Integrating multiple hematologic inflammatory indices into a principal component analysis–derived inflammatory burden score for coronary artery disease

BACKGROUND: Complete blood count-derived inflammatory indices have been associated with coronary artery disease (CAD), but their substantial overlap may limit interpretability when assessed individually. We used principal component analysis (PCA) to integrate six commonly used inflammatory indices into a composite Inflammatory Burden Score (IBS) and evaluated its association with angiographically confirmed CAD alongside the Prognostic Nutritional Index (PNI). METHODS: This retrospective cross-sectional study included 1,261 individuals (477 with CAD and 784 without CAD) who underwent invasive coronary angiography at a tertiary referral center. PCA was applied to six standardized inflammatory indices, and the first principal component was retained as the IBS. Hierarchical logistic regression models were constructed to evaluate the incremental contribution of IBS and PNI beyond conventional cardiovascular risk factors. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC), calibration metrics, and likelihood-ratio testing. XGBoost and SHAP analyses were performed as complementary explainable machine-learning approaches. RESULTS: The first principal component explained 68.1% of the total variance across inflammatory indices, with all variables contributing positively to the composite score. IBS was significantly higher among participants with CAD (p < 0.001). In univariable analyses, both IBS (OR 1.11, 95% CI 1.05-1.17, p < 0.001) and PNI (OR 0.05, 95% CI 0.02-0.14, p < 0.001) were associated with CAD. After adjustment for conventional cardiovascular risk factors, IBS (OR 1.12, 95% CI 1.06-1.19) and PNI (OR 0.06, 95% CI 0.02-0.20) remained independently associated with CAD. Addition of either IBS or PNI significantly improved model fit (both likelihood-ratio test p < 0.001). The AUC increased from 0.727 in the conventional risk-factor model to 0.734 after inclusion of IBS and to 0.737 after inclusion of PNI. The XGBoost model achieved an AUC of 0.802, while SHAP analysis identified age and sex as the dominant predictors, followed by IBS and PNI. CONCLUSIONS: In this single-center cross-sectional cohort, a PCA-derived inflammatory burden score and the Prognostic Nutritional Index were independently associated with angiographically confirmed CAD beyond conventional cardiovascular risk factors. Both measures contributed similarly in explainable machine-learning analyses. These findings are exploratory and require prospective external validation before clinical application.