HAS-BLED vs ORBIT: bleeding risk scores in atrial fibrillation

What HAS-BLED and ORBIT each count

HAS-BLED and ORBIT overlap on age, kidney function, bleeding history and antiplatelet drugs, but differ in what else they count and how they weigh it. HAS-BLED, from Pisters, Lip and colleagues in Chest in 2010, gives 1 point each for nine items: uncontrolled hypertension, abnormal kidney function, abnormal liver function, previous stroke, bleeding history or predisposition, labile INR, age over 65, drugs (antiplatelets or NSAIDs) and alcohol excess, for a total of 0 to 9. The ORBIT bleeding score, from O'Brien and colleagues in the European Heart Journal in 2015, has five items with weights: age 75 or older (1), reduced hemoglobin or hematocrit or a history of anemia (2), bleeding history (2), insufficient kidney function (1) and treatment with an antiplatelet (1), for a total of 0 to 7. HAS-BLED includes several factors that can change, such as blood pressure, INR control and alcohol; ORBIT is built mostly from factors that cannot.

How the ORBIT bleeding score works

The ORBIT bleeding score was derived from 7,411 patients taking oral anticoagulants in the Outcomes Registry for Better Informed Treatment of Atrial Fibrillation (ORBIT-AF), enrolled at 176 US sites and followed for a median of 2 years, with a major bleeding rate of 4.0 per 100 person-years. O'Brien and colleagues kept the five predictors with the strongest link to major bleeding and weighted them by that strength. Reduced hemoglobin means under 13 g/dL in men or under 12 g/dL in women, reduced hematocrit means under 40% in men or under 36% in women, and insufficient kidney function means an eGFR under 60 mL/min/1.73 m². The total sorts patients into three groups. A score of 0 to 2 is low risk, with 2.4 major bleeds per 100 patient-years in the derivation cohort; a score of 3 is medium risk, with 4.7; and a score of 4 to 7 is high risk, with 8.1.

Which score predicts bleeding better?

HAS-BLED and ORBIT predict major bleeding about equally well when pooled, and both discriminate only modestly. In the ORBIT derivation paper, the five-item ORBIT score had a C-index of 0.67 in ORBIT-AF. In an external test in the ROCKET-AF trial population, it discriminated similarly to HAS-BLED and ATRIA but was markedly better calibrated, meaning its predicted bleeding rates matched the observed rates more closely; the authors concluded it predicted major bleeding better than HAS-BLED. Later studies did not settle the question. A 2023 meta-analysis by Liu and colleagues in Frontiers in Cardiovascular Medicine pooled 17 studies of anticoagulated patients with atrial fibrillation: the pooled C-index for major bleeding was 0.63 for ORBIT and 0.61 for HAS-BLED, with no statistical difference, and the two were also equal for clinically relevant and intracranial bleeding. Results were the same whether patients took vitamin K antagonists or direct oral anticoagulants. A C-index in the low 0.60s means many patients who bleed score low on either tool.

Which guideline uses which

European and UK guidance have taken different routes on bleeding scores. The 2020 European Society of Cardiology (ESC) atrial fibrillation guideline recommended HAS-BLED, using a score of 3 or more to flag high bleeding risk for earlier and more frequent review, and it said bleeding risk alone should not decide against anticoagulation without an absolute contraindication. The 2021 NICE guideline on atrial fibrillation (NG196) recommends ORBIT for assessing bleeding risk when starting or reviewing anticoagulation, a choice the Liu meta-analysis cites as its reason for re-testing the two scores. The 2024 ESC guideline moved away from scores for this decision: it advises against using bleeding risk scores to decide on starting or withdrawing anticoagulation, and asks clinicians to find and manage modifiable bleeding risk factors instead. That favours HAS-BLED's items as a checklist, because uncontrolled blood pressure, labile INR, alcohol and interacting drugs can be changed, while ORBIT's age, anemia and bleeding history mostly cannot.

Worked examples: same patient, two scores

Because the two scores count different things, they can agree for one patient and split for another. Take a 78-year-old woman with a hemoglobin of 11.5 g/dL, an eGFR of 52, no previous bleeding, controlled blood pressure and a daily antiplatelet drug. ORBIT gives 1 for age, 2 for low hemoglobin, 1 for kidney function and 1 for the antiplatelet, a score of 5 and the high-risk group. HAS-BLED gives 1 for age, 1 for anemia as a bleeding predisposition and 1 for the drug, a score of 3, also high risk; her creatinine is below the 200 µmol/L kidney cut-off, so that item stays at 0. Now take a 70-year-old man with a systolic blood pressure of 170 mmHg, alcohol excess and a labile INR, but normal hemoglobin and kidney function. HAS-BLED scores 4, high risk, while ORBIT scores 0, low risk. His bleeding risk rests on factors HAS-BLED counts and ORBIT leaves out, and most of them can be changed.

Score either patient with the HAS-BLED calculator; the eGFR for ORBIT's kidney item comes from the eGFR calculator.