Major haemorrhage in people with heart failure and atrial fibrillation: community-based cohort study

Problem

Heart failure (HF) is a risk factor for major haemorrhage but is not included in key anticoagulation bleeding risk prediction scores for atrial fibrillation (AF), such as ORBIT or HAS-BLED. People with HF also typically have a poor prognosis and traditional analytical methods may over-estimate the prognostic significance of a variable if the competing risk of death is not accounted for. We aimed to report the relative risk of first major haemorrhage in people with HF and AF, compared to people with AF without HF, accounting for the competing risk of all-cause mortality.

Approach

Primary care cohort study of 2,178,162 patients aged ≥45 years in the English Clinical Practice Research Datalink (2000-2018), linked to secondary care data from Hospital Episode Statistics and mortality data from the Office for National Statistics. We excluded people with a history of major haemorrhage prior to the study index date. We conducted competing risks and landmark analyses alongside traditional survival analysis.

Findings

We included 60,270 people with HF and AF, 79,461 with HF only and 126,251 with AF only. Over 7.56 years median follow-up, 72,196 patients (3.3%) had a first major haemorrhage and 276,319 (12.7%) died. The cumulative incidence function of major haemorrhage in people with HF and AF was 2.25% (95%CI 1.96-2.58) at 1-year follow-up, 8.22% (95%CI 7.65-8.93) at 5-year follow-up and 12.0% (95%CI 11.3-12.9) at 10-year follow-up (Figure 1). By comparison among people with AF without HF the respective cumulative probability was 1.45% (95%CI 1.31-1.61) at 1-year, 6.47% (95%CI 6.15-6.82) at 5-year and 11.2% (95%CI 10.8-11.7) at 10-year follow-up (Figure 1). Incidence rates of intracranial and gastrointestinal bleeding per 1,000 person years at risk were also highest in people with HF and AF. In a fully-adjusted Cox model, the hazard ratio for major haemorrhage was higher among people with HF and AF (HR 2.52, 95%CI 2.44-2.61) than people with AF without HF (HR 1.87, 95%CI 1.82-1.92). HF remained associated with an increased relative risk of major haemorrhage in a sub-group analysis of people with AF who were prescribed an oral anticoagulant (fully adjusted Cox model HR 1.68, 95%CI 1.59-1.78). However, in a Fine and Gray model accounting for the competing risk of death, the hazard of major haemorrhage was similar for people with AF without HF (HR 1.82, 95%CI 1.77-1.87) or HF and AF (HR 1.71, 95%CI 1.66-1.78).

Consequences

People with HF and AF are at increased risk of major haemorrhage compared to those with AF without HF. Current bleeding prediction scores may under-estimate this risk and so miss the opportunity to treat modifiable bleeding risk factors. However, people with HF and AF often have a poor prognosis so may be exposed to this higher bleed risk for a relatively short window and prognosis is important to consider when assessing bleeding risk. Future research could seek to develop bleed risk prediction scores using competing risks methods.