Clinical background
The decision on the duration of antiplatelet treatment after percutaneous coronary intervention (PCI) is a trade-off between ischaemic and bleeding risk. Patients at high bleeding risk have historically either been excluded from clinical trials or treated with a bare metal stent to allow the shortest possible dual antiplatelet therapy (DAPT), a strategy that has proved inferior to modern drug-eluting stents even with a short DAPT duration. At the same time, the lack of a uniform definition of "high bleeding risk" has hampered both trial design and clinical decision-making.
The ARC-HBR criteria were developed to fill this void. Unlike earlier bleeding risk scores such as PRECISE-DAPT and PARIS, which rest on statistical models with moderate discrimination (a c-statistic of 0.64 to 0.73 in their respective derivation cohorts), ARC-HBR is a consensus-based definition produced by a 31-member consortium with representatives from academic research organisations, the US FDA, the Japanese medicines agency and European notified bodies [1]. The aim was to create a standardised, pragmatic definition that can be used both in clinical trials and in clinical practice.
Applying the ARC-HBR criteria
ARC-HBR is not a scoring scale but a binary classification. A patient is defined as being at high bleeding risk if at least one major criterion or at least two minor criteria are met:
Major criteria are those judged in isolation to confer a BARC 3 or 5 bleeding risk of at least 4 per cent at one year, or an intracranial bleeding risk of at least 1 per cent at one year. Minor criteria are those judged in isolation to confer an increased bleeding risk but with a BARC 3 or 5 bleeding risk below 4 per cent at one year [1].
The major criteria include long-term treatment with an oral anticoagulant, an eGFR below 30 mL/min/1.73 m², an Hb below 11 g/dL, recent bleeding requiring transfusion or hospital care, severe thrombocytopenia, cirrhosis with portal hypertension, active malignancy and previous spontaneous intracranial bleeding. The minor criteria include age 75 years or older, moderate renal impairment (eGFR 30 to 59), an Hb of 11 to 12.9 g/dL in men and 11 to 11.9 g/dL in women, spontaneous bleeding requiring hospital care in the past year, long-term treatment with NSAIDs or steroids, and previous ischaemic stroke [1].
The thresholds for high bleeding risk — 4 per cent for BARC 3 to 5 bleeding and 1 per cent for intracranial bleeding at one year — were chosen because the 1-year risk of major bleeding in DAPT trials that excluded high bleeding risk patients was below 3 per cent, whereas trials that included such patients showed a BARC 3 to 5 bleeding risk of 4 per cent or higher [1].
The derivation did not take place in a traditional statistical cohort. The criteria were based on a structured literature review and consensus from two meetings, in Washington, DC, in April 2018 and in Paris in October 2018. This is important: ARC-HBR is not a model derived in a specific population, and its performance must therefore be evaluated in external validation studies.
Interpretation in practice
The classification is binary and leads to two courses of action:
| Classification | Criteria met | Clinical action |
|---|---|---|
| High bleeding risk | At least 1 major or at least 2 minor criteria | Consider a shortened DAPT duration (1 to 3 months), early P2Y12 inhibitor monotherapy, and careful follow-up of bleeding risk factors |
| Not high bleeding risk | 0 major criteria and 0 to 1 minor criterion | Standard DAPT duration according to current guidelines, with individual assessment of ischaemic risk |
An important finding from the validation studies is that risk increases stepwise with the number of criteria met. In a cohort from Mount Sinai, bleeding risk rose progressively with the number of times the ARC-HBR definition was met, suggesting that the criteria have additive prognostic value beyond the binary classification [3]. The meta-analysis of 68,874 patients confirmed this pattern and proposed that the number of criteria met could be used for finer risk stratification [2].
Validation and performance
Several external validation studies have been published since the consensus document appeared in 2019.
Bern PCI Registry (12,121 patients, 2009 to 2016): 39.4 per cent were classified as high bleeding risk. BARC 3 or 5 bleeding at one year occurred in 6.4 per cent of high bleeding risk patients compared with 1.9 per cent of the others (p < 0.001). The device-oriented composite endpoint (DOCE) was also higher: 12.5 per cent versus 6.1 per cent. ARC-HBR had higher sensitivity than PRECISE-DAPT (63.8 per cent versus 53.1 per cent) and PARIS (31.9 per cent) but lower specificity (62.7 per cent versus 71.3 per cent and 86.5 per cent respectively) [4].
