Not only does the presence of islet autoantibodies have prognostic significance, selected patients may be eligible for teplizumab (Tzield), a monoclonal antibody approved by the U.S. Food and Drug Administration in 2022 to delay the onset of type 2 diabetes in high-risk individuals. An international randomized, placebo-controlled trial tested a 14-day intravenous infusion of teplizumab in 76 nondiabetic relatives of patients with type 1 diabetes (including 55 children 8 to 17 years of age) with two or more autoantibodies and impaired glucose tolerance. The intervention group had a median time to diagnosis of type 1 diabetes of 48.4 months, compared with 24.4 months in the placebo group. A subsequent trial showed that teplizumab preserves β-cell function in patients with newly diagnosed type 1 diabetes but does not improve clinical outcomes.
Teplizumab’s exceptionally high price ($13,850 per vial; $194,000 for a 14-day course) and limited benefits—delaying type 1 diabetes onset for an average of 2 years—means that other strategies are needed to improve outcomes in at-risk children. Notably, 80% of patients with type 1 diabetes have no affected relatives. However, it is not known whether screening the general population for type 1 diabetes with genetic or autoantibody tests would lead to more benefits than harms.
One alternative to population-wide screening is using machine learning to predict type 1 diabetes in primary care. A June 2024 study in The Lancet used electronic health records from more than 2 million children in Wales to develop and validate a predictive algorithm for type 1 diabetes based on 26 symptoms, history elements, and timing of primary care visits. When the algorithm was set to generate an electronic alert at 10% of visits, 7 in 10 children with type 1 diabetes would have triggered an alert in the 90 days before diagnosis and been diagnosed an average of 9 days earlier. Fewer alerts (e.g., 5% of visits) reduced the number of early diagnoses. Although the researchers asserted that implementing the algorithm would “substantially reduce the proportion of patients with new-onset type 1 diabetes presenting in [DKA],” they acknowledged that further studies are needed to test the feasibility of this strategy and its relationship to “alert fatigue” and clinician burnout in practice.