Admission Biomarker Risk Stratification

> Research explainer: This briefing examines verified primary research published 55 days before the briefing date. It is not a same-day research update and does not provide medical advice.

Evidence

The study screened 512 consecutive adult hospitalizations for acute heart failure at a tertiary care center in Vietnam between January and August 2021. Of these, 478 records had sufficient baseline data, and 430 patients with ascertainable 1-year vital status formed the analytic cohort. During the year after the index hospitalization, 84 of 430 patients died, a reported mortality proportion of 19.5%. [pmid:42384715]

The investigators compared admission NT-proBNP, analyzed as its natural logarithm, with AHEAD score categories for association with 1-year all-cause mortality. AHEAD is a five-component score incorporating atrial fibrillation, anemia, age over 70 years, renal dysfunction, and diabetes mellitus. The study grouped scores as 0–1, 2, and 3 or higher. [pmid:42384715]

Admission ln(NT-proBNP) was strongly associated with mortality after adjustment for admission covariates: adjusted hazard ratio 2.63 per 1-unit increase in ln(NT-proBNP), with a 95% confidence interval of 2.05 to 3.37. The authors described this as approximately a hazard ratio of 1.95 for each doubling of NT-proBNP. [pmid:42384715]

AHEAD categories also showed univariable associations with mortality. Relative to scores of 0–1, the reported univariable hazard ratios were 1.95 for score 2 and 3.61 for scores of 3 or higher. After adjustment for ln(NT-proBNP) and admission covariates, these estimates were attenuated to 1.03 and 1.81, respectively. [pmid:42384715]

For discrimination, the reported Harrell C-index was 0.758 for admission ln(NT-proBNP) and 0.608 for AHEAD categories. The combined model improved discrimination and reclassification compared with AHEAD alone, but adding AHEAD to ln(NT-proBNP) did not materially improve discrimination relative to ln(NT-proBNP) alone; the reported delta C-index was 0.000. [pmid:42384715]

The investigators also divided NT-proBNP values into cohort-specific tertiles. The highest tertile, above 6,385 pg/mL, identified groups described as high risk regardless of AHEAD category. This is a descriptive stratification result from this cohort rather than a universally validated threshold. [pmid:42384715]

Analysis — Biomarker versus score

The central comparison is not whether AHEAD had any prognostic signal, but whether it added meaningful discrimination once admission ln(NT-proBNP) was available. In this dataset, the answer was limited: AHEAD categories were associated with mortality before adjustment, yet their estimates weakened after accounting for the biomarker and admission covariates. The higher C-index for ln(NT-proBNP) than for AHEAD supports the same interpretation at the model-performance level. The combined model’s advantage over AHEAD alone should therefore not be read as evidence that the score improved on NT-proBNP; the reported delta C-index versus NT-proBNP alone was zero. AHEAD may still organize information about comorbidity burden, because its components cover clinical characteristics distinct from a biomarker measurement. But this study’s reported results support a narrower conclusion: for 1-year all-cause mortality discrimination in its analytic cohort, admission ln(NT-proBNP) was the stronger of the two measures, and the score did not materially increase discrimination beyond it. [pmid:42384715]

The tertile finding adds a practical interpretive layer without changing that comparison. Patients in the top cohort-defined NT-proBNP tertile were classified as high risk across AHEAD categories, suggesting that the biomarker level dominated the reported stratified pattern. Yet the analysis does not establish that a particular measurement should determine management, nor does it test whether using either approach changes outcomes. Its contribution is prognostic description within hospitalized acute heart failure, using admission data and a 1-year all-cause mortality endpoint. [pmid:42384715]

Limitations

This was a retrospective observational study from one tertiary care center in Vietnam, so its estimates and model performance may not transfer to other hospitals, health systems, or patient populations. The analytic cohort excluded 34 records with incomplete baseline data and 48 patients whose 1-year vital status could not be ascertained; those exclusions can affect how fully the final cohort represents all screened admissions. [pmid:42384715]

The outcome was 1-year all-cause mortality. The study does not establish why deaths occurred, whether NT-proBNP or AHEAD caused risk, or whether changing care based on either measure would alter outcomes. The threshold above 6,385 pg/mL was a cohort-specific tertile cut point, not an externally validated clinical cutoff. The authors state that external validation is warranted. Accordingly, these findings should be read as evidence about comparative prognostic performance in this defined cohort, not as medical advice, a treatment-effect estimate, or a patient-specific conclusion. [pmid:42384715]