Bleeding risk scores in atrial fibrillation—how should they influence decision-making for anticoagulants?

Editorial to accompany: A new bleeding risk score specifically developed for direct oral anticoagulants in a geriatric population. Bleeding risk tools can be useful to identify risk factors but should not be used in isolation to decide whether to prescribe anticoagulation. Addressing modifiable risk factors and shared decision-making are crucial to optimise patient care. Atrial fibrillation (AF) is a common cardiac condition, and the prevalence of this rises rapidly with advancing age, affecting up to 17% of people aged ≥80 years [1]. AF is associated with a five-fold increase in the risk of stroke in the general population; however, for older people, the risk of stroke with AF is higher still [2]. The stroke risk attributable to AF also increases with age, rising from 2.8% in those aged 60–69 years to 23.5% in those aged 80–89 years in the Framingham study [2]. Stroke prevention with anticoagulation is therefore an important consideration when treating older people with AF. The benefits of anticoagulation in older patients, including those with frailty, are well evidenced [3–5] and the direct oral anticoagulants (dabigatran, rivaroxaban, apixaban, and edoxaban) have now largely surpassed vitamin K antagonists such as warfarin as first-line treatment due to their favourable dosing schedules and safety profiles [6, 7]. Despite their efficacy, anticoagulants are not without their risks with the primary concern being bleeding, and clinicians can be reluctant to prescribe these medications to patients they see as high risk, for example, those at risk of falls [8]. Many scores have been developed to guide clinicians in assessing risk of bleeding for people with AF and international guidelines have long advocated the use of these tools before starting anticoagulation [7]. However, there is now a move away from using risk scores to decide on whether to start or stop an anticoagulant. There are relatively few absolute contraindications to DOACs, and stopping anticoagulants for older patients can significantly increase their risk of adverse outcomes [9]. The focus should instead be on addressing modifiable risk factors and taking an individualised approach to monitoring and decision-making for high-risk patients [6]. Despite this, risk scores can still be useful in highlighting high-risk patients who may require closer monitoring, to identify patients in routine health data that may benefit from review, or for explaining risk to aid shared decision-making. Several bleeding risk scores have been produced over the last 20 years, all using different development cohorts and subsequently validated in external cohorts. The most commonly cited scores being HAS-BLED [10], ORBIT-AF [11], ATRIA [12], and HEMORR2HAGES [13] which were all developed using cohorts of patients taking warfarin. As we now move to the era of DOACs, there is a question as to whether these scores are still valid or whether other risk factors may now be more important. Some also include irrelevant variables for DOAC patients such as labile INR [10] or are complicated to use in practice due to inclusion of genetic components [13]. The most recent bleeding risk score, named the ‘DOAC Score’ [14] was developed and validated using data from a number of DOAC trials and registries, this was shown to perform better than HAS-BLED in external validation, although the overall discrimination was only modest for both scores [14]. There is also a question as to how well the currently available scores predict risk in older patients, the mean age in the HAS-BLED development cohort was only 66 years [10]. However, for other scores the average age ranged from 75 to 80 years so may be more applicable for the older population [11–14]. The new A4C score, described by Piccoli and colleagues [15], aims to address some of these concerns by developing a score that incorporates data from the comprehensive geriatric assessment. The average age of their development cohort was significantly higher than other cohorts at 86 years, with over half of their patients being aged over 85 years, giving a better insight into bleeding risk for the very old. The final score included age (<85 years = 0; 85–95 years = 1; >95 years = 3), anaemia (haemoglobin <110 g/l = 2), low albumin (<35 g/l = 1), amiodarone use (=1) and low creatinine clearance (≥50 ml/min = 0; 30–50 ml/min = 1; <30 ml/min = 2) as risk factors for bleeding with higher scores corresponding with higher risk of bleeding. The A4C score showed moderate discrimination for major bleeding in the training cohort (0.73, 95% CI 0.64–0.82), falling to 0.66 (95% CI 0.58–0.75) in the validation cohort, but still outperformed other scores (HAS-BLED, ATRIA and HEMORR2HAGES) in both cohorts. Two bleeding risk factors identified by Piccoli and colleagues are aligned with those included in the other risk scores. Anaemia is included in both ATRIA [12] and HEMORR2HAGES [13]. Advancing age is included in all scores, with most assigning 1 or 2 points for ages ≥75 years [11–13], HAS-BLED assigns 1 point for age ≥ 65 years [10]. The DOAC score assigns higher points for age band categories from 65 years up to a top band of ≥80 years and the A4C only scores for ages ≥85 but with a top band of ≥90 years. Amiodarone and low albumin are unique to the A4C but other concomitant medications are included in other scores, and low albumin may be related to kidney or liver disease, malnutrition and low body weight which are risk factors included in other scores. In conclusion, the A4C score adds to the clinician arsenal that can be used to assess bleeding risk for their older patients when considering the risks of anticoagulation for AF. However, it needs to be validated externally in larger international cohorts to truly evaluate its utility. No matter how many risk scores are developed, it is important to recognise the myriad of factors that may increase a patient’s risk of bleeding and how combinations of factors may increase the risk for individual patients. There is a need to address the modifiable risk factors and discuss with the patient their priorities to complement the use of these tools in deciding whether anticoagulation should be prescribed or not. Developing these tools in conjunction with patients and their carers to produce information would be beneficial to aid shared decision-making. Neither author have any conflicts of interest to declare. Anneka Mitchell, NIHR Senior Clinical and Practitioner Research Award, NIHR304550 is funded by the NIHR for research time to complete this work. The views expressed in this publication are those of the authors and not necessarily those of the NIHR, NHS or the UK Department of Health and Social Care.