61-OR: Machine Learning Framework Predicts Metabolic Subphenotypes Using Glucose Dynamics during Oral Glucose Tolerance Test

Abstract

Listen

Translate article

Type 2 diabetes (T2D) is classically defined by measures of glycemia which do not take into account its physiologic basis, which differs between individuals. Defining the relative contribution of fundamental physiologic processes to glucose dysregulation would facilitate targeted treatment approaches for diabetes prevention and treatment. We propose that muscle insulin resistance, beta cell dysfunction, incretin defect, and hepatic insulin resistance are present to varying degrees in individuals predisposed to T2D, who can be thus classified according to their metabolic subphenotype. While this transformative approach to T2D is appealing, traditional metabolic tests are invasive and unavailable outside research units. We tested the ability of the shape of the glucose response curve during an oral glucose tolerance test (OGTT) to identify metabolic subphenotypes. 32 participants underwent gold-standard metabolic tests including frequently-sampled 3-hour OGTT, isoglycemic intravenous glucose infusion for quantification of incretin effect, steady-state plasma glucose measurement of insulin resistance, and calculated hepatic insulin resistance. We developed a machine learning framework to predict metabolic subphenotypes of an individual using dynamic features of the glucose time series during the OGTT. Metabolic testing revealed extensive inter-individual heterogeneity and the existence of four major metabolic subphenotypes. The features of the glucose time series identified insulin resistance, beta cell deficiency, and incretin deficiency with auROCs of 94%, 76%, and 75%, respectively. These were superior to standard clinical and laboratory measures of metabolic phenotypes. We suggest that identification of distinct metabolic subphenotypes using features of the glucose time series during OGTT may both enhance early identification of at-risk individuals as well as inform targeted therapeutic approaches to prevent and treat T2D. Disclosure A.A.Metwally: None. D.Perelman: None. H.Park: None. A.L.Gloyn: Other Relationship; Genentech, Inc., Roche Pharmaceuticals. T.Mclaughlin: Board Member; January, Inc., Research Support; Merck & Co., Inc., Novo Nordisk, Stock/Shareholder; Eiger BioPharmaceuticals. M.Snyder: Stock/Shareholder; January, Inc. Funding National Institutes of Health (R01DK110186-01) National Institute of Diabetes and Digestive and Kidney Diseases (U01-DK105535; U01-DK085545, UM1DK126185)