Advertisement

Recalibration of the HAS-BLED Score

Should Hemorrhagic Stroke Account for One or Two Points?
      After a hemorrhagic stroke, it is uncertain whether this event scores one point (either for stroke or bleeding) or two points (one point each for stroke and bleeding) on the bleeding risk score termed HAS-BLED (hypertension, abnormal renal/liver function [one or two points], stroke, bleeding history or predisposition, labile international normalized ratio [INR], elderly [> 65 years], drugs/alcohol concomitantly [one or two points]). We investigated the value of a recalibration of the HAS-BLED score to account for two points from a hemorrhagic stroke. Data were analyzed from the Danish nationwide cohort of patients with incident atrial fibrillation (AF) from January 1999 to December 2013. The primary outcome in this observational study was major bleeding. The original and the recalibrated HAS-BLED scores were assessed, and the event rates of major bleeding were calculated. The predictive accuracy of major bleeding was compared by using C-statistics, the net reclassification index (NRI), and integrated discrimination improvement (IDI). An event rate for major bleeding of 4.3 per 100 person-years was recorded in the 210,299 patients with AF. The C-statistics for the two scores were modest: 0.613 (95% CI, 0.607-0.619) for the original score and 0.616 (95% CI, 0.610-0.622) for the recalibrated score. The NRI was 10.0% (95% CI, 7.6-12.4). The relative IDI was 23.6% (95% CI, 15.7-31.5), reflecting that the recalibrated HAS-BLED score more accurately predicted bleeding events. Recalibration of the “S” component in the HAS-BLED score (counting two points for a hemorrhagic stroke) resulted in an increase in the C-statistics, NRI, and IDI. This approach could potentially aid physicians in more accurate assessments of bleeding risk in patients with AF.

      Key Words

      Abbreviations:

      AF (atrial fibrillation), IDI (integrated discrimination improvement), INR (international normalized ratio), NRI (net reclassification index), OAC (oral anticoagulant)
      Oral anticoagulant (OAC) treatment substantially reduces the risk of stroke and all-cause mortality in patients with atrial fibrillation (AF).
      • Hart R.G.
      • Pearce L.A.
      • Aguilar M.I.
      Meta-analysis: antithrombotic therapy to prevent stroke in patients who have nonvalvular atrial fibrillation.
      However, the decision to treat these patients relies on the expected risk of stroke weighed against the expected risk of bleeding. Contemporary guideline recommendations on OAC treatment are based on the CHA2DS2-VASc score (assessing the risk of thromboembolism [congestive heart failure, hypertension, age ≥ 75 years, diabetes mellitus, stroke/transient ischemic attack, vascular disease, age 65-74 years, sex category]) regarding whether to initiate treatment.
      • Camm A.J.
      • Lip G.Y.
      • De Caterina R.
      • et al.
      and the ESC Committee for Practice Guidelines-CPG; Document Reviewers
      2012 Focused update of the ESC guidelines for the management of atrial fibrillation: an update of the 2010 ESC guidelines for the management of atrial fibrillation—developed with the special contribution of the European Heart Rhythm Association.
      • January C.T.
      • Wann F.L.
      • Joseph F.
      • et al.
      2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: executive summary.
      The European Society of Cardiology further recommends formal assessment of the bleeding risk by using the HAS-BLED score.
      • Camm A.J.
      • Kirchhof P.
      • Lip G.Y.
      • et al.
      Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC).
      This score summarizes to a maximum of nine points (hypertension, abnormal renal/liver function [one or two points], stroke, bleeding history or predisposition, labile international normalized ratio [INR], elderly [> 65 years], drugs/alcohol concomitantly [one or two points]).
      • Pisters R.
      • Lane D.A.
      • Nieuwlaat R.
      • de Vos C.B.
      • Crijns H.J.
      • Lip G.Y.
      A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the Euro Heart Survey.
      A stroke currently contributes one point, and prior major bleeding (or its predisposition) contributes one point. Accurate assessment of bleeding risk and optimal treatment guidance are pivotal in this frail population of patients with AF who sustain intracranial bleeding, especially given the strong associations to disability and mortality.
      • Marsh E.B.
      • Gottesman R.F.
      Brain hemorrhage: restarting anticoagulation after intracranial hemorrhage.
      • Nielsen P.B.
      • Larsen T.B.
      • Skjøth F.
      • Gorst-Rasmussen A.
      • Rasmussen L.H.
      • Lip G.Y.
      Restarting anticoagulant treatment after intracranial haemorrhage in patients with atrial fibrillation and the impact on recurrent stroke, mortality and bleeding: a nationwide cohort study.
      After a hemorrhagic stroke, it is it is unclear whether it should count one point (either for stroke or bleeding) or two points (one point each for stroke and bleeding) on the HAS-BLED score. We investigated the value of a recalibration of the HAS-BLED score to account for two points from a hemorrhagic stroke.

