More calculation, limited gain: systemic immune-inflammation index versus simpler blood count–derived indices for predicting acute pancreatitis severity and short-term mortality (a systematic review and meta-analysis)

Blood-count-derived inflammatory indices are accessible candidates for early risk stratification in acute pancreatitis (AP), but whether the systemic immune-inflammation index (SII) provides superior stand-alone discrimination compared with simpler indices is uncertain. We evaluated SII for AP severity and short-term mortality and compared it directly with the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR) and systemic inflammation response index (SIRI). We systematically searched MEDLINE/PubMed, Embase, Scopus and Google Scholar through July 2026 for adult studies evaluating admission or first-24-hour SII for acute pancreatitis severity or mortality. Severity outcomes were stratified by Revised Atlanta and non-Atlanta definitions. Random-effects meta-analysis pooled AUCs, while threshold-based sensitivity and specificity were synthesised using bivariate models. Comparative discrimination between SII and other markers measured in the same cohorts was assessed using within-study AUC differences. Heterogeneity, risk of bias and certainty of evidence were evaluated using I², the Quality Assessment of Prognostic Accuracy Studies tool and GRADE principles, respectively. Twenty-five studies included 12,850 participants. Across 19 severity studies (11,534 participants), pooled SII AUCs were 0.663 (95% CI 0.514–0.786; I 2 = 66.4%) for Revised Atlanta severe AP, 0.724 (0.578–0.835; I 2 = 91.9%) for mild versus moderately severe/severe AP, and 0.713 (0.653–0.766; I 2 = 46.5%) for non-Atlanta severity definitions. In paired comparisons, SII showed similar discrimination to NLR and SIRI but higher discrimination than PLR (AUC difference + 0.064). Five general AP cohorts (918 participants; 71 deaths) yielded a pooled mortality AUC of 0.796 (0.591–0.914; I 2 = 66.8%), although between-study uncertainty was substantial. Mortality comparisons were sparse. Certainty of evidence was low for the comparison with PLR and very low for the other principal comparisons. SII showed definition-dependent and heterogeneous discrimination. It did not outperform NLR or SIRI, showed a modest low-certainty advantage over PLR for severity, and had unstable mortality accuracy. Given the high risk of bias across much of the evidence base, current evidence is insufficient to support SII as a stand-alone decision tool; any adjunctive role remains uncertain pending prospective validation of prespecified thresholds. PROSPERO, CRD420261464739.