Association Between Anion Gap and Mortality in Critically Ill Patients with Atrial Fibrillation: A Propensity Score-Matched Study

Background Serum anion gap (AG) is associated with mortality in critical illnesses, yet its prognostic significance specifically in intensive care unit (ICU) patients with atrial fibrillation (AF) remains unclear. This study aimed to investigate the associations between AG and mortality in this high-risk population. Methods We identified critically ill patients with AF from the MIMIC-IV database and stratified them by AG tertiles. Outcomes included 28-day and 365-day mortality. Multivariable Cox regression, propensity score matching (PSM), and restricted cubic splines were employed to examine the association between AG and mortality. Survival differences were evaluated using Kaplan-Meier analysis. Subgroup analyses assessed the consistency of associations, and ROC analysis quantified the incremental predictive value of AG. Results Among 14,635 eligible patients, elevated AG was significantly associated with increased mortality both before and after propensity score matching. In fully adjusted models, each 1-unit increase in AG was associated with a 5% higher risk of 28-day mortality (HR 1.05, 95% CI 1.04–1.05) and a 4% increased risk of 365-day mortality (HR 1.04, 95% CI 1.03–1.05). Patients in the highest AG tertile had substantially increased mortality risk compared to the lowest tertile (28-day HR 2.19, 95% CI 1.90–2.52; 365-day HR 1.98, 95% CI 1.74–2.25). Consistent dose-response relationships were observed across all analytical methods. Subgroup analyses presented in forest plots demonstrated the robustness of this association across various clinical strata. Additionally, AG significantly improved the predictive accuracy of conventional illness severity scores. Conclusions Elevated AG is independently associated with increased 28- and 365-day mortality in critically ill patients with AF. AG provides significant incremental prognostic value to established risk assessment tools such as SOFA and OASIS scores, and may serve as a readily available biomarker for improving risk stratification in this population.