Association between early intravenous magnesium sulfate administration and mortality according to systemic immune-inflammation index in critically ill patients with acute heart failure: A retrospective cohort study

Background Magnesium sulfate is commonly administered in acute heart failure (AHF). We investigated the association between early intravenous magnesium sulfate and mortality across baseline systemic immune-inflammation index (SII) strata in critically ill patients with AHF. Methods We conducted a retrospective study using the MIMIC-IV database of critically ill patients with AHF. Early intravenous magnesium sulfate was defined as treatment initiated from 6 h before to 24 h after intensive care unit admission. SII was categorized into quartiles. The primary and secondary outcomes were 30-day and 90-day all-cause mortality. Multivariable Cox regression, restricted cubic spline, and sensitivity analyses were performed. Results Among 4293 patients, 1561 (36.4%) received intravenous magnesium sulfate. Magnesium sulfate administration was associated with lower 30-day (adjusted hazard ratio [aHR] 0.68, 95% CI 0.56–0.82) and 90-day mortality (aHR 0.74, 95% CI 0.63–0.86). For 30-day mortality, multivariable Cox regression demonstrated significant associations in the lowest (aHR 0.64, 95% CI 0.43–0.95) and highest (aHR 0.52, 95% CI 0.36–0.75) SII quartiles, whereas no significant associations were identified in the intermediate quartiles. Comparable patterns were observed for 90-day mortality. Restricted cubic spline analyses showed visual separation between treatment groups at the extremes of the SII distribution, although formal interaction testing was not significant. Sensitivity analyses consistently reproduced the association in the highest SII quartile. Conclusions Stratified analyses suggested that the association between early magnesium sulfate administration and lower mortality was most evident in the highest SII quartile and less consistent in the lowest quartile. These findings are hypothesis-generating and require prospective validation. Trial registration Not applicable.