Nonlinear correlation between lactate levels and 28-day all-cause mortality in patients with sepsis complicated by acute respiratory distress syndrome (ARDS): a retrospective study based on the MIMIC-IV database.

BACKGROUND

Sepsis-associated acute respiratory distress syndrome (ARDS) is a life-threatening condition with high mortality. Although lactate is a widely used prognostic marker in critical illness, its association with mortality in this specific patient population remains inconsistent, and the nature of this relationship has not been systematically characterized. This study aimed to systematically investigate the association between admission serum lactate levels and 28-day all-cause mortality in this cohort and to derive lactate-based risk strata for clinical prognosis.

METHODS

A retrospective cohort of 3,214 septic patients with ARDS was identified from the Medical Information Mart for Intensive Care-IV (MIMIC-IV, v3.0) database. Patients were stratified into four groups by lactate quartiles. Restricted cubic spline (RCS) analysis, the Boruta algorithm for covariate selection, multivariate Cox regression, and propensity score matching (PSM) were applied to explore the lactate-mortality association and control confounders.

RESULTS

RCS analysis confirmed a significant U-shaped, nonlinear relationship between admission lactate levels and 28-day mortality (P for nonlinearity < 0.001). In quartile-based analyses using multivariable Cox regression, mortality was significantly lower in the lowest lactate group (Group 1: <1.50 mmol/L) compared with higher quartiles (Group 2: 1.50-2.25 mmol/L; Group 3: 2.25-4.05 mmol/L; Group 4: ≥4.05 mmol/L) (19.4%, 26.1%, 28.7%, and 25.8%, respectively; overall P = 0.001). After full covariate adjustment, compared with Group 1, mortality risk was significantly elevated in Group 2 (adjusted hazard ratio [aHR] = 1.25, 95% CI: 1.01-1.54; P = 0.039) and Group 3 (aHR = 1.39, 95% CI: 1.13-1.72; P = 0.002), but not in Group 4 (aHR = 0.88, 95% CI: 0.70-1.11; P = 0.295). Kaplan-Meier survival analysis further demonstrated the lowest survival probability in Group 2 and Group 3, with convergence of survival curves between Group 1 and Group 4 occurring around day 14. After propensity score matching to control for confounders, the non-linear relationship remained robust, and no significant mortality difference was observed between Group 1 and Group 4, further supporting the U-shaped association.

CONCLUSIONS

This study reveals a U-shaped relationship between lactate and 28-day mortality in sepsis-associated ARDS, providing a significant refinement to the conventional linear prognostic model. The derived lactate thresholds provide a novel framework for risk stratification and may help guide phenotype-specific clinical management.