      Methods and Materials

      Data from three Danish nationwide registries were used to conduct an observational cohort study investigating the original HAS-BLED score and the recalibrated HAS-BLED score. As previously performed, we identified patients with nonvalvular AF discharged from the hospital from January 1, 1999, to December 31, 2013, and excluded patients who experienced a thromboembolism or major bleeding within 7 days after discharge.
      • Lip G.Y.
      • Nielsen P.B.
      • Skjøth F.
      • Lane D.A.
      • Rasmussen L.H.
      • Larsen T.B.
      The value of the European Society of Cardiology guidelines for refining stroke risk stratification in patients with atrial fibrillation categorized as low risk using the anticoagulation and risk factors in atrial fibrillation stroke score: a nationwide cohort study.
      The primary end point was major bleeding, defined as a composite of intracranial bleeding (including traumatic intracranial bleeding events), gastrointestinal bleeding, acute anemia, bleeding from the urinary tract, and airway bleeding (e-Table 1 presents the relevant International Classification of Diseases, 10th Revision, codes). Both primary and secondary diagnoses were included, but emergency department diagnoses were not included because of low validity.
      • Johnsen S.P.
      • Overvad K.
      • Sørensen H.T.
      • Tjønneland A.
      • Husted S.E.
      Predictive value of stroke and transient ischemic attack discharge diagnoses in the Danish National Registry of Patients.
      Patients were followed up from 7 days after hospital discharge and up to 1 year later or to occurrence of death, a major bleeding event, or end-of-study period, whichever came first.
      Two different HAS-BLED scores (original and recalibrated) were calculated at hospital discharge; only patients with a hemorrhagic stroke (nontraumatic intracranial bleeds) were reclassified in the recalibrated HAS-BLED score. Crude event rates (total number of events divided by accrued person-time) stratified according to scores ranging from 0 to 8 were reported. No information was available on INR values; hence the “L” component was excluded from the calculations. To compare the predictive power of the scores, the (Harrell’s) C-statistics were calculated and compared.
      • Newson R.
      Comparing the predictive powers of survival models using Harrell’s C or Somers' D.
      Estimates of bleeding risk were obtained in a competing risk of death setting by using information directly from the cumulative incidence function.
      • Wolbers M.
      • Koller M.T.
      • Witteman J.C.
      • Steyerberg E.W.
      Prognostic models with competing risks: methods and application to coronary risk prediction.
      Use of competing risk analyses are advised, especially in an elderly and fragile population, to obtain adequate risk estimates that are not biased due to the competing risk.
      • Koller M.T.
      • Schaer B.
      • Wolbers M.
      • Sticherling C.
      • Bucher H.C.
      • Osswald S.
      Death without prior appropriate implantable cardioverter-defibrillator therapy: a competing risk study.
      To further compare the individual-level changes in risk assessment from the two scores, a net reclassification index (NRI) was calculated, also in a competing risk setting.
      • Pencina M.J.
      • D’Agostino R.B.
      • Steyerberg E.W.
      Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers.
      • Klein J.P.
      • Andersen P.K.
      Regression modeling of competing risks data based on pseudovalues of the cumulative incidence function.
      Briefly, the NRI evaluates the proportion of patients with a correct/incorrect change in risk according to being a case (patient with an event) or a noncase. Finally, the integrated discrimination improvement (IDI) was calculated relative to the original HAS-BLED score to assess the separation in predicted risk for events and nonevents.
      • Pencina M.J.
      • D’Agostino R.B.
      • Vasan R.S.
      Statistical methods for assessment of added usefulness of new biomarkers.
      Bootstrap CIs for C-statistics, NRI, and IDI were calculated by using 100 bootstrap samples. Data were analyzed by using Stata version 13.1 (Stata Corporation). Register-based studies do not require ethical approval in Denmark.
      Three sensitivity analyses were performed. The first analysis stratified patients according to baseline OAC treatment or initiation within the first 7 days after baseline. This method was chosen in an attempt to indirectly assess if the missing “L” component from the data could influence the results (recognizing that patients not treated with OAC would not contribute to this component). The second analysis restricted the follow-up time to 6 months to assess performance of the recalibrated HAS-BLED score compared with the original HAS-BLED score in a relatively short follow-up time. The motivation for this analysis was established by conceding that an intracranial hemorrhage can be associated with increased event rates of recurrent major bleeding (including recurrent intracranial bleeding).
      • Aguilar M.I.
      • Hart R.G.
      • Kase C.S.
      • et al.
      Treatment of warfarin-associated intracerebral hemorrhage: literature review and expert opinion.
      In the final analysis, the study period was restricted to the five most recent calendar years (2009-2013). Given the increasing availability of imaging technologies (MRI/CT scans) in this period, the validity of hemorrhagic stroke and/or hemorrhagic bleeding diagnosis was deemed to increase; hence, the proportion of reclassified patients (based on hemorrhagic stroke) would be more accurate.

