Clinical background
The decision on oral anticoagulation in atrial fibrillation rests on a trade-off between the risk of thromboembolic stroke and the risk of major bleeding. Thromboembolic risk is captured by CHA₂DS₂-VASc, but bleeding risk is at least as complex: it depends on renal and liver function, blood pressure control, concomitant medication, age and previous bleeding, and these factors interact in ways that are hard to survey clinically. HAS-BLED was constructed to quantify the one-year risk of major bleeding in patients with atrial fibrillation and thereby to make the trade-off more systematic [1].
The score is not an instrument for deciding whether a patient should receive anticoagulation. Instead it should flag patients at increased bleeding risk in whom modifiable factors can be addressed and follow-up intensified. A patient with a high bleeding risk often has a high thromboembolic risk as well, and it is then that the net benefit of anticoagulation is greatest [3,4].
Calculating HAS-BLED
Each variable contributes 0 or 1 point, and the total score runs from 0 to 9:
The variables are defined as follows:
- Hypertension: uncontrolled hypertension, defined as a systolic blood pressure >160 mmHg
- Abnormal renal function: dialysis, kidney transplantation or a serum creatinine >200 µmol/L (>2.26 mg/dL)
- Abnormal liver function: cirrhosis, bilirubin >2× the upper limit of normal, or transaminases >3× the upper limit of normal
- Stroke: previous stroke in the history
- Bleeding: a history of bleeding or a bleeding predisposition, defined as previous major bleeding or anaemia
- Labile INR: time in therapeutic range (TTR) <60%, relevant only with vitamin K antagonists
- Age >65 years
- Drugs: concomitant antiplatelet agent or NSAID
- Alcohol ≥8 units per week
The derivation cohort consisted of 3,978 patients with atrial fibrillation in the Euro Heart Survey on AF, a prospective European observational study. Over one year of follow-up there were 53 major bleeds (1.5%). The outcome modelled was major bleeding, defined as intracranial bleeding, bleeding requiring hospital admission, a fall in haemoglobin >2 g/dL and/or transfusion [1].
Interpretation in practice
| Score | Risk category | Clinical action |
|---|---|---|
| 0–2 | Low to moderate bleeding risk | Routine follow-up; continue anticoagulation as indicated |
| ≥3 | High bleeding risk | Address modifiable factors, consider closer follow-up and INR monitoring with vitamin K antagonists |
A score of 3 or higher should prompt an active search for, and correction of, modifiable risk factors: optimise blood pressure control, review concomitant antiplatelet agents or NSAIDs, improve the TTR in patients on warfarin, and address alcohol intake. The score must not be used as a reason to avoid anticoagulation. For most patients with atrial fibrillation the benefit of stroke prevention outweighs the bleeding risk, particularly when the factors driving the score up are themselves modifiable [3,4].
An important addition is that the risk is not static. A patient with a low score today may cross the threshold by growing older, developing impaired renal function or being started on an antiplatelet agent. HAS-BLED should therefore be recalculated at every clinical contact, not only when anticoagulation is started [3,4].
Validation and performance
In the derivation cohort, HAS-BLED achieved a c-statistic of 0.72 for major bleeding over one year [1]. In external cohorts, discrimination has fallen. A meta-analysis of 39 studies found a pooled c-statistic of 0.63 (95% CI 0.61–0.65) for major bleeding, with no significant difference between patients treated with vitamin K antagonists (c-statistic 0.63, CI 0.61–0.65) and direct oral anticoagulants (c-statistic 0.63, CI 0.59–0.67) [3]. A separate meta-analysis focused specifically on the comparison between HAS-BLED and ORBIT (17 studies) confirmed that the two scoring systems performed equivalently, with a c-statistic of 0.61 (CI 0.59–0.63) for HAS-BLED versus 0.63 (CI 0.60–0.66) for ORBIT, with no statistical difference [4].
As regards calibration, an earlier meta-analysis (11 studies) found that HAS-BLED overpredicted bleeding risk in the low-risk group (risk ratio 1.16, CI 0.63–2.13, p = 0.64) but underpredicted it in the moderate-risk group (risk ratio 0.66, CI 0.51–0.86, p = 0.002) and the high-risk group (risk ratio 0.88, CI 0.70–1.10, p = 0.27) [2]. Calibration data have been inconsistent between studies in the later meta-analyses, and there is insufficient evidence to say that the score calibrates well in all populations [3,4].
In a prospective Spanish cohort of 1,433 patients with atrial fibrillation on rivaroxaban (the EMIR registry), HAS-BLED achieved an AUC of 0.770 (CI 0.693–0.847) for major bleeding, with better calibration than ORBIT in this DOAC-treated population [5].
In summary, HAS-BLED performs at least as well as ORBIT, ATRIA and HEMORR₂HAGES for bleeding risk prediction, and significantly better than CHADS₂ and CHA₂DS₂-VASc, which are stroke scores and not bleeding scores [2,3].
Limitations
Labile INR is not applicable to patients treated with direct oral anticoagulants. This is the most obvious limitation: the variable was constructed for vitamin K antagonists and has no counterpart for DOACs. The meta-analyses show, however, that the loss of this variable does not significantly affect predictive performance, possibly because the TTR correlates in part with other factors in the score, but it does mean that in practice HAS-BLED is reduced to 8 variables in DOAC-treated patients [3,4].
The score was derived in a cohort in which 1.5% experienced major bleeding over one year. In populations with a higher baseline bleeding risk — for example with severely impaired renal function, or advanced age with multiple comorbidities — the underlying risk is higher than the score directly reflects, and calibration deteriorates, particularly in the moderate- and high-risk bands [2,3].
The commonest misuse is to interpret a high score as a reason to withhold anticoagulation. That is plainly wrong: the score identifies patients who need extra attention and correction of modifiable factors, not patients who should not be anticoagulated. Bleeding risk is often highest in those who also have the highest thromboembolic risk, and it is there that the benefit of anticoagulation is greatest [3,4].
Another pitfall is to use CHADS₂ or CHA₂DS₂-VASc as a proxy for bleeding risk. These stroke scores have significantly poorer discrimination for bleeding outcomes than HAS-BLED and should not be used for that purpose [2,3].
References
- Pisters R et al. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the Euro Heart Survey. Chest 2010. PMID: 20299623
- Zhu W et al. The HAS-BLED score for predicting major bleeding risk in anticoagulated patients with atrial fibrillation: a systematic review and meta-analysis. Clin Cardiol 2015. PMID: 26418409
- Gao X et al. Diagnostic accuracy of the HAS-BLED bleeding score in VKA- or DOAC-treated patients with atrial fibrillation: a systematic review and meta-analysis. Front Cardiovasc Med 2021. PMID: 34881309
- Liu X et al. HAS-BLED vs. ORBIT scores in anticoagulated patients with atrial fibrillation: a systematic review and meta-analysis. Front Cardiovasc Med 2022. PMID: 36684554
- Esteve-Pastor MA et al. Predicting performance of the HAS-BLED and ORBIT bleeding risk scores in patients with atrial fibrillation treated with rivaroxaban: observations from the prospective EMIR registry. Eur Heart J Cardiovasc Pharmacother 2022. PMID: 36318457
