Utility of diabetes-associated autoantibodies for classification of new onset diabetes in children and adolescents.

OBJECTIVE: To determine whether measuring diabetes-associated autoantibodies (DAA) in pediatric new onset diabetes (NODM) can be restricted to patients with equivocal diabetes type. RESEARCH DESIGN AND METHODS: Retrospective analysis of all patients with NODM admitted to Boston Children's Hospital from 1 October 2007 to 1 July 2013 who had measurement of DAA [glutamic acid decarboxylase, insulin, insulinoma-associated antigen 2 (IA-2)]. Data collection included initial diagnosis of diabetes type before DAA results and at follow-up. We used logistic regression to predict type 1 diabetes (T1D) and developed a clinical score to classify diabetes type. RESULTS: Of 1089 patients (45.4% female, 76.7% White, age 10.6 ± 4.5 yr), initial diagnosis was 1021 (93.8%) T1D, 42 (3.9%) type 2 diabetes (T2D), and 26 (2.4%) other. Of 993 patients with clinical T1D, 78 (7.9%) were DAA-, and of 42 patients with clinical T2D, 12 (28.6%) were DAA+. Type of diabetes was reclassified at follow-up in less than 6% of patients. Data from a subset of 736 patients were used to develop a scoring system to predict T1D. Using weight z-score, age, and race, the scoring system had 91.7% sensitivity, 82% specificity, and a positive predictive value of 98.6%, and suggested DAA measurement was unnecessary in 85.3% of patients. Findings were similar in a validation cohort of 234 patients. CONCLUSIONS: Application of a simple scoring system may reduce to ∼15% the number of DAA measurements needed to classify diabetes type, resulting in substantial cost savings. Clinical judgment should guide the decision to measure DAA.