# Sex-stratified early biomarker model identifies lactate as the key predictor of in-hospital deterioration in acute heart failure

## TL;DR

This study finds that lactate is the strongest early predictor of in-hospital deterioration in acute heart failure patients, with sex-specific differences in biomarker performance.

## Contribution

The study introduces a sex-stratified model identifying lactate as a superior early biomarker for in-hospital deterioration in acute heart failure.

## Key findings

- Lactate was the most robust predictor of in-hospital deterioration with an odds ratio of 9.38 per log10 increase.
- In men, NT-proBNP and lactate were significant predictors, while no biomarker reached significance in women.
- NLR predicted outcomes in the non-HFrEF subgroup, and model performance had an AUC of ∼0.71–0.73.

## Abstract

Early identification of patients at risk for deterioration during hospitalization for acute heart failure (AHF) is essential for guiding intensive monitoring and advanced therapies. Biomarkers such as NT-proBNP and troponin I are routinely used, yet their comparative prognostic performance—particularly when stratified by sex—remains uncertain. Other emerging biomarkers, including lactate and the neutrophil-to-lymphocyte ratio (NLR), have also been linked to adverse outcomes, but their value relative to established cardiac markers has not been clearly defined.

We conducted a retrospective cohort study using de-identified electronic medical records from 2010 to March 2025 at a tertiary care center. Patients aged ≥16 years with clinician-documented AHF and available admission biomarkers were eligible. The primary endpoint was a composite of in-hospital death, mechanical ventilation, extracorporeal membrane oxygenation (ECMO), or intra-aortic balloon pump (IABP). Broad and strict endpoints were examined in sensitivity analyses. Multivariable logistic regression models, sex-stratified analyses, and penalized regressions with bootstrap resampling were performed.

Among 143 eligible patients (81 men, 62 women), the primary endpoint occurred in 46.9%. In our cohort, women experienced a slightly higher crude rate of in-hospital deterioration compared with men (48.4% vs. 45.7%). Lactate was the most robust predictor across all models, with an odds ratio of 9.38 (95% CI 2.47–35.63; p = 0.001) per log10 increase and a clear dose–response (event rates 39.8%, 55.2%, and 85.7% across lactate strata ≤2, 2–4, and >4 mmol/L; p-trend = 0.002). In sex-stratified models, NT-proBNP (OR 2.87; p = 0.029) and lactate (OR 28.98; p = 0.003) were significant in men, while no biomarker reached significance in women. NLR predicted outcomes in the non-HFrEF subgroup. Model performance was modest (AUC ∼0.71–0.73) but calibration was good. Findings remained consistent in winsorized and bootstrap sensitivity analyses.

In this single-center AHF cohort, lactate emerged as the most consistent early biomarker associated with in-hospital deterioration, with stronger prognostic performance than the other evaluated cardiac markers. Sex-stratified and phenotype-specific findings (NT-proBNP and lactate in men, NLR in non-HFrEF) were exploratory and did not show significant sex–biomarker interaction. These results support incorporating lactate into early risk stratification and highlight the need for larger multicenter validation studies.