A Machine Learning Model for Predicting Diabetic Nephropathy Based on TG/Cys-C Ratio and Five Clinical Indicators

Area of Science:

• Nephrology
• Diabetology
• Medical Informatics
• Machine Learning in Medicine

Background:

• Differentiating diabetic nephropathy (DN) from non-diabetic renal disease (NDRD) is clinically challenging.
• Accurate diagnosis is crucial for effective management of kidney disease in patients with diabetes mellitus.

Purpose of the Study:

• To develop and validate a machine learning model for the differential diagnosis of DN and NDRD.
• To identify key clinical features that distinguish DN from NDRD in patients with type 2 diabetes mellitus and proteinuria.

Main Methods:

• A predictive model was built using clinical data from 100 type 2 diabetes mellitus patients (50 DN, 50 NDRD) confirmed by renal biopsy.
• Recursive Feature Elimination (RFE) was employed for feature selection.
• The model was validated internally and externally using data from additional patient cohorts.

Main Results:

• The logistic regression (LR) model, incorporating TG/Cys-c ratio, SBP, diabetes duration, DR, HbA1c, and Hb, demonstrated strong performance.
• Internal validation yielded an AUC of 0.9305, accuracy of 0.8333, sensitivity of 0.8283, and specificity of 0.8701.
• External validation achieved superior results with an AUC of 0.9642, accuracy of 0.9455, sensitivity of 0.9615, and specificity of 0.9310.

Conclusions:

• The developed machine learning model, named PDN (Prediction of Diabetic Nephropathy), effectively differentiates DN from NDRD.
• This tool can assist clinicians in making accurate diagnoses and optimizing treatment strategies for patients with diabetic kidney disease.
• An online platform for the PDN model is available at http://cppdd.cn/service/PDN.