Predictive value of systemic inflammation response index for atherosclerotic cardiovascular disease risk in patients with hypercholesterolemia: a machine learning study with dual-cohort validation.

BACKGROUND: Residual cardiovascular risk persists in patients with hypercholesterolemia despite lipid-lowering therapy, underscoring the importance of inflammation in ASCVD development. This study evaluated the relationship between Systemic Inflammation Response Index (SIRI) (a composite biomarker derived from neutrophil, monocyte, and lymphocyte counts) and ASCVD in patients with hypercholesterolemia. And to develop an interpretable machine learning (ML) model for predicting ASCVD risk in patients with hypercholesterolemia. METHODS: This study utilized data from the National Health and Nutrition Examination Survey (2001-2018), including a total of 6,645 patients with hypercholesterolemia. Additionally, a independent external cohort of 357 patients from Guizhou Provincial People's Hospital served as the validation cohort. Then, we used ML to analyze the effect of SIRI on ASCVD in patients with hypercholesterolemia and established four risk prediction models. Area under the receiver operating characteristic curve (AUC) was used to evaluate model performance. Shapley Additive Explanations (SHAP) were applied for model interpretation, and a web-based application was developed for clinical use. RESULTS: Our findings indicated that SIRI is consistently associated with ASCVD in hypercholesterolemia patients in both cohorts. SIRI and other seven features were used to construct ML models. The XGBoost model achieved an AUC of 0.8001 in the internal validation cohort and 0.7030 in the external cohort. The model retained strong clinical relevance. SHAP analysis highlighted elevated SIRI levels as an important predictor of ASCVD risk in patients with hypercholesterolemia. The inclusion of novel inflammatory markers such as SIRI enhanced the model's discriminative capability. CONCLUSIONS: The study findings revealed that high SIRI levels were an independent risk factor for ASCVD in patients with hypercholesterolemia. In addition, this study constructed the first interpretable ML model combined with SIRI for ASCVD prediction in hypercholesterolemia patients. The model demonstrated acceptable performance and moderate generalizability. While its external specificity was limited, the model may still serve as a useful risk stratification aid to support early identification of high-risk individuals with hypercholesterolemia.