Anion gap trajectory patterns and prognosis in sepsis patients: A multicenter retrospective study

The anion gap (AG), a well-established indicator of metabolic acidosis, has been associated with adverse outcomes in critical illnesses, yet its temporal dynamics in sepsis remain poorly characterized. This retrospective cohort study analyzed patients with sepsis from the Medical Information Mart for Intensive Care IV (MIMIC-IV) v3.1 database and the eICU database to evaluate the prognostic value of AG trajectory patterns. Using latent class trajectory modeling, distinct AG trajectory profiles were identified, and their associations with mortality were assessed through Kaplan–Meier survival analysis and multivariable Cox proportional hazards regression, adjusting for key covariates including age, gender, Acute Physiology Score III, Charlson Comorbidity Index, model for end-stage liver disease, Obesity Admission Severity of Illness Score, sequential organ failure assessment, and systemic inflammatory response syndrome. In the MIMIC-IV (n = 4875) and eICU (n = 3863) cohorts, sepsis patients were categorized into 3 distinct AG trajectories using latent class trajectory modeling. Kaplan–Meier analyses confirmed significant survival differences across trajectory classes at all time points (all P  <.001). Multivariable Cox regression demonstrated that class 3 independently predicted higher mortality in MIMIC-IV: 28 days (adjusted Hazard ratio (H) = 1.64, P  <.001), 90 days (aHR = 1.60, P  <.001), and 365 days (aHR = 1.56, P  <.001). class 2 showed no significant associations. Consistent findings were observed in eICU, where class 3 was linked to increased 28-day mortality (aHR = 1.46, P  = .008). Three AG trajectories exist in sepsis, with class 3 identifying patients at high short- and long-term mortality risk. Validated across 2 databases, AG trajectory classification may enhance sepsis risk stratification.