Validation of a score for predicting bleeding events during acute coronary syndromes

Abstracts

FUNDAMENTO: Sangramento é uma complicação grave em pacientes tratados para síndromes coronarianas agudas (SCA) com antitrombóticos e terapias invasivas. Consequentemente, o benefício dessas terapias deve ser analisado contra os potenciais riscos de complicações hemorrágicas. Desta forma, a determinação de um escore para estimar o risco individual de sangramento pode representar uma importante ferramenta na tomada de decisões clínicas. OBJETIVO: Criar e validar um escore de risco de sangramento para pacientes com SCA. MÉTODOS: Foram utilizados preditores independentes de sangramento relatados pelo Registro GRACE. Variáveis com odds ratio (OR) > 2,5 nesse Registro somaram 3 pontos (histórico anterior de sangramento), OR=1,5-2,4 somaram 2 pontos (clearance de creatinina< 30 ml/min, sexo feminino) e aquelas com OR < 1,5 somaram 1 ponto (clearance entre 30 e 60 ml/min, a cada 10 anos de idade > 30, infra ou supra-desnivelamento do segmento ST, doença arterial periférica e fumo). O escore foi validado em uma coorte de 383 indivíduos com SCA. Sangramento hospitalar foi definido como queda de hematócrito > 10%, transfusão de sangue > 2 unidades, sangramento intracerebral ou sangramento fatal. RESULTADOS: A incidência de eventos hemorrágicos foi de 3,1% e a estatística-C do escore foi 0,66 (IC95% = 0,52-0,80), indicando capacidade preditiva para esses eventos. Aqueles com escore > 7 apresentaram 6% de incidência de sangramento, comparados com 1,9% se o escore era < 7 (RR=3,2; IC95%=1,04-9,9; p=0,03). Houve uma interação entre um escore > 7 e um maior risco imposto pelo tratamento com Clopidogrel (p=0,02), bloqueadores IIb/IIIa (p=0,06) e revascularização cirúrgica (p < 0,001). CONCLUSÃO: O escore discrimina o risco de sangramento e é potencialmente útil na tomada de decisão clínica em SCA.

Hemorragia; risco; síndrome coronariana aguda

FUNDAMENTO: Sangrado es una complicación grave en pacientes tratados por síndromes coronarios agudos (SCA) con antitrombóticos y terapias invasivas. Consecuentemente, el beneficio de esas terapias debe ser analizado contra los potenciales riesgos de complicaciones hemorrágicas. De esta forma, la determinación de un escore para estimar el riesgo individual de sangrado puede representar una importante herramienta en la toma de decisiones clínicas. OBJETIVO: Crear y validar un escore de riesgo de sangrado para pacientes con SCA. MÉTODOS: Fueron utilizados predictores independientes de sangrado relatados por el Registro GRACE. Variables con odds ratio (OR) > 2,5 en ese Registro sumaron 3 puntos (histórico anterior de sangrado), OR=1,5-2,4 sumaron 2 puntos (clearance de creatinina < 30 ml/min, sexo femenino) y aquellas con OR < 1,5 sumaron 1 punto (clearance = 30-60 ml/min, a cada 10 años de edad > 30, infra o supra-desnivel del segmento ST, enfermedad arterial periférica y tabaco). El escore fue validado en una cohorte de 383 individuos con SCA. Sangrado intrahospitalario fue definido como caída de hematocrito > 10%, transfusión de sangre > 2 unidades, sangrado intracerebral o sangrado fatal. RESULTADOS: La incidencia de eventos hemorrágicos fue de 3,1% y la estadística-C del escore fue 0,66 (IC95% = 0,52-0,80), indicando capacidad predictiva para esos eventos. Aquellos con escore > 7 presentaron 6% de incidencia de sangrado, comparados con 1,9% si el escore era < 7 (RR=3,2; IC95%=1,04-9,9; P=0,03). Hubo una interacción entre un escore > 7 y un mayor riesgo impuesto por el tratamiento con Clopidogrel (P=0,02), bloqueadores IIb/IIIa (P=0,06) y revascularización quirúrgica (P<0,001). CONCLUSIÓN: El escore discrimina el riesgo de sangrado y es potencialmente útil en la toma de decisión clínica en SCA.

