82-OR: A T1D Genetic Risk Score Combined with Clinical Features and Autoantibodies Enables Accurate Diabetes Classification in a Racial/Ethnically Diverse Population: The Search for Diabetes in Youth Study

Abstract

Accurate classification of diabetes type guides correct treatment. We recently showed a type 1 diabetes (T1D) genetic risk score (GRS-1), combined with clinical features and biomarkers discriminates between T1D and type 2 (T2D) and MODY in European adults. We aimed to test the ability of an improved 67 SNP T1D score (GRS-2) to discriminate T1D in an ethnically diverse U.S. pediatric population. We included 1818 SEARCH Study participants who had a C-peptide at a median (IQR) of 8 (6-9) years post diagnosis. T1D was defined based on severe insulin deficiency (fasting C-peptide <0.25 ng/ml) at follow-up. We generated genetic risk scores from SNP data, and assessed discriminative power of the T1D GRS-2 using the area under the receiver operating curve (AUC) in the three largest ethnic groups in SEARCH: white (n=1101, 95% T1D), black (n=228, 59% T1D) and Hispanic (n=257, 77% T1D). We then assessed the discriminative power of GRSs combined with autoantibody and clinical data. The T1D GRS-2 was discriminative of T1D across all three ethnic groups AUC (95% CI) for GRS-1 vs. GRS-2 was 0.88 (0.84, 0.93) vs. 0.88 (0.83, 0.92) in white, 0.82 (0.77, 0.87) vs. 0.87 (0.83, 0.92) in blacks, and 0.87 (0.82, 0.92) vs. 0.9 (0.86, 0.95) in Hispanics). Combined risk scores were most accurate at classifying T1D. AUC for different scores were: 0.93 (0.91, 0.95) for type assigned by provider at diagnosis, 0.95 (0.93, 0.96) for age and BMI, 0.95 (0.94, 0.97) for GAD/IA-2/ZNT8 autoantibodies, and 0.99 (0.99, 1.00) for a combined approach using autoantibodies, GRS-2 and clinical features. These results were similar across all race/ethnic groups. A new T1D GRS-2 is discriminative of T1D in a racial/ethnically diverse pediatric population. Using GRS2 in a combined model of clinical features, autoantibodies and genetics offers near perfect classification of T1D and could be used to accurately classify diabetes type. Disclosure R.A. Oram: Other Relationship; Self; Randox Laboratories Ltd. S.A. Sharp: None. C. Pihoker: None. L.A. Ferrat: None. G. Imperatore: None. S. Saydah: None. A.H. Williams: None. L.E. Wagenknecht: None. J.M. Lawrence: None. M.N. Weedon: None. R. Dagostino: Consultant; Self; Acelity, Amgen Inc. W. Hagopian: Research Support; Self; Novo Nordisk A/S. J. Divers: None. D. Dabelea: None. Funding National Institutes of Health (UC4DK108173)