Author + information
- Received April 1, 2013
- Revision received August 17, 2013
- Accepted August 26, 2013
- Published online December 10, 2013.
- Vanessa Roldán, MD, PhD∗,
- Francisco Marín, MD, PhD†,
- Sergio Manzano-Fernández, MD, PhD†,
- Pilar Gallego, MD∗,
- Juan Antonio Vílchez, PhD†,
- Mariano Valdés, MD, PhD†,
- Vicente Vicente, MD, PhD∗ and
- Gregory Y.H. Lip, MD‡∗ ()
- ∗Hematology and Medical Oncology Unit, Hospital Universitario Morales Meseguer, University of Murcia, Murcia, Spain
- †Department of Cardiology, Hospital Universitario Virgen de la Arrixaca, University of Murcia, Murcia, Spain
- ‡University of Birmingham Centre for Cardiovascular Sciences, City Hospital, Birmingham, United Kingdom
- ↵∗Reprint requests and correspondence:
Prof. Gregory Y. H. Lip, University of Birmingham Centre for Cardiovascular Sciences, City Hospital, Dudley Road, Birmingham B18 7QH, United Kingdom.
Objectives The aim of this study was to test the hypothesis that a specific bleeding risk score, HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/alcohol concomitantly), was better at predicting major bleeding compared with CHADS2 (congestive heart failure, hypertension, 75 years of age or older, diabetes mellitus, and previous stroke or transient ischemic attack) and CHA2DS2-VASc (congestive heart failure, hypertension, 75 years of age and older, diabetes mellitus, previous stroke or transient ischemic attack, vascular disease, 65 to 74 years of age, female) in anticoagulated atrial fibrillation (AF) patients.
Background The CHADS2 and CHA2DS2-VASc scores are well-validated stroke risk prediction scores for AF, but are also associated with increased bleeding and mortality.
Methods We recruited 1,370 consecutive AF patients (49% male; median age, 76 years) receiving oral anticoagulation therapy from our outpatient anticoagulation clinic, all of whom were receiving acenocoumarol and had an international normalized ratio between 2.0 and 3.0 during the preceding 6 months. During follow-up, major bleeding events were identified by the 2005 International Society on Thrombosis and Haemostasis criteria. Model performance was evaluated by calculating the C-statistic, and the improvement in predictive accuracy was evaluated by calculating the net reclassification improvement and integrated discrimination improvement.
Results After a median follow-up of 996 (range, 802 to 1,254) days, 114 patients (3.0%/year) presented with a major bleeding event; 31 of these events were intracranial hemorrhages (0.8%/year). Based on the C-statistic, HAS-BLED had a model performance superior to that of both CHADS2 and CHA2DS2-VASc (both p < 0.001). Both net reclassification improvement and integrated discrimination improvement analyses also show that HAS-BLED was more accurately associated with major bleeding compared with CHADS2 and CHA2DS2-VASc scores.
Conclusions In anticoagulated AF patients, a validated specific bleeding risk score, HAS-BLED, should be used for assessing major bleeding. The practice of using CHADS2 and CHA2DS2-VASc as a measure of high bleeding risk should be discouraged, given its inferior predictive performance compared with the HAS-BLED score.
Oral anticoagulation (OAC) is highly effective in reducing the risk of stroke in atrial fibrillation (AF) patients (1). Decisions on thromboprophylaxis have been based on stroke risk, as assessed by different stroke risk stratification schemes (2), such as the CHADS2 (congestive heart failure, hypertension, 75 years of age or older, diabetes mellitus, and previous stroke or transient ischemic attack) score (3). More recently, the CHA2DS2-VASc (congestive heart failure, hypertension, 75 years of age and older, diabetes mellitus, previous stroke or transient ischemic attack, vascular disease, 65 to 74 years of age, female) score has been used in guidelines, with particular focus on the initial identification of “truly low risk” patients as the initial decision-making step (2,4,5).
Stroke risk and bleeding risk are closely related, and the CHADS2 score closely correlates with bleeding rate (6,7). This has led to many clinicians occasionally using the CHADS2 (and more recently CHA2DS2-VASc) as an indicator of bleeding risk, which could lead to the low use of OAC in those with high CHADS2 (or CHA2DS2-VASc) scores (8–10). In some prescribing recommendations for the novel oral anticoagulants, the lower dose of, for example, dabigatran is recommended at high bleeding risk as quantified by a high CHADS2 score (11,12).
