stages is crucial for gaining insights into disease etiology, possible prevention, and treatment strategies. The UK Biobank (UKB) is a valuable resource for studying complex diseases, including T2D, with comprehensive data from half a million volunteer participants. However, the UKB data present some unique challenges due to their semicompeting risks structure, involving 2 nonterminal events (T2D and complications) and one terminal event (death). In this paper, we propose a new shared gamma frailty-based semicompeting risks model within the Bayesian framework to account for subsequent nonterminal and terminal events and enable appropriate analysis. We further propose incorporating prevalent cases, that is, individuals with diabetes at enrollment, to gain more insights into the progression to complications and complications to death. To integrate prevalent cases, we introduce a power prior approach that leads to improved model fit and more efficient estimates. Simulation results demonstrate the efficacy of our modeling framework. We apply our method to identify the impacts of various risk factors at different stages of T2D development.