Considerations when choosing an appropriate bleeding risk assessment tool for patients with atrial fibrillation

Abstract

Oral anticoagulation is integral to the management of patients with atrial fibrillation (AF) to reduce the risk of thromboembolism. Stroke risk should be assessed using the CHA2DS2‐VASc score and men with a CHA2DS2‐VASc score of ≥1 and women with CHA2DS2‐VASc score ≥2 should be offered anticoagulation.1.Lip G.Y.H. Banerjee A. Boriani G. et al.Antithrombotic therapy for atrial fibrillation: CHEST guideline and expert panel report.Chest. 2018; 154: 1121-1201Abstract Full Text Full Text PDF PubMed Scopus (573) Google Scholar However, use of anticoagulation increases the risk of bleeding events and thus bleeding risk must be taken into account when initiating anticoagulation, especially because anticoagulation‐related major bleeding in AF patients has been associated with a substantial increase in the risk of death, ischemic stroke, and myocardial infarction.2.Held C. Hylek E.M. Alexander J.H. et al.Clinical outcomes and management associated with major bleeding in patients with atrial fibrillation treated with apixaban or warfarin: insights from the ARISTOTLE trial.Eur Heart J. 2015; 36: 1264-1272Crossref PubMed Scopus (115) Google Scholar, 3.Lip G.Y.H. Lane D.A. Bleeding risk assessment in atrial fibrillation: observations on the use and misuse of bleeding risk scores.J Thromb Haemost. 2016; 14: 1711-1714Crossref PubMed Scopus (93) Google Scholar Intracranial hemorrhage, the most serious form of bleeding, was linked with a hazard ratio (HR) of 121.5 (95% confidence interval [CI], 91.3‐161.8) for death and a HR of 22.0 (95% CI, 9.9‐48.8) for stroke or myocardial infarction.2.Held C. Hylek E.M. Alexander J.H. et al.Clinical outcomes and management associated with major bleeding in patients with atrial fibrillation treated with apixaban or warfarin: insights from the ARISTOTLE trial.Eur Heart J. 2015; 36: 1264-1272Crossref PubMed Scopus (115) Google Scholar Several risk prediction models to determine risk of bleeding have been developed based on various clinical, biological, and/or genetic markers.4.Pisters R. Lane D.A. Nieuwlaat 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; 138: 1093-1100Abstract Full Text Full Text PDF PubMed Scopus (3292) Google Scholar, 5.Fang M.C. Go A.S. Chang Y. et al.A new risk scheme to predict warfarin‐associated hemorrhage: the ATRIA (Anticoagulation and Risk Factors in Atrial Fibrillation) study.J Am Coll Cardiol. 2011; 58: 395-401Crossref PubMed Scopus (651) Google Scholar, 6.Hijazi Z. Oldgren J. Lindback J. et al.The novel biomarker‐based ABC (age, biomarkers, clinical history)‐bleeding risk score for patients with atrial fibrillation: a derivation and validation study.Lancet. 2016; 387: 2302-2311Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar, 7.Gage B.F. Yan Y. Milligan P.E. et al.Clinical classification schemes for predicting hemorrhage: results from the National Registry of Atrial Fibrillation (NRAF).Am Heart J. 2006; 151: 713-719Crossref PubMed Scopus (809) Google Scholar, 8.Zulkifly H. Lip G.Y.H. Lane D.A. Bleeding risk scores in atrial fibrillation and venous thromboembolism.Am J Cardiol. 2017; 120: 1139-1145Abstract Full Text Full Text PDF PubMed Scopus (38) Google Scholar The HAS‐BLED score incorporates common bleeding risk factors for patients with AF4.Pisters R. Lane D.A. Nieuwlaat 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; 138: 1093-1100Abstract Full Text Full Text PDF PubMed Scopus (3292) Google Scholar and is recommended by some guidelines for use in clinical practice.1.Lip G.Y.H. Banerjee A. Boriani G. et al.Antithrombotic therapy for atrial fibrillation: CHEST guideline and expert panel report.Chest. 2018; 154: 1121-1201Abstract Full Text Full Text PDF PubMed Scopus (573) Google Scholar, 9.National Institute for Health and Care Excellence. Guideline on atrial fibrillation: management; 2014: p. 1‐45.Google Scholar However, bleeding risk scores have been used inappropriately by the ill‐informed as an excuse to withhold anticoagulation. A high bleeding risk score (e.g., a HAS‐BLED score of ≥3) is not a contraindication for oral anticoagulants, but instead should prompt responsible clinicians to undertake the necessary steps to reduce this risk and address modifiable risk factors such as uncontrolled hypertension, poor control of International Normalized Ratios (labile INRs) if receiving a vitamin K antagonist, concomitant use of medications that increase the risk of bleeding including aspirin or nonsteroidal anti‐inflammatory drugs, and harmful alcohol consumption.10.Kirchhof P. Benussi S. Kotecha D. et al.2016 ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS.Eur Heart J. 2016; 37: 2893-2962Crossref PubMed Scopus (5124) Google Scholar Other potentially modifiable risk factors for bleeding such as anemia, impaired renal and/or hepatic function, and reduced platelet count or function should also be considered.10.Kirchhof