      Results

      The study population comprised 210,299 patients with AF (5,898 patients were excluded due to a thromboembolic/bleeding event within 7 days after discharge); they had a median age of 74 years (interquartile range, 65-82), and 46.6% were women. During 1 year of follow-up, 7,602 (3.62%) bleeding events were observed. The mean HAS-BLED scores were 2.13 for the original score and 2.14 for the recalibrated score. The overall crude event rate of the primary end point was 4.3 per 100 person-years; Table 1 displays the event rates stratified according to points for the two scores. A total of 1,479 patients were reclassified in the recalibrated HAS-BLED score analysis based on having sustained a hemorrhagic stroke when the observation time commenced. The predictive accuracy of the two scores in terms of the C-statistics was moderate: 0.613 (95% CI, 0.607-0.619) for the original HAS-BLED score and 0.616 (95% CI, 0.610-0.622) for the recalibrated HAS-BLED score, respectively. The NRI was 10.0% (95% CI, 7.6-12.4), displaying a significant improvement of correct classification by the recalibrated HAS-BLED scored compared with the original score. The relative IDI was 23.6% (95% CI, 15.7-31.5), reflecting that the recalibrated HAS-BLED score more accurately predicted bleeding events.
      Table 1Rates Per 100 Person-Years for Major Bleeding Events According to HAS-BLED Scores for 1 Year of Follow-up in Patients With Incident AF
      PointsOriginal HAS-BLED ScoreRecalibrated HAS-BLED Score
      Person-TimeEventsEvent RatePerson-TimeEventsEvent Rate
      019,6111930.9819,6111930.98
      141,8641,1272.6941,7991,1112.69
      251,9332,1794.2051,7032,1494.16
      343,7692,3925.4643,6462,3685.43
      417,3591,2947.4517,4721,3167.53
      53,6923619.783,96540210.14
      64255212.254565912.94
      722418.2323417.36
      8120220
      9 (labile INR)NANANANANANA
      Any score178,6767,6024.3
      Categorized risk
       0-2 (low risk)113,4083,4993.09113,1133,4533.05
       ≥3 (high risk)65,2684,1036.2965,5634,1496.33
      Observation time starts after a quarantine period of 7 d relative to atrial fibrillation (AF) discharge. HAS-BLED = hypertension, abnormal renal/liver function (one or two points), stroke, bleeding history or predisposition, labile international normalized ratio (INR), elderly (> 65 y), drugs/alcohol concomitantly [one or two points]); NA = not available.
      When stratifying the cohort to patients who received no OAC treatment vs those treated, there was an increase in the bleeding rates per 100 person-years: 3.97 for no treatment and 4.73 for OAC treated. The sensitivity analysis resulted in similar NRI findings; for OAC-treated patients, it was 9.0% (95% CI, 4.8-12.3), and for untreated patients, it was 10.1% (95% CI, 9.0-11.2). The C-statistics were likewise unaffected by this stratification (data not shown). When applying a shorter follow-up period (ie, 6 months), a considerably higher NRI of 35.0% (95% CI, 32.9-37.1) was observed. The C-statistics for the original and recalibrated HAS-BLED scores, however, displayed modest increases (0.617 and 0.621, respectively). Restricting the study period to the last 5 years reduced the study population to 78,699 patients and essentially doubled the NRI (ie, 20.7% [95% CI, 17.9-23.5]).

