Predictive model for sarcopenia in chronic kidney disease: a nomogram and machine learning approach using CHARLS data

Area of Science:

• Nephrology
• Geriatrics
• Biostatistics

Background:

• Sarcopenia is a common complication in chronic kidney disease (CKD) patients, leading to adverse clinical outcomes.
• Early identification and management of sarcopenia are crucial for improving patient prognosis in CKD.

Purpose of the Study:

• To develop and validate a predictive model for sarcopenia risk in individuals with CKD.
• To identify key predictors of sarcopenia in the CKD population using machine learning techniques.

Main Methods:

• Utilized data from the China Health and Retirement Longitudinal Study (CHARLS) with 1,092 CKD patients.
• Employed LASSO regression and logistic regression for predictor identification, constructing a nomogram for risk prediction.
• Applied machine learning algorithms, including Gradient Boosting Machine (GBM) with Bayesian optimization, for model development and validation using ROC and DCA.

Main Results:

• Identified age, waist circumference, LDL-C, HDL-C, triglycerides, and diastolic blood pressure as significant predictors of sarcopenia.
• The optimized GBM model achieved high predictive accuracy with an AUC of 0.933 in the training set and 0.932 in the validation set.
• SHAP analysis highlighted age and waist circumference as the most influential factors in sarcopenia prediction.

Conclusions:

• The developed nomogram and GBM model offer reliable tools for predicting sarcopenia risk in CKD patients.
• These models can aid clinicians in early risk assessment and effective management strategies for sarcopenia in the CKD population.