Developing a blood fatty acid-based algorithm as an early predictor of insulin resistance: Applying machine learning to harmonized data from prospective cohort studies

Project Summary Metabolic diseases, such as Metabolic Syndrome (MetSyn) and type 2 diabetes (T2DM) affect ~37% and ~10% of Americans, respectively, with a substantial impact on the health, quality of life and financial security of these individuals. Without the means to identify high risk individuals conveniently and cheaply, many will find care too little and too late: more than 80% of individuals with pre-diabetes don’t know they have it. Thus, there is a need for accessible and inexpensive early predictive biomarkers of MetSyn and/or T2DM to facilitate the early identification of high-risk individuals, providing the time necessary to make meaningful lifestyle changes to slow or prevent disease progression. Hemoglobin A1C (HbA1c) and fasting plasma glucose are traditional markers of hyperglycemia and insulin resistance, however, these markers have significant limitations in that they are measures of existing, not impending, disease. Other tests of insulin resistance (e.g., oral glucose tolerance tests and hyperinsulinemic euglycemic clamps) are inefficient, cumbersome for the patient, and expensive; thus, they are also not viable options as early screening tools. Emerging evidence suggests that erythrocyte (RBC) fatty acid (FAs) profiles may serve as an early signal of impending hyperglycemia up to 5 years before T2DM develops regardless of whether insulin sensitivity, insulin secretion and glycemic status are known. As a clinical laboratory that specializes in providing FA measurements, interpretation and customized behavioral interventions, OmegaQuant Analytics (OQA) supports a large and growing customer base of researchers, clinicians, businesses, and individuals. OQA has measured full FA profiles (28 FAs) on numerous randomized clinical trials and prospective cohort studies. Furthermore, OQA is a leader in the at-home FA testing market through its innovative dried blood spot collection system, testing almost 40,000 samples annually. Through a partnership with the Fatty Acid Research Institute (FA expertise; biostatistical support; data access), we propose to develop highly predictive metabolic indices from FA profiles using an innovative approach leveraging existing prospective cohort data. We will do this by (Aim 1) predicting future MetSyn or T2DM status from an RBC FA signature using a harmonized dataset from five leading prospective cohort studies yielding a sample of 10,264 individuals (including 1,173 minorities). Prospective statistical predictions using the harmonized data sets will yield the FA Metabolic Index (FAMI) and FA Diabetes Index (FADI). We will then (Aim 2) explore how FAMI and FADI can be leveraged to profitability by creating interpretative reports for FAMI and FADI, providing an understanding and actionable set of steps to change dietary behaviors to modify MetSyn/T2DM risk and exploring willingness of clinicians, laboratories and individuals to purchase the tests. These two simple, early-warning tests will allow for targeted intervention of individuals at high-risk for developing MetSyn and/or T2DM, ultimately leading to substantial reductions in the prevalence and societal burden of this disease.