Type 1 diabetes risk factors, risk prediction and presymptomatic detection: Evidence and guidance for screening

Abstract

Type 1 diabetes is recognized as a chronic disease with a presymptomatic phase that does not require insulin therapy and a clinical phase where insulin treatment becomes necessary. The presymptomatic phase is characterized by the presence of autoantibodies targeting pancreatic islet beta cell antigens (islet autoantibodies). This phase is further classified into three stages: Stage 1, defined by normoglycaemia; Stage 2, characterized by dysglycaemia; and Stage 3, marked by hyperglycaemia, which typically presents clinically and necessitates insulin therapy. The prospect of therapies to delay the onset of clinical disease and insulin treatment has been a driver of research into the presymptomatic phase since the discovery of islet autoantibodies. With the recent approval of teplizumab as a therapy to delay disease progression, attention has increasingly focused on diagnosing individuals with Stage 1 and Stage 2 type 1 diabetes. However, diagnosing an asymptomatic condition that affects fewer than 1 in 200 individuals poses significant challenges. As we enter this new era of diagnosis, it is crucial to refine diagnostic approaches to ensure accuracy and effectiveness. This review summarizes current evidence and guidance while emphasizing the need for continued research alongside broader application of screening. PLAIN LANGUAGE SUMMARY: Type 1 diabetes is an autoimmune disease that affects approximately 0.5% of individuals. In this publication, the authors provide a comprehensive overview of strategies for identifying individuals in the pre-symptomatic, early stages of the disease. Early-stage type 1 diabetes can be detected by the presence of autoantibodies against specific proteins in the blood, signaling an ongoing disease process before clinical symptoms appear. Genetic factors also contribute to the development of these autoantibodies and the disease itself. The paper explores how these markers are used for early identification, emphasizing optimal screening ages and the role of confirmation tests in preventing misdiagnosis. A key consideration in early diagnosis is that disease progression varies-some individuals develop clinical diabetes rapidly, while others may take many years. The authors discuss additional tests that can help predict how soon a diagnosed individual may require insulin treatment. Finally, the paper highlights ongoing challenges in optimizing screening for wider application and the complexities of integrating research-based screening into routine clinical practice.

Keywords: cohort study; cost‐effectiveness; population study; primary care; type 1 diabetes.

FIGURE 1

Effect of a priori probability on screening. The example shows the application of an islet autoantibody screening test that has a threshold for positivity set to the 99th centile of healthy controls (99% specificity). The threshold identifies 90% of the disease group (90% sensitivity). In a case‐control setting often used to evaluate the performance of a test (A), false positives are infrequent. Case B shows the performance of the same test in general population screening, where the a priori disease prevalence is 0.4% (4000 from 1 million tested). With 99% specificity in health and 90% sensitivity in disease, the test is expected to identify 3600 type 1 diabetes cases (90% of the total cases) plus another 9960 who will not develop type 1 diabetes, with a positive predictive value (PPV) of 27%.

FIGURE 2

Early‐stage type 1 diabetes screening strategy decision tree options. Option A starts with islet autoantibody testing in all children with a series of AND commands leading to second assays and samples depending upon the screening test and confirmation test results. Option B includes an a priori selection based on genetic risk with OR commands that identify a subset of the population to be tested for islet autoantibodies following the AND command decision tree. The efficiency of screening, and in particular for option B, is dependent upon the recall rates achieved for autoantibody testing.