The Mount Sinai cohort (9,623 patients, 2014 to 2017): 44.4 per cent were classified as high bleeding risk. The primary bleeding endpoint at one year was 9.1 per cent in high bleeding risk patients versus 3.2 per cent in the others (p < 0.001). High bleeding risk patients also had a significantly higher rate of thrombotic events and all-cause mortality [3].
Japanese cohort (3,410 patients, 2010 to 2013): ARC-HBR showed higher diagnostic sensitivity for bleeding events than PRECISE-DAPT, PARIS and the CREDO-Kyoto bleeding score, in line with the findings from the Bern cohort [5].
Meta-analysis (9 studies, 68,874 patients): 39.2 per cent were classified as high bleeding risk. The relative risk of major bleeding was 2.70 (95 per cent CI 2.35 to 3.10; p < 0.0001). The pooled c-statistic was 0.69 (95 per cent CI 0.61 to 0.75), indicating moderate discrimination. Calibration was suboptimal with a tendency to underestimate bleeding risk (pooled observed-to-expected ratio 1.47; 95 per cent CI 0.82 to 2.60). A sensitivity analysis of 5 studies and 46,712 patients confirmed that the criteria have additive prognostic value when several are met. Acute coronary syndrome as the presentation was the only baseline characteristic significantly to affect predictive performance [2].
In summary, ARC-HBR performs better than older bleeding risk scores in terms of sensitivity, but at the price of lower specificity. Discrimination is moderate and calibration tends to underestimate the actual bleeding risk, particularly in patients with acute coronary syndrome.
Limitations
ARC-HBR is a consensus definition, not a statistically derived model. This means that the criteria and their thresholds rest on expert judgement and a literature review rather than on formal regression in a specific cohort. The moderate c-statistic (0.69 in the meta-analysis) reflects this [2].
The criteria apply to patients undergoing PCI. They are not validated for patients undergoing coronary artery bypass surgery or for medical management of coronary artery disease without intervention.
A recurring finding is that ARC-HBR classifies a large proportion of PCI patients as high bleeding risk, 39 to 44 per cent in the validation cohorts [2,3,4]. This gives high sensitivity but also means that a substantial proportion of patients are flagged as high bleeding risk without necessarily needing shortened DAPT. Lower specificity compared with PRECISE-DAPT and particularly PARIS means that ARC-HBR captures more patients who do not bleed, which can lead to undertreatment with respect to ischaemic protection if the criteria are applied mechanically [4].
The weakness in calibration, with a tendency to underestimate bleeding risk, is particularly marked in patients with acute coronary syndrome [2]. This should be borne in mind when assessing patients undergoing PCI on an emergency indication.
A practical problem is that several criteria require information that is not always available when PCI is being planned, for example ongoing NSAID or steroid treatment or a detailed bleeding history. Incomplete data can lead to patients being wrongly classified as not at high bleeding risk.
References
- Urban P, Mehran R, Colleran R, et al. Defining High Bleeding Risk in Patients Undergoing Percutaneous Coronary Intervention: A Consensus Document From the Academic Research Consortium for High Bleeding Risk. Circulation 2019;140(3):240-261. PMID: 31116032
- Montalto C, Munafò AR, Arzuffi L, et al. Validation of the ARC-HBR criteria in 68,874 patients undergoing PCI: A systematic review and meta-analysis. Hellenic J Cardiol 2022;66:59-66. PMID: 35550178
- Cao D, Mehran R, Dangas G, et al. Validation of the Academic Research Consortium High Bleeding Risk Definition in Contemporary PCI Patients. J Am Coll Cardiol 2020;75(21):2711-2722. PMID: 32466887
- Ueki Y, Bär S, Losdat S, et al. Validation of the Academic Research Consortium for High Bleeding Risk (ARC-HBR) criteria in patients undergoing percutaneous coronary intervention and comparison with contemporary bleeding risk scores. EuroIntervention 2020;16(5):371-379. PMID: 32065586
- Okabe K, Miura K, Shima Y, et al. Comparison and Validation of Long-Term Bleeding Events for Academic Bleeding Risk (ARC-HBR) Criteria and Contemporary Risk Scores for Percutaneous Coronary Intervention With a Second-Generation Drug Eluting Stent. Circ J 2022;86(9):1379-1387. PMID: 35400715
