Natriuretic peptides stratify competing first events: bleeding, stroke, and death in atrial fibrillation with acute heart failure
J Yamamoto, K Nakamura, Y Enomoto, H Hayama, M Yamamoto, H Hara, Y Hiroi, H HaraAbstract
Background
In patients hospitalised for acute heart failure (AHF) with atrial fibrillation (AF), early hazards diverge between ischaemic stroke, major bleeding, and death. Clinicians need a simple safety-first signal to identify patients in whom bleeding or death is likely to occur before stroke when tailoring oral anticoagulation.
Purpose
To determine whether baseline natriuretic peptides (NP) (i) identify a "safety-first" trajectory where bleeding or death precedes stroke, (ii) discriminate this trajectory over time, and (iii) provide clinically interpretable information for risk-guided management.
Methods
We analysed 696 consecutive AHF–AF admissions from a multicentre registry (2014–2020) with available natriuretic peptide measurements. First events were classified as bleeding-first, stroke-first, or death-first under a competing-risk framework. The primary endpoint was safety-first (bleeding-first or death-first), with stroke-first treated as a competing event and no-event censored; secondary endpoints were the three components. NP were measured as B-type natriuretic peptide or N-terminal pro-B-type natriuretic peptide and log-transformed as NPlog = loge(NP + 1). Fine–Gray models estimated subdistribution hazard ratios (sHRs) per 1-unit NPlog, with 95% confidence intervals (CI). One-year cumulative incidence was also summarised across NP tertiles. Time-dependent area under the curve (AUC) at 1 and 3 years was calculated for safety-first, bleeding-first and stroke-first.
Results
Median age was 81 years [interquartile range 73–87]; 48.3% were women. Oral anticoagulation was prescribed at discharge in 81.6%, and 63.6% of treated patients received direct oral anticoagulants. Over a median follow-up of 512 days, the first event was bleeding in 82 patients, stroke in 38, and death in 134. Across NP tertiles, 1-year cumulative incidence increased for bleeding-first (7.3%, 9.4%, and 12.6%) and death-first (11.4%, 12.1%, and 15.5%), but not for stroke-first (4.3%, 5.1%, and 2.4%). Higher NPlog was independently associated with the safety-first endpoint: sHR per 1-unit NPlog 1.12 (95% CI 1.02–1.24, p=0.023). Component-specific sHRs were 1.02 (0.88–1.19, p=0.77) for bleeding-first, 1.10 (0.98–1.24, p=0.12) for death-first and 0.82 (0.60–1.11, p=0.20) for stroke-first, indicating that NP primarily track a bleed- or death-dominated pattern rather than ischaemic stroke. Discrimination of NPlog alone for the safety-first endpoint was modest but improved over time (AUC 0.56 at 1 year and 0.69 at 3 years); performance was similar for bleeding-first and poor for stroke-first.
Conclusions
In AF-complicated AHF, higher baseline NP primarily flag patients whose subsequent course is dominated by bleeding or death rather than stroke, providing a pragmatic safety-first signal to support cautious anticoagulation and prioritised bleed-mitigation and frailty-conscious care.FirstEvent_NPtertilesFor image description, please refer to the figure legend and surrounding text.FGforest_NPlogFor image description, please refer to the figure legend and surrounding text.