      Discussion

      In this nationwide cohort study reflecting clinical practice, the original HAS-BLED score was recalibrated to account for two points if the stroke type was hemorrhagic. The recalibrated HAS-BLED score exhibited improved accuracy, as displayed by an improved NRI of 10% and a relative IDI of 23%.
      Although the recalibration resulted in only a modest increase in the C-statistics, the potential for improved accuracy is higher. Only a very small proportion of this large cohort was reclassified: of 2,218 patients who sustained an intracranial hemorrhage event, 1,479 (0.7% of the total study population) were reclassified. The remaining 739 patients who were not reclassified had all experienced previous events (ischemic stroke/transient ischemic attack and a major bleeding event [or anemia]).
      The recalibrated HAS-BLED score could potentially be more complicated to count than the original score. The physician who is presented with a patient who has sustained a hemorrhagic stroke must take into account whether this patient already has one point attributed to prior bleeding and prior stroke. Conversely, counting a hemorrhagic stroke as two points (one for stroke and one for bleeding) appears intuitively applicable given the nature of the outcome. Important to note, the recalibrated score performs best within a short follow-up time in terms of NRI, which could be related to the patients who were actually reclassified (those with a hemorrhagic stroke); however, this finding could represent the early excess risk associated with such high-risk patients.
      The present study has some limitations. We did not have access to INR values from the registries; hence, we excluded a potential pivotal component from the HAS-BLED score. An observational design was used to assess the HAS-BLED score, but we cannot rule out miscoding of comorbidity and concomitant medication. In addition, we did not have access to imaging data, and we cannot rule out erroneously coded hemorrhagic strokes. The HAS-BLED score was assessed at baseline; however, risk assessment in patients with AF is a continuum, and risk (both for bleeding and thromboembolic events) does not remain static in these patients.

      Conclusions

      Recalibration of the HAS-BLED score, counting two points for a hemorrhagic stroke, resulted in an improved accuracy of predicting major bleeding events. This approach could potentially aid physicians in a more accurate assessment of bleeding risk in patients with AF.

      Acknowledgments

      Financial/nonfinancial disclosures: The authors have reported to CHEST the following: G. Y. H. L. has served as a consultant for Bayer, Astellas, Merck, AstraZeneca, Sanofi, BMS/Pfizer, and Boehringer Ingelheim; and has been on the speakers bureaus for Bayer, BMS/Pfizer, Boehringer Ingelheim, and Sanofi. T. B. L. has served as an investigator for Janssen Scientific Affairs, LLC and Boehringer Ingelheim; and has been on the speakers bureaus for Bayer, BMS/Pfizer, Roche Diagnostics, Boehringer Ingelheim, and Takeda Pharma. None declared (P. B. N.).
      Role of sponsors: The sponsor had no role in the design of the study, the collection and analysis of the data, or the preparation of the manuscript.
      Additional information: e-Table 1 can be found in the Supplemental Materials section of the online article.

