Comparative analysis of six inflammatory markers in epilepsy severity: SIRI shows the strongest association

Abstract

Objective biomarkers for assessing epilepsy severity are lacking. This study compared the ability to identify higher NHS3 scores using six peripheral blood parameters — including the systemic inflammatory response index (SIRI) and mean platelet volume (MPV) — and evaluated whether combining SIRI and MPV improves performance. This retrospective cross-sectional study consecutively enrolled 364 hospitalized patients with epilepsy at a single center (December 2023–May 2025). Higher NHS3 scores were defined as a National Hospital Seizure Severity Scale (NHS3) score ≥ 9 (the sample median). Spearman correlation, receiver operating characteristic (ROC) analysis with DeLong tests, and multivariable logistic regression were performed. Six markers were examined: SIRI, NLR (neutrophil-to-lymphocyte ratio), MLR (monocyte-to-lymphocyte ratio), SII (systemic immune-inflammation index), PAR (platelet-to-albumin ratio), and MPV. SIRI showed the strongest positive correlation with NHS3 score ( r  = 0.402, P <  0.001), while MPV showed a significant negative correlation ( r = -0.377, P <  0.001). For identifying higher NHS3 scores, SIRI achieved an AUC of 0.685, indicating a nominally higher performance than NLR, MLR, SII, and PAR (nominal P <  0.05 for all comparisons). MPV yielded an AUC of 0.675, comparable to SIRI ( P  = 0.770) and not significantly different from PAR ( P =  0.017). The combined SIRI + MPV model achieved an AUC of 0.719. However, DeLong’s test indicated that this combined model did not provide a statistically significant improvement over SIRI alone ( P  = 0.120). In separate multivariable logistic regression models adjusting for clinical covariates, both SIRI (OR = 1.51, 95% CI: 1.09–2.09, P =  0.013) and MPV (OR = 0.70, 95% CI: 0.57–0.86, P  = 0.001) remained independently associated with higher NHS3 scores. Among the six markers, SIRI showed the strongest positive correlation with NHS3 scores ( r  = 0.402) and the highest AUC (0.685). MPV exhibited a significant negative correlation ( r = -0.377, AUC = 0.675), with its performance comparable to that of SIRI ( P  = 0.770). In adjusted multivariable models, SIRI (OR = 1.51, P  = 0.013) and MPV (OR = 0.70, P  = 0.001) remained independently associated with higher scores. These findings suggest that systemic inflammation, reflected by SIRI, and platelet activity, reflected by MPV, may represent complementary pathways associated with epilepsy severity. Given the cross-sectional design, these results are hypothesis-generating and warrant prospective validation.