Medicine (Baltimore). 2026 Jun 26;105(26):e49526. doi: 10.1097/MD.0000000000049526.
ABSTRACT
The association between early magnesium sulfate administration and outcomes in sepsis remains incompletely understood, particularly in relation to differences in systemic immune-inflammatory status. In this retrospective cohort study using Medical Information Mart for Intensive Care IV (version 3.1), adult critically ill patients with sepsis were included. Systemic immune-inflammation index (SII) was calculated from platelet, neutrophil, and lymphocyte counts obtained within a peri-intensive care unit window (6 hours before to 24 hours after intensive care unit admission). Early magnesium administration was defined as intravenous magnesium sulfate initiated within the same window. The primary outcome was 30-day all-cause mortality. Propensity score matching was performed to balance baseline characteristics. Associations were examined overall and across SII quartiles using stratified multivariable Cox proportional hazards models, with nonlinear relationships explored using restricted cubic splines. Among 12,087 patients, 6039 received early intravenous magnesium. After propensity score matching, 3396 matched pairs were analyzed. Early magnesium administration was associated with lower 30-day mortality overall (15.8% vs 18.3%, P = .007). In stratified analyses, adjusted hazard ratios were 0.66 (95% confidence interval: 0.51-0.87), 0.65 (0.49-0.85), 0.75 (0.59-0.94), and 0.85 (0.70-1.04) across quartiles 1 to 4. Restricted cubic spline analyses showed descriptive attenuation at higher SII values (P for interaction = .208), without evidence of a statistically definitive interaction. Early magnesium administration was associated with lower 30-day mortality in critically ill patients with sepsis. Although formal interaction testing did not reach statistical significance, adjusted hazard ratios showed a descriptive gradient of attenuation across increasing SII quartiles, directionally consistent across sensitivity analyses. These findings are hypothesis-generating and require prospective validation.
PMID:42363536 | DOI:10.1097/MD.0000000000049526
