Heterogeneity in paediatric diabetes clinical presentation and its link to disease severity
Abstract
Background
Diabetes subtypes with different clinical profiles have been found and replicated in adults. However, clinical heterogeneity of paediatric-onset diabetes remains unexplored. Better capture of different aetiologies of the disease could prove a powerful tool towards precision medicine.
Methods
We performed data-driven clustering analysis in patients with newly diagnosed diabetes (n = 3,064) from the Norwegian Childhood Diabetes Registry. Patients were stratified by autoantibody status and clustered based on five clinical variables: Age at diagnosis, fasting glucose, HbA1c, fasting C-peptide Z-score, and BMI Z-score. We assessed inter-cluster differences regarding severity of disease, early treatment needs, polygenic scores (PS), and serum biomarkers.
Findings
We identified two clusters of autoantibody-positive and three clusters of autoantibody-negative patients: 1) Childhood severe autoimmune diabetes (CSAID; n = 1482, 48·37 %), characterized by childhood-onset diabetes, milder disease presentation, unaltered lipidic profiles, higher Type 1 diabetes PS, and autoantibody positivity; 2) Adolescence severe autoimmune diabetes (ASAID; n = 1252, 40·86 %) with adolescent-onset of diabetes, more severe disease presentation, higher diabetic ketoacidosis, lipidic profiles signaling towards prolonged metabolic disease, and autoantibody positivity; 3) Childhood severe insulin-deficient diabetes (CSIDD; n = 106, 3·46 %) and 4) Adolescence severe insulin-deficient diabetes (ASIDD; n = 144, 4·70 %) with similar characteristics to CSAID and ASAID, respectively, but no autoantibody positivity; and 5) Adolescence severe insulin-resistant diabetes (ASIRD; n = 80, 2·61 %), the oldest group, with the highest C-peptide, BMI Z-score, and Type 2 diabetes PS. CSAID and ASAID presented similarities to previously described endotypes for type 1 diabetes.
Interpretation
We grouped patients with paediatric diabetes into five subgroups with varying clinical severity, genetic risk, and metabolic profiles. Similarities between autoantibody-positive and -negative clusters underscore the importance of adopting a personalised, multivariate approach to diabetes management that extends beyond autoantibody status. We hypothesise that our clusters may be connected to previously described endotypes of type 1 diabetes, facilitating patient classification without the need for pancreatic biopsies. Further understanding of this concept could help define the mechanisms involved in disease initiation, time to diagnosis, and progression.
Related articles
Related articles are currently not available for this article.