Beyond Stages: Predicting Individual Time Dependent Risk for Type 1 Diabetes

Abstract

Abstract Background Essentially all individuals with multiple autoantibodies will develop clinical type 1 diabetes. Multiple autoantibodies (AABs) and normal glucose tolerance define stage 1 diabetes; abnormal glucose tolerance defines stage 2. However, the rate of progression within these stages is heterogeneous, necessitating personalized risk calculators to improve clinical implementation. Methods We developed 3 models using TrialNet's Pathway to Prevention data to accommodate the reality that not all risk variables are clinically available. The small model included AAB status, fasting glucose, hemoglobin A1c, and age, while the medium and large models added predictors of disease progression measured via oral glucose tolerance testing. Findings All models markedly improved granularity regarding personalized risk missing from current categories of stages of type 1 diabetes. Model-derived risk calculations are consistent with the expected reduction of risk with increasing age and increase in risk with higher glucose and lower insulin secretion, illustrating the suitability of the models. Adding glucose and insulin secretion data altered model predicted probabilities within stages. In those with high 2-hour glucose, a high C-peptide markedly decreased predicted risk; a lower C-peptide obviated the age-dependent risk of 2-hour glucose alone, providing a more nuanced estimate of the rate of disease progression within stage 2. Conclusion While essentially all those with multiple AABs will develop type 1 diabetes, the rate of progression is heterogeneous and not explained by any individual single risk variable. The model-based probabilities developed here provide an adaptable personalized risk calculator to better inform decisions about how and when to monitor disease progression in clinical practice.