Specific bleeding risk scores are available for patients with AF (13), and the HAS-BLED score is now recommended in European and Canadian AF guidelines to estimate major bleeding risk in anticoagulated AF patients (5,13–15). HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/alcohol concomitantly) has been shown to perform better than other bleeding risk scores (such as HEMORR2HAGES [Hepatic or Renal Disease, Ethanol Abuse, Malignancy, Older Age, Reduced Platelet Count or Function, Re-Bleeding, Hypertension, Anemia, Genetic Factors, Excessive Fall Risk and Stroke], ATRIA [Anticoagulation and Risk Factors in Atrial Fibrillation]) for predicting serious bleeding in vitamin K antagonist (VKA) and non-VKA anticoagulated clinical trial cohorts of AF patients (16), as well as real-world clinical practice (17,18). HAS-BLED is also the only risk score predictive of intracranial bleeding in AF (16) and non-AF (19) patients. Also, HAS-BLED has been related to major bleeding during bridging (20) and percutaneous coronary interventions (21,22) in both AF and non-AF cohorts.
In the present study, we tested the hypothesis that a specific bleeding risk score, HAS-BLED, was better at predicting major bleeding compared with CHADS2 and CHA2DS2-VASc in anticoagulated AF patients.
We recruited consecutive patients with permanent or paroxysmal AF receiving OAC therapy from our outpatient anticoagulation clinic. We studied patients who were entered into our anticoagulation clinic database in 2007 and the first trimester of 2008. The various clinical parameters needed to calculate the HAS-BLED and CHA2DS2-VASc scores were available on our database, and, thus, the various scores were applied retrospectively to the cohort for the present analysis.
All patients were receiving anticoagulation therapy with acenocoumarol and consistently achieved an international normalized ratio (INR) between 2.0 and 3.0 during the previous 6 months of clinic visits. Patients with prosthetic heart valves, acute coronary syndrome, stroke (ischemic or embolic), valvular AF, or any hemodynamic instability as well as patients who had hospital admission or surgical intervention in the preceding 6 months were excluded from the study. A complete medical history was recorded. Follow-up was performed through visits to the anticoagulation clinic.
The HAS-BLED bleeding risk score (14) was calculated as a measure of baseline bleeding risk, as the result of adding 1 point to hypertension, abnormal renal/liver function (1 point each), stroke, bleeding history or predisposition, labile INR, elderly (65 years of age and older), and drugs/alcohol concomitantly (1 point for each one). Based on our inclusion criteria at entry, labile INR was quantified as 0 in every patient.
Baseline stroke risk was assessed using the CHADS2 and CHA2DS2-VASc scores (3,4).
Major bleeding events were defined by the 2005 International Society on Thrombosis and Haemostasis criteria (23).
Continuous variables were tested for normality by the Kolmogorov-Smirnov test. Continuous variables are presented as a mean ± SD or median (interquartile range, as appropriate, and categorical variables as a percentage. Cox models were used to determine the associations between clinical scores and bleeding as well as and mortality.
Model performance was evaluated by calculating C-statistic, and the improvement in predictive accuracy was evaluated by calculating the net reclassification improvement (NRI) and integrated discrimination improvement (IDI), as described by Pencina et al. (24), where the categories of probability for events are defined based on the HAS-BLED or CHADS2 or CHA2DS2-VASc scores. We used the method described by Hanley and McNeil (25,26) for the comparison of correlated C-statistic.
A p value <0.05 was accepted as statistically significant. Statistical analyses were performed using SPSS version 15.0 for Windows (SPSS, Inc., Chicago, Illinois) and SAS version 9.2 (SAS Institute Inc., Cary, North Carolina).
We included 1,370 patients (47% male; median age, 76 years; interquartile range, 71 to 81 years) (Table 1). Median follow-up was 996 days (interquartile range, 802 to 1,254 days). During this period, 114 patients (3.0%/year) presented with a major bleeding event; of these, 31 were intracranial hemorrhages (0.8%/year). A total of 160 patients (4.3%/year) died during the follow-up, 18 (0.4%/year) as a result of a hemorrhage.
Data on the CHA2DS2VASc and HAS-BLED scores of patients who experienced major bleeding events are provided in Online Tables 1 and 2. The cumulative event rates are shown in Kaplan-Meier survival curves as Online Figures 1a and 1b.
Univariate and multivariate analyses
In Table 2, we show the univariate and multivariate analyses, exploring the different variables associated with major bleeding events.
On univariate analysis, the CHADS2 and CHA2DS2-VASc scores were predictive of bleeding events, with a hazard ratio (HR): 1.31 (95% confidence interval [CI]: 1.14 to 1.52; p < 0.001) and an HR: 1.22 (95% CI: 1.09 to 1.37; p = 0.001), respectively. The HAS-BLED score was predictive of major bleeds, with an HR: 1.94 (95% CI: 1.66 to 2.28; p < 0.001), which was higher than for the CHADS2 or CHA2DS2-VASc scores.