P. Benussi S. Kotecha D. et al.2016 ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS.Eur Heart J. 2016; 37: 2893-2962Crossref PubMed Scopus (5124) Google Scholar The HAS‐BLED score is then used to flag up those “high‐risk” patients for early review and follow‐up (e.g., 4 weeks, rather than 4‐6 months). It is vital that patients with AF who are at high risk of bleeding are identified as early as possible, so potentially modifiable risk factors can be addressed and the patients can be appropriately monitored. The dynamic nature of risk also necessitates that regular assessments of bleeding (and stroke risk) are performed (Figure 1) because current evidence has demonstrated that changes in risk profiles are important predictors of adverse events in patients with AF.11.Chao T.‐.F. Lip G.Y.H. Liu C.‐.J. et al.Relationship of aging and incident comorbidities to stroke risk in patients with atrial fibrillation.J Am Coll Cardiol. 2018; 71: 122-132Crossref PubMed Scopus (121) Google Scholar, 12.Chao T.‐.F. Lip G.Y.H. Lin Y.‐.J. et al.Incident risk factors and major bleeding in patients with atrial fibrillation treated with oral anticoagulants: a comparison of baseline, follow‐up and delta HAS‐BLED scores with an approach focused on modifiable bleeding risk factors.Thromb Haemost. 2018; 118: 768-777Crossref PubMed Scopus (102) Google Scholar, 13.Yoon M. Yang P.‐.S. Jang E. et al.Dynamic changes of CHA2DS2‐VASc score and the risk of ischaemic stroke in Asian patients with atrial fibrillation: a nationwide cohort study.Thromb Haemost. 2018; 118: 1296-1304Crossref PubMed Scopus (81) Google Scholar Bleeding risk scores vary in their complexity and simplicity to implement in clinical practice.8.Zulkifly H. Lip G.Y.H. Lane D.A. Bleeding risk scores in atrial fibrillation and venous thromboembolism.Am J Cardiol. 2017; 120: 1139-1145Abstract Full Text Full Text PDF PubMed Scopus (38) Google Scholar Although the inclusion of additional factors may increase the accuracy of risk models, this is often at the expense of practicality and ease of calculation. Indeed, risk prediction, especially with incorporation of various biomarkers, can improve prediction (at least statistically) but at the cost of additional complexity, cost, and reduced practicality.14.Esteve‐Pastor M.A. Roldan V. Rivera‐Caravaca J.M. Ramirez‐Macias I. Lip G.Y.H. Marin F. The use of biomarkers in clinical management guidelines: a critical appraisal.Thromb Haemost. 2019; 119: 1901-1919Crossref PubMed Scopus (51) Google Scholar There also exists a complex relationship between thrombosis and bleeding and several risk factors are shared (e.g., age, renal dysfunction, malignancy), which complicates clinical decisions. For the vast majority of patients with AF, the net clinical benefit of oral anticoagulation to reduce risk of stroke outweighs the risk of major bleeding, including among patients identified as being at high risk of bleeding.15.Friberg L. Rosenqvist M. Lip G.Y.H. Net clinical benefit of warfarin in patients with atrial fibrillation: a report from the Swedish atrial fibrillation cohort study.Circulation. 2012; 125: 2298-2307Crossref PubMed Scopus (374) Google Scholar In the current issue of the Journal of Thrombosis and Haemostasis, Chang et al16.Chang G. Xie Q. Ma L. et al.Accuracy of HAS‐BLED and other bleeding risk assessment tools in predicting major bleeding events in atrial fibrillation: a network meta‐analysis.J Thromb Haemost. 2019; PubMed Google Scholar report the results from a network meta‐analysis of 18 studies (n = 321 888 patients) comparing the sensitivity and specificity of the HAS‐BLED model and other risk assessment models for predicting major bleeding events in patients with AF. Overall, Chang et al16.Chang G. Xie Q. Ma L. et al.Accuracy of HAS‐BLED and other bleeding risk assessment tools in predicting major bleeding events in atrial fibrillation: a network meta‐analysis.J Thromb Haemost. 2019; PubMed Google Scholar show that the European score based only on modifiable bleeding risk factors, ABC and mOBRI models had high sensitivity but low specificity, whereas the ORBIT, ATRIA, Shireman, and GARFIELD‐AF registry models had high specificity, but low sensitivity. The network meta‐analysis clearly demonstrates that the HAS‐BLED model was the most balanced in terms of sensitivity and specificity, slightly surpassing the HEMORR2HAGES model. These results provide an overview of the different bleeding risk models available and also quantitative evidence supporting the results of a previous systematic review, which also concluded that the HAS‐BLED provides the best prediction model for assessment of bleeding risk.17.Borre E.D. Goode A. Raitz G. et al.Predicting thromboembolic and bleeding event risk in patients with non‐valvular atrial fibrillation: a systematic review.Thromb Haemost. 