      Supplementary Data

      References

        • Hart R.G.
        • Pearce L.A.
        • Aguilar M.I.
        Meta-analysis: antithrombotic therapy to prevent stroke in patients who have nonvalvular atrial fibrillation.
        Ann Intern Med. 2007; 146: 857-867
        • Camm A.J.
        • Lip G.Y.
        • De Caterina R.
        • et al.
        • and the ESC Committee for Practice Guidelines-CPG; Document Reviewers
        2012 Focused update of the ESC guidelines for the management of atrial fibrillation: an update of the 2010 ESC guidelines for the management of atrial fibrillation—developed with the special contribution of the European Heart Rhythm Association.
        Europace. 2012; 14: 1385-1413
        • January C.T.
        • Wann F.L.
        • Joseph F.
        • et al.
        2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: executive summary.
        J Am Coll Cardiol. 2014; 64: 2246-2280
        • Camm A.J.
        • Kirchhof P.
        • Lip G.Y.
        • et al.
        Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC).
        Eur Heart J. 2010; 31: 2369-2429
        • Pisters R.
        • Lane D.A.
        • Nieuwlaat R.
        • de Vos C.B.
        • Crijns H.J.
        • Lip G.Y.
        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: 1093-1100
        • Marsh E.B.
        • Gottesman R.F.
        Brain hemorrhage: restarting anticoagulation after intracranial hemorrhage.
        Nat Rev Neurol. 2011; 7: 130-132
        • Nielsen P.B.
        • Larsen T.B.
        • Skjøth F.
        • Gorst-Rasmussen A.
        • Rasmussen L.H.
        • Lip G.Y.
        Restarting anticoagulant treatment after intracranial haemorrhage in patients with atrial fibrillation and the impact on recurrent stroke, mortality and bleeding: a nationwide cohort study.
        Circulation. 2015; 132: 517-525
        • Lip G.Y.
        • Nielsen P.B.
        • Skjøth F.
        • Lane D.A.
        • Rasmussen L.H.
        • Larsen T.B.
        The value of the European Society of Cardiology guidelines for refining stroke risk stratification in patients with atrial fibrillation categorized as low risk using the anticoagulation and risk factors in atrial fibrillation stroke score: a nationwide cohort study.
        Chest. 2014; 146: 1337-1346
        • Johnsen S.P.
        • Overvad K.
        • Sørensen H.T.
        • Tjønneland A.
        • Husted S.E.
        Predictive value of stroke and transient ischemic attack discharge diagnoses in the Danish National Registry of Patients.
        J Clin Epidemiol. 2002; 55: 602-607
        • Newson R.
        Comparing the predictive powers of survival models using Harrell’s C or Somers' D.
        Stata J. 2010; 10: 339-358
        • Wolbers M.
        • Koller M.T.
        • Witteman J.C.
        • Steyerberg E.W.
        Prognostic models with competing risks: methods and application to coronary risk prediction.
        Epidemiology. 2009; 20: 555-561
        • Koller M.T.
        • Schaer B.
        • Wolbers M.
        • Sticherling C.
        • Bucher H.C.
        • Osswald S.
        Death without prior appropriate implantable cardioverter-defibrillator therapy: a competing risk study.
        Circulation. 2008; 117: 1918-1926
        • Pencina M.J.
        • D’Agostino R.B.
        • Steyerberg E.W.
        Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers.
        Stat Med. 2011; 30: 11-21
        • Klein J.P.
        • Andersen P.K.
        Regression modeling of competing risks data based on pseudovalues of the cumulative incidence function.
        Biometrics. 2005; 61: 223-229
        • Pencina M.J.
        • D’Agostino R.B.
        • Vasan R.S.
        Statistical methods for assessment of added usefulness of new biomarkers.
        Clin Chem Lab Med. 2010; 48: 1703-1711
        • Aguilar M.I.
        • Hart R.G.
        • Kase C.S.
        • et al.
        Treatment of warfarin-associated intracerebral hemorrhage: literature review and expert opinion.
        Mayo Clin Proc. 2007; 82: 82-92

      Linked Article

      • Deep vs Lobar Intracerebral Hemorrhage on HAS-BLED Scoring System
        ChestVol. 149Issue 6
        • Preview
          We have read with great interest the article entitled “Recalibration of the HAS-BLED Score: Should Hemorrhagic Stroke Account for One or Two Points?” in a recent issue of CHEST (February 2016).1 The authors noted that after a hemorrhagic stroke, it is unclear whether it should count 1 point (either for stroke or bleeding) or 2 points (1 point each for stroke and bleeding) on the bleeding risk score HAS-BLED (hypertension, abnormal renal/liver function [1 or 2 points], stroke, bleeding history or predisposition, labile international normalized ratio, elderly [> 65 years], drugs/alcohol concomitantly [1 or 2 points]).
        • Full-Text
        • PDF
      • Response
        ChestVol. 149Issue 6
        • Preview
          We thank Ince and Senel1 for their interest in our research. The objective of our commentary article2 was to clarify an opacity incorporated into the bleeding risk score HAS-BLED (hypertension, abnormal renal/liver function [1 or 2 points], stroke, bleeding history or predisposition, labile international normalized ratio, elderly [> 65 years], drugs/alcohol concomitantly [1 or 2 points]). This question is encountered in everyday clinical practice: does a hemorrhagic stroke count as a bleeding event in addition to the stroke event? Using the well-validated nationwide Danish registries, we found that the recalibrated HAS-BLED score more accurately predicted major bleeding in patients sustaining an intracranial hemorrhage (ICH).
        • Full-Text
        • PDF