On receiver-operating characteristic curve analyses, the HAS-BLED had a similar C-statistic (0.69 ± 0.03; p < 0.001) as the multivariable model (0.71 ± 0.03; p < 0.001) tested in Table 2. The C-statistics for HAS-BLED and the multivariable model were significantly higher than those for CHADS2 (0.59 ± 0.03; p = 0.002) or CHA2DS2-VASc (0.58 ± 0.03; p = 0.006) (both comparisons, p < 0.001) (Table 3).
On multivariate analysis, both CHADS2 and CHA2DS2-VASc scores lost their statistical significance after adjusting by HAS-BLED score (Table 4).
Predictive performance for major bleeding
As detailed in Table 3, the HAS-BLED score showed a model performance (based on C-statistics) superior to both that of CHADS2 and CHA2DS2-VASc scores (both p < 0.001).
Both NRI and IDI analyses show how the HAS-BLED score was more accurately associated with major bleeding episodes than to both the CHADS2 (both p < 0.001) and CHA2DS2-VASc (all p values <0.001 for IDI and NRI).
Based on reclassification analyses, the probability of correctly predicting serious bleeding events using the HAS-BLED score was particularly reflected in the percentage of events correctly reclassified (Table 4).
In this study, we confirm our hypothesis that the HAS-BLED score had modest but significantly better prediction accuracy than stroke stratification scores (CHADS2 and CHA2DS2-VASc) for major bleeding events in anticoagulated AF patients. Thus, a well-validated specific bleeding risk score, HAS-BLED, should be used for assessing major bleeding in AF patients. The practice of using CHADS2 and CHA2DS2-VASc scores as a measure of high bleeding risk should thus be avoided due to its inferior predictive performance compared with that of the HAS-BLED score.
The HAS-BLED bleeding score is a simple and useful tool for application in daily clinical practice. Since its first description in 2010 (14), this score has been validated in several populations and has been shown to outperform several older (and more complicated) bleeding risk scores (16–19,27). In a recent analysis of an unselected nationwide cohort of hospitalized patients with AF, for example, the simple HAS-BLED score was comparable to the older (and more complex) HEMORR2HAGES in predicting bleeding risk (28). Similarly HAS-BLED has been demonstrated to be better than the new ATRIA risk score (16–19).
Stroke risk is closely related to bleeding risk, and OAC therapy needs to balance the benefit from stroke prevention against serious bleeding risk. Many thromboembolic risk factors have also been identified as bleeding risk factors, and thus, risk factors for OAC-associated bleeding, are often also indications for OAC use in AF patients as well (14,29). We previously showed how the HAS-BLED bleeding score, although having good predictive value for assessing major bleeding risk in AF patients (performing as well as a multivariate model), was also modestly predictive of adverse cardiovascular events and death in anticoagulated AF patients (but HAS-BLED performed less well compared with a multivariate model) (30). Indeed, the HAS-BLED score has not been designed to predict thromboembolic events, and for that purpose, the CHADS2 or CHA2DS2-VASc scores have been demonstrated to be much better (30,31).
Our analysis also suggests that there may be limited utility in developing a stratification scheme for predicting both thrombotic and bleeding events. Although all valid stratification schemes (CHADS2/CHA2DS2-VASc and HAS-BLED) share some risk factors in common, the HAS-BLED score is clearly oriented toward predicting serious bleeding (previous hemorrhagic episode, kidney or liver impairment, concomitant aspirin/nonsteroidal anti-inflammatory drug use, labile INR). Another advantage of the HAS-BLED score is that it is relatively simple and user-friendly, yet offers useful predictive capacity for bleeding and makes clinicians think about the potentially reversible risk factors for bleeding.
Of note, the HAS-BLED score is recommended in major guidelines (5,13,15) to “flag” patients potentially at risk of bleeding and to make clinicians think about correcting the potentially reversible bleeding risk factors. The guidelines strongly emphasize that a high HAS-BLED score should not be used as a reason to withhold anticoagulation therapy (5).
There could be some selection bias because patients were on stable OAC therapy at entry (i.e., all INRs were 2.0 to 3.0 in the preceding 6 months to allow for some population homogeneity at study entry), and, thus, patients with unstable anticoagulation who are more prone to have adverse events were excluded. Similarly, they were all “experienced” patients with coumarin treatment. It is well known that the risk of bleeding while receiving anticoagulation therapy or having a thromboembolic event is highest during the period immediately after treatment with coumarin is initiated (32). Nonetheless, it must not be forgotten that even patients with stable OAC (i.e., with nonlabile INRs) may sustain major bleeds (including intracranial bleeding), and numerous studies show how labile INRs are related to increased bleeding risk (33).