2018; 118: 2171-2187Crossref PubMed Scopus (121) Google Scholar The study by Chang et al16.Chang G. Xie Q. Ma L. et al.Accuracy of HAS‐BLED and other bleeding risk assessment tools in predicting major bleeding events in atrial fibrillation: a network meta‐analysis.J Thromb Haemost. 2019; PubMed Google Scholar highlights some of the strengths and weaknesses of individual bleeding risk scores, particularly that those with a high sensitivity often had low specificity and vice versa. Although the HAS‐BLED and HEMORR2HAGES models were balanced, both demonstrated only modest values of sensitivity and specificity. Modest values in both attributes may cause several limitations; indeed, scores with high sensitivity, but low specificity or scores with high specificity, but low sensitivity may be more useful in certain situations.18.Trevethan R. Sensitivity, specificity, and predictive values: foundations, pliabilities, and pitfalls in research and practice.Front Public Health. 2017; 5: 307Crossref PubMed Scopus (559) Google Scholar For instance, the ORBIT model had high specificity and would identify patients at high risk of bleeding. Conversely, the mOBRI model had high sensitivity and could be used to confidently identify patients at low risk of bleeding. When using the various bleeding risk models as screening tools, it is important to consider the characteristics of the cohort of interest, including the tendency for patients to have a high risk of bleeding (Bayes’ theorem).19.Westbury C.F. Bayes’ rule for clinicians: an introduction.Front Psychol. 2010; 1: 192Crossref PubMed Scopus (34) Google Scholar As such, Chang et al16.Chang G. Xie Q. Ma L. et al.Accuracy of HAS‐BLED and other bleeding risk assessment tools in predicting major bleeding events in atrial fibrillation: a network meta‐analysis.J Thromb Haemost. 2019; PubMed Google Scholar appropriately suggest that among patients at high risk of bleeding, a model with high sensitivity is desirable, whereas among patients at low risk of bleeding, a model with high specificity is desirable. However, sensitivity and specificity may be of little practical use for clinicians when considering the probability of bleeding events in individual patients. During such situations, the predictive probabilities (positive and negative predictive value [NPV]) of the various models are more important.18.Trevethan R. Sensitivity, specificity, and predictive values: foundations, pliabilities, and pitfalls in research and practice.Front Public Health. 2017; 5: 307Crossref PubMed Scopus (559) Google Scholar Thus, a model with high sensitivity or specificity may have little value if it has low positive and NPV, such that clinicians are unable to produce reliable results from their assessments. To expand on this, a model with high specificity but a low NPV may incorrectly classify a significant proportion of patients as low risk of bleeding when they should have been identified at high risk instead (false negatives). In practice, this misclassification could result in anticoagulated patients with AF not receiving the appropriate monitoring and a lack of attention on modifiable risk factors for bleeding. Of note, the HAS‐BLED score also draws attention to several potentially reversible risk factors (uncontrolled hypertension; labile INR; concomitant use of nonsteroidal anti‐inflammatory drugs or excess alcohol) when compared with other models such as mOBRI, ATRIA, ABC, and GARFIELD‐AF, which mostly consist of nonmodifiable risk factors. No single model for assessing bleeding risk for patients with AF has both high sensitivity and high specificity and provides a best fit for every clinical situation. The important thing is that a formal bleeding risk assessment is undertaken in all patients initiating oral anticoagulants and that bleeding risk scores, such as HAS‐BLED, are memorable acronyms/tools to assist in identifying and addressing modifiable risk factors and identifying which patients will require closer/more frequent follow‐up to reduce risk. In summary, bleeding risk scores in patients with AF should not be used as an excuse to avoid initiation of oral anticoagulation. Furthermore, anticoagulants should not be avoided solely because the patient is at an increased risk of falls. Patients with AF identified as high risk of bleeding should be adequately monitored and appropriate strategies should be implemented to reduce the patient's risk of bleeding when commencing anticoagulation. W. Y. Ding and S. L. Harrison wrote the first draft and all authors provided critical revision. D.A. Lane reports receiving investigator‐initiated educational grants from Bristol Myers Squibb (BMS) and Boehringer Ingelheim; serving as a speaker for Boehringer Ingelheim, Bayer, and BMS/Pfizer; and consulting for Boehringer Ingelheim, Bayer, BMS/Pfizer, and Daiichi‐Sankyo. G.Y.H. Lip reports being a consultant for Bayer/Janssen, BMS/Pfizer, Medtronic, Boehringer Ingelheim, Novartis, Verseon and Daiichi‐Sankyo and a speaker for Bayer, BMS/Pfizer, Medtronic, Boehringer Ingelheim, and Daiichi‐Sankyo. No fees are directly received personally.