Selecting individuals for genetic testing of monogenic diabetes remains challenging in clinical practice. A study published in Diabetes Care developed a probability calculator that integrates clinical features and biomarkers to support the identification of monogenic diabetes.

Two prediction models were developed using a Bayesian recalibration mixture model approach. One model focused on early-insulin-treated individuals. This group served as a proxy for type 1 diabetes. The second model focused on not-early-insulin-treated individuals. This group served as a proxy for type 2 diabetes. Initial model development used case-control data. This included 594 individuals with monogenic diabetes and 597 with non-monogenic diabetes. The models initially used only clinical features.

The models were then recalibrated using population-based data from the Using pharmacogeNetics to Improve Treatment in Early-onset Diabetes (UNITED) study (n=1,299). Biomarkers were added at this stage. These included C-peptide and islet autoantibodies. External validation was performed in an independent population-based cohort of 1,025 individuals.

In early-insulin-treated individuals, the model that included biomarkers showed higher discrimination. The receiver operating characteristic area under the curve (ROCAUC) was 0.98 (95% credible interval 0.95–0.98). This was higher than the model using clinical features alone (ROCAUC 0.80; 95% credible interval 0.71-0.82; P<0.001). It was also higher than biomarkers alone (ROCAUC 0.96; 95% CI 0.95-0.97). In not-early-insulin-treated individuals, the model showed good discrimination (ROCAUC 0.86; 95% credible interval 0.85-0.88).

Both models showed good calibration. They also maintained strong discrimination in external validation. ROCAUC values were 0.98 (95% credible interval 0.98-0.98) for early-insulin-treated individuals and 0.92 (95% credible interval 0.900.93) for not-early-insulin-treated individuals. Using a ≥5% probability threshold to guide testing resulted in positive test rates of 16% to 19%. The calculator integrates clinical features and biomarkers. It provides a tool to support the selection of individuals for monogenic diabetes diagnostic testing. It is available as an online calculator for White European individuals diagnosed with diabetes at ≤35 years.