A Bayesian population analysis of the development of type 2 diabetes in the offspring of diabetic parents

Abstract

Disease progression of type 2 diabetes (T2D) has received considerable attention, but little is known about the disease development of T2D. The purposes of this study were to identify disease development variables (DDV) for development of T2D and to compare corresponding models for disease development. All subjects included in this study were the offspring of diabetic parents and were followed up to 25years. Repeated fasting blood samples were collected during the follow-up. Longitudinal data of four DDVs, namely fasting blood glucose (FBG), fasting serum insulin (FSI), homeostatic model assessment of insulin resistance (HOMA-IR) and body mass index (BMI) were recorded and compared. According to the diabetes status at the end of the follow-up, the data analysis involved a progressor group of 25 subjects, and a non-progressor group of 127 subjects. The temporal changes in the four DDVs over the time course of the disease development were evaluated by a single-slope and a dual-slope population-based Bayesian model. A dual-slope model based on FBG was found to be the best disease development model. For non-progressors, the FBG baseline stayed at 69.2 [66.5, 72.1]mg/dl (Bayes estimate [95% Bayesian credible set]) and increased with age by a rate of 0.227mg/dl [0.149, 0.3] per year. For the progressors, the FBG increase with age the same rate as non-progressors and started to have an additional increase of 2.27 [0.505, 4.52]mg/dl per year, starting 8.73 [-10.8, -6.93] years before the diagnosis of T2D. No significant longitudinal increasing or decreasing temporal pattern was found for FSI, HOMA-IR and BMI by the population-based Bayesian approach. The proposed model, which enables a quantitative, time-based evaluation of the development of T2D in this higher risk population, may be used to quantify the effect of interventions/prevention strategies such as drug treatment and lifestyle changes.