# Association between antithrombin levels and prognosis in patients with sepsis: a retrospective cohort study based on the MIMIC-IV and MIMIC-III databases

## TL;DR

Low antithrombin levels in sepsis patients are linked to higher mortality and organ failure, suggesting it could help identify high-risk patients.

## Contribution

Identifies specific antithrombin activity thresholds (55% overall, 64% in hypertensive patients) associated with increased mortality in sepsis.

## Key findings

- Patients with antithrombin <55% had significantly higher 28-day mortality, ICU mortality, and organ dysfunction.
- Hypertensive patients with antithrombin <64% showed elevated risks of mortality and complications.
- Antithrombin activity is a potential prognostic biomarker for sepsis risk stratification.

## Abstract

Sepsis is a critical determinant of mortality in critical patients. Antithrombin (AT) plays a pivotal role as a serine protease inhibitor with dual anticoagulant and anti-inflammatory functions, yet its precise role in prognostic stratification remains undefined. This study aimed to investigate the association between AT activity and clinical outcomes in sepsis and to identify critical prognostic thresholds.

We conducted a retrospective cohort study of 222 septic patients from the MIMIC-IV and MIMIC-III databases. AT activity was measured within the first 24 h following sepsis diagnosis, with the primary outcome defined as 28-day all-cause mortality. For preliminary description, AT activity was categorized into tertiles. The primary analysis utilized restricted cubic splines (RCS) to model the dose–response relationship and identify risk thresholds. Multivariable Cox regression models were employed to adjust for demographics, comorbidities, and SOFA score. Subgroup and survival analyses were performed to evaluate effect modification and visualize outcome differences across threshold-defined risk groups. To visually compare survival outcomes between patient groups defined by the RCS-derived risk thresholds, we generated Kaplan–Meier curves and employed log-rank tests.

A non-linear relationship between AT activity and 28-day mortality was identified, with a marked increase in risk observed below approximately 55% in the overall cohort. Patients with AT activity < 55% had significantly higher 28-day mortality (34.2% vs. 14.4%, p = 0.001), ICU mortality (33.3% vs. 9.0%, p < 0.001), and incidences of disseminated intravascular coagulation (DIC) (22.5% vs. 3.6%, p < 0.001) and acute kidney injury (AKI) (78.4% vs. 62.2%, p = 0.013). Subgroup analysis revealed a significant interaction with hypertension. In the hypertensive subgroup, a similarly elevated risk zone was observed below approximately 64% AT activity. Hypertensive patients below this level had markedly increased 28-day mortality (42.3% vs. 9.62%, p < 0.001), ICU mortality (38.5% vs. 5.77%, p < 0.001), and incidences of DIC (19.2% vs. 1.92%, p < 0.001).

Reduced AT activity was significantly associated with higher mortality and organ dysfunction in sepsis. Risk thresholds were observed at approximately 55% for the overall cohort and 64% among hypertensive patients. Patients below these levels exhibited significantly increased mortality and higher incidences of DIC and AKI. These findings support AT activity as a prognostic biomarker for risk stratification and highlight its potential to inform future management strategies for high-risk patients.

The online version contains supplementary material available at 10.1186/s40560-026-00862-x.