Indeed, our observed greater importance of previous bleeding and liver dysfunction components in the development of major bleeds is of interest, but this should be confirmed with other large cohorts, especially because our real-world patients would have been initially selected as being (probably) suitable for the initiation of anticoagulation and hence referred to our anticoagulation clinic. Nonetheless, the strength of our study is its inclusion of consecutive patients attending our anticoagulation clinic, and even when avoiding 1 point on the HAS-BLED score due to the labile INR criterion in our well-controlled VKA cohort, this score still maintains its predictive power of major bleeding events.
We have demonstrated that the predictive value and usefulness of the HAS-BLED score for predicting major bleeding is better than that of the CHADS2 or CHA2DS2-VASc scores, in a large cohort of consecutive patients taking acenocoumarol. The use of CHADS2 and CHA2DS2-VASc as a measure of high bleeding risk in AF should be discouraged, given its inferior predictive performance compared with the HAS-BLED score. Thus, a well-validated specific bleeding risk score, HAS-BLED, should be used for assessing major bleeding in anticoagulated AF patients.
For supplemental tables and figures, please see the online version of this article.
This work was partially supported by Sociedad Española de Cardiología, RD06/0014/039, (RECAVA) from ISCIII, and PI11/1256-FEDER from ISCIII. Dr. Roldán has received funding for consultancy and lecturing from Bristol-Myers Squibb, Bayer, and Boehringer Ingelheim. Dr. Marín has received funding for research, consultancy, and lecturing from Abbott, Boston Scientific, Bayer, AstraZeneca, Daiichi Sankyo, Bristol-Myers Squibb/Pfizer Inc., and Boehringer Ingelheim. Dr. Lip has received funding for research, consultancy, and lecturing from different manufacturers of drugs used for the treatment of atrial fibrillation, including AstraZeneca, Bayer, Boehringer Ingelheim, Astellas, sanofi-aventis, and Daiichi Sankyo. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- congestive heart failure, hypertension, 75 years of age or older, diabetes mellitus, and previous stroke or transient ischemic attack
- congestive heart failure, hypertension, 75 years of age and older, diabetes mellitus, previous stroke or transient ischemic attack, vascular disease, 65 to 74 years of age, female
- confidence interval
- hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/alcohol concomitantly
- hazard ratio
- integrated discrimination improvement
- international normalized ratio
- net reclassification improvement
- oral anticoagulation
- vitamin K antagonist
- Received April 1, 2013.
- Revision received August 17, 2013.
- Accepted August 26, 2013.
- American College of Cardiology Foundation
- Lip G.Y.
- Camm A.J.,
- Lip G.Y.,
- De Caterina R.,
- et al.,
- ESC Committee for Practice Guidelines
- Hylek E.M.,
- Evans-Molina C.,
- Shea C.,
- Henault L.E.,
- Regan S.
- Wilke T.,
- Groth A.,
- Mueller S.,
- et al.
- Holt T.A.,
- Hunter T.D.,
- Gunnarsson C.,
- Khan N.,
- Cload P.,
- Lip G.Y.
- ↵MIMS Malaysia: Full prescribing information of Pradaxa. http://www.mims.com.my/MALAYSIA/drug/info/Pradaxa/?type=full#SpecialPrecautions. Accessed March 6, 2013.
- Pradaxa Prescribing Information, Taiwan Food and Drug. Available at: Administration. http://www.fda.gov.tw/MLMS/(S(nmvmo245foj0qw55kb0ae4bi))/ShowFile.aspx?LicId=02025458&Seq=002&Type=9. Accessed March 6, 2013.
- Lip G.Y.,
- Andreotti F.,
- Fauchier L.,
- et al.,
- European Heart Rhythm Association
- Skanes A.C.,
- Healey J.S.,
- Cairns J.A.,
- et al.,
- Canadian Cardiovascular Society Atrial Fibrillation Guidelines Committee
- Apostolakis S.,
- Lane D.A.,
- Guo Y.,
- Buller H.,
- Lip G.Y.
- Lip G.Y.,
- Banerjee A.,
- Lagrenade I.,
- Lane D.A.,
- Taillandier S.,
- Fauchier L.
- Lip G.Y.,
- Lin H.J.,
- Hsu H.C.,
- et al.
- Ruiz-Nodar J.M.,
- Marín F.,
- Roldan V.,
- et al.
- Schulman S.,
- Kearon C.,
- Subcommittee on Control of Anticoagulation of the Scientific and Standardization Committee of the International Society on Thrombosis and Haemostasis
- Lip G.Y.,
- Frison L.,
- Halperin J.L.,
- Lane D.A.
- Hughes M.,
- Lip G.Y.,
- on behalf of the Guideline Development Group for the NICE national clinical guideline for management of atrial fibrillation in primary and secondary care
- Gallego P.,
- Roldán V.,
- Torregrosa J.M.,
- et al.
- Wan Y.,
- Heneghan C.,
- Perera R.,
- et al.