ORBIT bleeding risk score in atrial fibrillation

Clinical background

The decision on anticoagulation in atrial fibrillation requires a trade-off between thromboembolic risk and bleeding risk. A thromboembolic risk score such as CHA₂DS₂-VASc identifies patients at high stroke risk, but gives no guidance on how large the bleeding risk is in that particular patient. The ORBIT bleeding risk score was developed to fill this gap: a simple, bedside-applicable tool that estimates the risk of major bleeding in patients who are already taking, or about to start, an oral anticoagulant for atrial fibrillation. The score is intended to be used alongside a thromboembolic risk score, not instead of one.

Calculating the ORBIT bleeding risk score

The score is the sum of five clinical variables, two of which carry double weight:

$$\text{ORBIT} = \mathbb{1}[\text{age} \geq 74] + 2 \cdot \mathbb{1}[\text{reduced Hb/Hct or anaemia}] + 2 \cdot \mathbb{1}[\text{major bleeding}] + \mathbb{1}[\text{eGFR} < 60] + \mathbb{1}[\text{antiplatelet}]$$

The variables and their definitions:

• Age ≥ 74 years: one point.
• Reduced haemoglobin/haematocrit or a history of anaemia: two points. The thresholds are Hb < 12 g/dL or haematocrit < 36% in women, Hb < 13 g/dL or haematocrit < 40% in men.
• A history of major bleeding (gastrointestinal or intracranial): two points.
• eGFR < 60 mL/min/1.73 m²: one point.
• Concomitant antiplatelet treatment: one point.

The score ranges from 0 to 7. The two most heavily weighted variables (reduced Hb and a history of bleeding) were also the strongest predictors in the derivation cohort, whereas age, impaired renal function and concomitant antiplatelet treatment had a weaker but statistically significant association with major bleeding [1].

The derivation cohort consisted of 7,411 patients with atrial fibrillation taking an oral anticoagulant at baseline in the ORBIT-AF registry, a prospective American cohort based in primary care and specialist practice at 176 centres, recruited between 2010 and 2012. The median age was 75 years and 42.4% were women. Treatment was warfarin in 93.5% and dabigatran in 6.5%. Over a median follow-up of 2 years there were 581 major bleeds (7.8%), corresponding to 4.0 events per 100 person-years. Major bleeding was defined by the ISTH criteria: fatal bleeding, symptomatic bleeding in a critical organ, or bleeding with a fall in Hb ≥ 20 g/L or a transfusion requirement of ≥ two units [1].

Interpretation in practice

In the derivation cohort, patients were divided into three risk categories on the basis of the total score:

A score of 0–2 does not mean that the bleeding risk is negligible, but it is considerably lower than at higher scores. At low risk, anticoagulation can generally be started or continued without the bleeding risk being a reason to withhold it, provided the thromboembolic risk warrants treatment. At medium or high risk, the modifiable risk factors on which the score rests should be addressed first: reduced haemoglobin should be investigated, concomitant antiplatelet treatment reconsidered and the eGFR followed. A high score must not automatically lead to withdrawal of anticoagulation, since the net clinical benefit usually outweighs the bleeding risk when the thromboembolic risk is increased. The score should rather form the basis of a structured discussion with the patient about risks and benefits, and for prioritising measures that can reduce the bleeding risk [1].

Validation and performance

In the derivation cohort the simple five-variable model had a c-index of 0.67 and the full model (12 variables) 0.69. In a sensitivity analysis for intracranial bleeding, the score performed with a c-index of 0.69 [1].

External validation was carried out in the ROCKET-AF trial, a randomised trial of 14,264 patients comparing rivaroxaban with warfarin. Discrimination was somewhat lower in this population for all the scores, but the pattern was the same: the full ORBIT model had the highest c-index, followed by the simple ORBIT score, ATRIA and HAS-BLED. Decisive was the calibration: ORBIT showed clearly superior calibration compared with HAS-BLED and ATRIA when set against the observed bleeding rates in ROCKET-AF. HAS-BLED showed relatively poor calibration for the low-risk categories and ATRIA for almost all risk groups [1].

In the RE-LY trial (18,113 patients, dabigatran 110 mg, 150 mg or warfarin) the superiority of the ORBIT score was confirmed. The c-index was 0.66 for major bleeding, 0.66 for life-threatening bleeding and 0.62 for intracranial bleeding, significantly better than HAS-BLED (differences in c-index 0.050, 0.053 and 0.048 respectively, all P < 0.05) [2]. ORBIT also showed the best calibration of the four scores compared.

In an Asian cohort of 961 patients with atrial fibrillation treated with direct oral anticoagulants (DOACs), the c-statistic was 0.64 for ORBIT and 0.65 for HAS-BLED, with no significant difference. Both scores showed moderate discrimination [3].

In a French primary care cohort (CACAO, 3,082 patients, DOAC or VKA), the c-statistic for all 13 scores tested was 0.41 to 0.66 for major bleeding in DOAC-treated patients with atrial fibrillation. No score showed satisfactory discrimination in primary care [4].

A meta-analysis of nine studies found that HAS-BLED categorised a larger proportion of patients as low risk with a lower observed bleeding rate than ORBIT, suggesting that ORBIT may be less conservative at low scores and thereby identify more patients at genuinely increased risk [5].

Limitations

The score was derived in an American cohort dominated by warfarin (93.5%), and the external validation in ROCKET-AF included patients randomised to rivaroxaban or warfarin. Transferability to patients on other DOACs (apixaban, edoxaban) is supported indirectly through RE-LY and Asian cohorts, but performance is generally moderate rather than good.

All the validation studies show a c-index between 0.62 and 0.67, which means that the score fails to classify approximately one-third of patients correctly. The score must therefore not be used as the sole basis for withholding anticoagulation in a patient at high thromboembolic risk. A high ORBIT score signals a need for risk-reducing measures, not necessarily withholding anticoagulation.

The score lacks variables for liver function and labile INR, both of which are included in HAS-BLED. In patients with severe liver impairment or a very labile INR, HAS-BLED may therefore be the more complete tool, even though ORBIT generally calibrates better [1].

Concomitant antiplatelet treatment is the only modifiable variable in the score. In patients on antiplatelet treatment for a strong indication (for example recent stenting), this trade-off must be made clinically and not according to a total score.

References

1. O'Brien EC, Simon DN, Thomas LE, et al. The ORBIT bleeding score: a simple bedside score to assess bleeding risk in atrial fibrillation. Eur Heart J. 2015;36(46):3258–3264. PMID: 26424865.
2. Proietti M, Hijazi Z, Andersson U, et al. Comparison of bleeding risk scores in patients with atrial fibrillation: insights from the RE-LY trial. J Intern Med. 2018;283(3):282–292. PMID: 29044861.
3. Wattanaruengchai P, Nathisuwan S, Karaketklang K, et al. Comparison of the HAS-BLED versus ORBIT scores in predicting major bleeding among Asians receiving direct-acting oral anticoagulants. Br J Clin Pharmacol. 2022;88(5):2203–2212. PMID: 34783372.
4. Gaboreau Y, Frappé P, Vermorel C, et al. Oral anticoagulant safety in family practice: prognostic accuracy of Bleeding Risk Scores (from the CACAO study). Fam Pract. 2024;41(1):9–17. PMID: 38281089.
5. Zeng J, Yu P, Cui W, et al. Comparison of HAS-BLED with other risk models for predicting the bleeding risk in anticoagulated patients with atrial fibrillation: A PRISMA-compliant article. Medicine. 2020;99(25):e20782. PMID: 32569222.