Estimating homeostasis model assessment for insulin secretion by using multiple adaptive regression spline in healthy...

The prevalence of type 2 diabetes (T2M) has been increasing drastically in recent two decades. One of the main underlying pathophysiology was decreased insulin secretion (ISEC). Even though there were many studies found the related factors affecting ISEC, no study used multiple adaptive regression spline (MARS) to build an equation estimating ISEC. In the present study, we used MARS to estimate hemostasis assessment model of β-cell (HOMA-β) in healthy Taiwanese men. Totally, there were 317 men enrolled. Participants who were taking medications related to metabolic syndrome were excluded. MARS was used to build an equation to estimate HOMA-β. Multiple linear regression (MLR) was taken as a bench mark for comparing the accuracy with MARS. The method with less estimation errors was considered to be more accurate. All the estimation errors were smaller for MARS. This indicated that MARS outperformed MLR. The equation built is shown below. The r² of this equation was 0.58. By using MARS, we built an equation which could accurately estimate HOMA-β in a healthy Taiwanese men cohort. The most important factor was HB, followed by TB, education level, sport area, GOT, GPT, and BF. This equation has a practical clinical use and could further explore which were the factors that were related to ISEC.