Childhood-onset diabetes is currently classified as autoimmune type 1 diabetes, type 2 diabetes, or monogenic diabetes (1–3). The pathophysiology of type 1 diabetes involves immune-mediated β-cell failure, while type 2 diabetes is characterized by nonautoimmune relative insulin deficiency for the degree of insulin resistance (4). The prevalence of both type 1 and type 2 diabetes continues to increase (5,6). Clinically, type 2 diabetes is distinguished from type 1 diabetes by the absence of autoimmunity and characteristic clinical features, including presence of obesity, signs of insulin resistance, and strong family history of the disease (1). Given the high prevalence of childhood obesity, including in youth with type 1 diabetes (7,8), the distinction between type 2 diabetes and type 1 diabetes, particularly at presentation or early in the disease course, may be challenging. Additionally, markers of autoimmunity may be absent in youth with type 1 diabetes but can be detected in youth with a clinical diagnosis of type 2 diabetes (9,10). In youth diagnosed clinically with type 2 diabetes, islet autoantibodies were reported in up to 10% upon screening for participation in the Treatment Options for Type 2 Diabetes in Adolescents and Youth (TODAY) clinical trial (11) and in up to 20% in the SEARCH for Diabetes in Youth (SEARCH) study (12).
Accurate diagnosis of diabetes type is important, as it has major implications for management decisions and prognosis. Indeed, in our studies of youth with obesity and a clinical diagnosis of type 2 diabetes, those with positive autoantibodies were characterized by greater insulin sensitivity but significantly worse β-cell function compared with those with negative autoantibodies, suggesting higher risk of progression to β-cell failure and need for a different therapeutic approach (13). In addition to markers of autoimmunity, C-peptide concentration, a marker of β-cell reserve, may help in the diagnosis and management of the disease. Retained C-peptide is associated with better glycemic control and less risk of complications in individuals with type 1 diabetes (14). C-peptide concentrations are higher in youth with type 2 diabetes with negative autoantibodies compared with autoantibody-positive peers (11,15). Stimulated C-peptide concentrations were related to glycemic control at diagnosis (16) and predicted long-term therapeutic outcomes in youth with type 2 diabetes (17). However, there is significant variability in the concentrations of C-peptide. Clear thresholds of these biomarkers are hard to establish given the need to examine insulin production relative to insulin sensitivity regardless of the type of diabetes, duration of diabetes, and importance of distinguishing fasting versus stimulated biomarkers. To improve diabetes classification, models that include both clinical and biochemical features are needed. Additionally, autoimmunity markers and the genetic risk profile (18), which may influence insulin sensitivity and/or β-cell function or other biologic pathways, need to be considered. Clinical prediction models that incorporate these principles may help in the classification of youth-onset diabetes.
The study by Jones et al. (19) in this issue of Diabetes Care presents such a clinical prediction model, combining clinical features and autoimmunity biomarkers obtained at diabetes diagnosis in 2,966 youth (age ≤19 years) from the prospective SEARCH study. The model included clinical measures at diagnosis (age, sex, BMI, waist circumference, and HDL cholesterol), islet autoantibody (GADA and IA-2A) status, and type 1 diabetes genetic risk score (T1DGRS). These data were available for 1,918 participants. The model aimed to identify participants with fasting C-peptide ≥250 pmol/L (≥0.75 ng/ml) obtained between 3 and 10 years (median 74 months) after diagnosis, defined as preserved endogenous insulin secretion consistent with type 2 diabetes. The model was found to be accurate in identifying participants with C-peptide ≥0.75 ng/mL (17% of participants; 2.3% vs. 53% of those with and without positive autoantibodies). The area under the receiver operator curve was 0.95 for the model that included clinical characteristics alone, rising to 0.97 with the inclusion of islet autoantibodies, and it remained similar with the addition of T1DGRS (0.98). The model performed well in youth with obesity (area under the receiver operator curve 0.88–0.96), across race and ethnicity subgroups (0.88–0.97), by autoantibody status (0.87–0.96), and by clinical provider–diagnosed diabetes type (0.81–0.92).
The strengths of this study are the large sample size, the racial/ethnic diversity, and the availability of prospective measures of β-cell function. The prediction model developed was based on clinical criteria reflective of insulin sensitivity (BMI, waist circumference, and HDL cholesterol) and markers of autoimmunity. Thus, it was helpful in improving prediction of persistence of endogenous insulin reserve prospectively. The model performed better than provider diagnosis of type of diabetes (which nonetheless had an accuracy of 87%) or pancreatic autoantibodies alone in children with obesity >10 years of age, suggesting that adoption of such models may prove helpful in diagnostically challenging situations.
However, the cutoff value of C-peptide chosen to indicate persistence of endogenous insulin production was based on adult studies. Youth with type 2 diabetes typically have higher C-peptide concentrations even in those with positive autoimmune markers (11). Testing of the models to include a higher C-peptide threshold more typical of youth-onset type 2 diabetes, or stimulated C-peptide concentrations, may yield different results. Such an approach may better assist in clinical decision-making where withdrawal of insulin therapy and adoption of other therapeutic agents is contemplated. On the other hand, insulin secretion declines over time with the progressive loss of β-cell function in youth-onset type 2 diabetes (17,20). The interval at which C-peptide concentrations are obtained may also affect the performance of the models in the classification of the type of diabetes.
Addition of T1DGRS did not improve the model’s performance, likely because other features included in the model were more informative of diabetes type, particularly in Black and White-Hispanic youth. Of note, a genetic risk score for type 2 diabetes generated from a genome-wide association study in individuals of European ancestry, available in a subset of participants, did not improve the overall discrimination of the models. It would be important to test whether type 2 diabetes genetic risk scores developed from multiancestry cohorts (21) link to specific pathways in the pathogenesis of the disease and may prove helpful in strengthening future prediction models.
This study provides an important contribution to incorporation of models that use clinical criteria and biomarkers to facilitate more accurate diagnosis of the type of diabetes in youth, and it has the potential to support clinical decision-making. Additional studies are needed to validate the model in different populations. Future work may consider inclusion of genetic risk scores from multiancestry cohorts and testing of different models for prediction of stimulated C-peptide concentrations or the disposition index, a more rigorous marker of β-cell reserve that reflects β-cell function relative to insulin sensitivity. Such models may prove effective in the diagnosis of the subtypes of diabetes and the biologic pathways involved, which may lead to refinement of the approach to therapy.
See accompanying article, p. 2110.
Article Information
Funding. F.B. is supported by grant funding from the Agricultural Research Service of the U.S. Department of Agriculture, National Institute of Diabetes and Digestive and Kidney Diseases, and the Eunice Kennedy Shriver National Institute of Child Health and Human Development.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Handling Editors. The journal editors responsible for overseeing the review of the manuscript were John B. Buse and Kristen J. Nadeau.