A clinical diagnostic model based on an eXtreme Gradient Boosting algorithm to distinguish type 1 diabetes

Area of Science:

• Endocrinology
• Diabetes Mellitus Research
• Computational Medicine

Background:

• Accurate early-phase classification of type 1 diabetes (T1DM) and type 2 diabetes (T2DM) is critical for personalized treatment strategies.
• Distinguishing T1DM from T2DM in newly diagnosed adults presents a clinical challenge, necessitating simpler diagnostic tools.

Purpose of the Study:

• To develop and validate a classification model for differentiating T1DM from T2DM in adults using accessible clinical variables.
• To identify key clinical predictors for accurate diabetes subtyping.

Main Methods:

• A cross-sectional study involving 15,206 adults with newly diagnosed diabetes in China.
• Diabetes classification based on postprandial C-peptide (PCP) and glutamic acid decarboxylase autoantibody (GADA) levels.
• Development of a diagnostic model using the eXtreme Gradient Boosting (XGBoost) algorithm, prioritizing variables by importance scores.

Main Results:

• The final model integrated age of onset, body mass index (BMI), and hemoglobin A1c (HbA1c), demonstrating superior performance.
• The model achieved an area under the receiver operating characteristic curve (ROC AUC) of 0.83, with a sensitivity of 0.77 and specificity of 0.76.
• Good calibration performance was observed, with an intercept of 0.02 and a slope of 0.90.

Conclusions:

• A classification model incorporating age of onset, BMI, and HbA1c effectively distinguishes T1DM from T2DM in adults.
• This model serves as a valuable tool for physicians to aid in diabetes subtyping and guide precise treatment decisions.