The inflammatory marker SIRI acts as a statistical mediator between ApoB/ApoA1 ratio and atrial fibrillation, and combined biomarkers show higher discriminative ability for the presence of atrial fibrillation than single indicators

Objective This study aimed to investigate whether the systemic inflammation response index (SIRI) acts as a mediator between the apolipoprotein B/A1 ratio and atrial fibrillation, and assess the combined discriminative ability of the two indicators for identifying AF. Method In this retrospective single-center analysis, a final cohort of 638 individuals was enrolled for analysis. Among them, persistent atrial fibrillation (PeAF) was diagnosed in 212 cases and paroxysmal AF (PAF) in 212. The reference group comprised 214 age-matched inpatients without AF. To investigate the associations among SIRI, the ApoB/ApoA1, and AF risk, we used a multivariable analysis. Additionally, fully adjusted logistic regression models with restricted cubic spline (RCS) terms were employed to explore potential nonlinear relationships between SIRI, the ApoB/ApoA1, and AF presence. Spearman analysis was used to assess their monotonic relationship. Mediation analysis was conducted with ApoB/ApoA1 as the exposure and SIRI as the mediator, with AF as the outcome. ROC analyses were performed to evaluate the discriminative accuracy of ApoB/ApoA1, SIRI, and their combination for AF detection. DeLong's test was used to compare the AUCs. Sensitivity, specificity, and Brier score were calculated across multiple cutoff points, with the Youden index identifying the optimal threshold. Finally, we performed a subgroup analysis stratified by the use of lipid-lowering drugs to evaluate the potential effect modification by lipid-lowering therapy. Result The mediation analysis indicated that SIRI served as a partial statistical mediator of the inverse ApoB/ApoA1–AF relationship. The ROC analysis showed AUCs of 0.728 for ApoB/ApoA1, 0.760 for SIRI and 0.799 for the combined model. When contrasted with ApoB/ApoA1 alone, the ApoB/ApoA1 + SIRI combination achieved a significantly higher AUC (DeLong test, P < 0.001). Compared with SIRI alone, the integrated model also yielded a notably larger AUC (DeLong test, P = 0.002). The ApoB/ApoA1 + SIRI model exhibited the highest AUC, along with improved specificity, the lowest Brier score, and positive discriminative value. Besides, these findings are not affected by lipid-lowering drug use. However, given the cross-sectional design, these finding reflects statistical interdependence rather than a causal sequence. Conclusion Elevated SIRI and reduced ApoB/ApoA1 were significantly associated with higher odds of AF. The statistical mediation effect of SIRI suggests a potential interrelationship among these variables. The combined model outperformed all others in discriminative performance, offering a practical exploratory tool for patient stratification. However, these findings require independent external validation and prospective confirmation.