Search

Objective: Biomarkers predicting risk of type 1 diabetes (Stage 3) among children with islet autoantibodies are greatly needed to prevent DKA and facilitate prevention therapies. Research Design and Methods: Children in the prospective TEDDY study (n=707) with confirmed diabetes-associated autoantibodies (GADA, IA-2A and/or IAA) and ≥2 HbA1c measurements were followed to diabetes or median age 11.1 years. Once confirmed autoantibody positive, HbA1c was measured quarterly. Cox models and receiver operative characteristic (ROC) analyses revealed the prognostic utility for risk of Stage 3 on a relative HbA1c increase from the baseline visit or an OGTT 2-hr plasma glucose (2-hPG). This HbA1c approach was then validated in the TrialNet Pathway to Prevention study (n=1190). Results: A 10% relative HbA1c increase from baseline best marked increased risk of Stage 3 in TEDDY (74% sensitive; 88% specific). Significant predictors of risk for HbA1c change were age and HbA1c at baseline test, genetic sex, maximum number of autoantibodies and maximum rate of HbA1c increase by time of change. The multivariable model featuring a HbA1c >10% increase and these additional factors revealed increased risk of Stage 3 in TEDDY (HR=12.74, 95%CI 8.7-18.6, p>10% increase performed similarly to an OGTT 2-hPG composite model (TEDDY AUC=0.88 and 0.85, respectively) and to the HbA1c model in TrialNet (AUC=0.82). Conclusion: An increase of >10% in HbA1c from baseline is as informative as OGTT 2-hPG in predicting risk of Stage 3 in youth with genetic risk and diabetes-associated autoantibodies.

IA sensitively predict future T1D, but onset often occurs many years after seroconversion. Early diagnosis results in less DKA and milder initial disease, especially in young children. Using the international TEDDY cohort, we asked if quarterly HbA1c testing could predict T1D among pre-pubertal subjects with ≥1 persistent IA (GADA, IA2A, IAA). Mean(SD) age at seroconversion and T1D onset was 43(29) and 73(22) months respectively. Of 8,504 HLA high risk children, 456 had persistent confirmed IA, had undergone at least 3 quarterly HbA1c tests in the prior 12 months, and were not on diabetes drugs or diets. Of these, 104 progressed to T1D and 352 did not. Subjects were split into training (292 total, 62 progressors) and test (164 total, 40 progressors) datasets with similar characteristics. The optimal maximum HbA1c cutpoint within 1 year pre-onset to predict T1D, by ROC analysis, was similar in training and test datasets (p=0.66). For both, HbA1c ≥5.6% gave the best Youden index. Other factors significant for T1D risk in the training dataset included age at HbA1c test, month of HbA1c test, continent and IA2A titer, but not gender, family history, HLA type, seroconversion age, IAA titer nor T1D-related SNP genotypes. Adjusted for significant factors, the optimal HbA1c cutpoint of ≥5.6% for quarterly testing was 91% sensitive, 91% specific, 73% PPV and 98% NPV to predict T1D within 1 year in children with HLA and IA risk. The final predictive model fit both training and test data similarly (p=0.28). Median(SD) time from first HbA1c ≥5.6% to diagnosis was 8.6(4.5) months, enabling monitoring and education for earlier treatment. Quarterly HbA1c surveillance is cost-effective and clinically accessible. A hierarchical plan for pediatric T1D prediction is proposed comprising initial genetic screening, then IA surveillance of those at genetic risk, then HbA1c surveillance of those with IA, and finally close glycemic surveillance of those with HbA1c ≥5.6% to ascertain metabolic onset. Disclosure M. Killian: None. K. Vehik: None. R.R. Little: None. H. Elding Larsson: None. M.J. Haller: None. M. Rewers: None. J. She: None. J. Toppari: None. B. Akolkar: None. J. Krischer: None. W. Hagopian: None.