Development of a machine learning-based nomogram for predicting atherosclerosis risk in patients with systemic lupus erythematosus

Author information

CER19466
Full Papers

purchase article

Received: 28/10/2025
Accepted : 02/03/2026
In Press: 25/03/2026

Abstract

OBJECTIVES:
To develop a prediction model and construct a nomogram via machine-learning algorithms for estimating the risk of atherosclerosis in patients with systemic lupus erythematosus (SLE).
METHODS:
The electronic medical records of SLE inpatients treated at the Affiliated Hospital of North Sichuan Medical College from 2013 to 2023 were collected. Following data cleaning and variable encoding, variables were initially screened using LASSO regression. Subsequently, the optimal model was selected, and SHAP plots were generated to identify the five most influential features. Finally, logistic regression was used for model construction and nomogram plotting. The model’s performance was evaluated using calibration curves, receiver operating characteristic (ROC) curves, and decision-curve analysis. A two-sided p<0.05 was considered statistically significant.
RESULTS:
The Adaboost model was identified as the optimal one. The key predictors of atherosclerosis in SLE patients were age, cystatin C, T lymphocytes, glucose and AFU. The nomogram showed extensive predictive coverage. The area under the ROC curve (AUC) with a 95% CI was 0.905 (0.880–0.930) in the training dataset and 0.890 (0.841–0.939) in the testing dataset. The model demonstrated strong applicability and significant clinical benefits, highlighting its excellent predictive performance and clinical value in assessing the atherosclerotic risk in SLE patients.
CONCLUSIONS:
This exploratory, hypothesis-generating single-centre study provides an internally validated, interpretable prediction approach for atherosclerosis in SLE. External and temporal validation in independent cohorts is required before clinical implementation.