Hemorragia; riesgo; síndrome coronario agudo

ORIGINAL ARTICLE

Escola Bahiana de Medicina, Hospital Português, Salvador, BA - Brasil

Mailing address

ABSTRACT

BACKGROUND: Bleeding is a major complication in patients treated for acute coronary syndromes (ACS) with antithrombotic and invasive therapies. Consequently, the benefit of such therapies should be balanced against the potential risk of hemorrhagic complications. Therefore, a score to estimate individual risk of bleeding might represent an important tool in clinical decision-making.

OBJECTIVE: This study aims to create and validate a bleeding risk score for patients with ACS.

METHODS: Independent predictors of bleeding reported by the GRACE Registry were utilized. Variables with odds ratio (OR) > 2.5 in that Registry added 3 points (previous history of bleeding), OR = 1.5-2.4 added 2 points (creatinine clearance < 30 ml/min, female gender) and those with OR < 1.5 added 1 point (clearance between 30 and 60 ml/min, each 10 years of age > 30, ST-deviation, peripheral artery disease and smoking). The score was validated in a cohort of 383 individuals with ACS. In-hospital bleeding was defined as hematocrit fall > 10%, blood transfusion > 2 units, intracerebral bleeding or fatal bleeding.

RESULTS: The incidence of bleeding events was 3.1% and the score's C-statistics was 0.66 (95% CI = 0.52-0.80), indicating a predictive ability towards these events. Those with a score > 7 had 6% incidence of bleeding, compared with 1.9% if the score was < 7 (RR = 3.2; 95%CI = 1.04-9.9; p = 0.03). There was an interaction between a score > 7 and greater risk imposed by treatment with Clopidogrel (p = 0.02), IIb/IIIa blockers (p = 0.06) and surgical revascularization (p < 0.001).

CONCLUSION: The score discriminates bleeding risk and is potentially useful in clinical decision-making during ACS.

Key words: Hemorrhage; risk; acute coronary syndrome.

Introduction

Acute coronary syndromes (ACS) are associated with significant rates of recurrent ischemic events in the short^1,2 and long-term³. To reduce the chance of recurrent events, antithrombotic therapy and invasive procedures are indicated, at the expense of increasing the risk of bleeding complications^4,5. In this context, independent predictors of bleeding have been reported by several studies^6-9. In order to apply this knowledge to clinical practice, these predictors can be used to create a score that allows physicians to estimate individual risk of bleeding.

Some facts suggested that a bleeding score would be useful in clinical decision-making during ACS. Bleeding is associated with a significant rise in mortality, varying from 6% to 17% of absolute increase in short-term death according to different studies^7-9; the very same treatments (antithrombotic therapy and invasive coronary procedures) intended to prevent recurrent ischemic events, paradoxically increase the incidence of bleeding^4,5. Finally, the withdrawn of antithrombotic therapy during bleeding events enhances patients vulnerability to ischemic events⁹. Therefore, it is important to compare the risk of ischemic events and the risk of bleeding events in each individual, in order to define how aggressive the treatment should be. For instance, those with low bleeding-risk and high ischemic-risk should benefit from aggressive antithrombotic therapy, while those with high bleeding-risk and low ischemic-risk should not receive intense antithrombotic therapy.

The present study created and tested a score to estimate the chance of bleeding complications in individuals admitted with ACS. The score was built based on previous literature regarding bleeding predictors and was validated in a local cohort of patients hospitalized with unstable angina, non-ST or ST-segment elevation acute myocardial infarction.

Methods

Sample selection

Consecutive patients with acute coronary syndromes, at least 18 years of age, admitted in the coronary care unit of our Hospital between 1999 and 2007 were considered candidates for the study. Inclusion criteria were defined as chest pain or equivalent within 24 hours of admission and at least one of the following characteristics: 1) electrocardiographic ischemic changes, defined as dynamic T wave inversion, ST-segment depression or ST-segment elevation; 2) elevation of serum markers of myocardial necrosis; 3) previous documentation of coronary artery disease. All participants provided written informed consent and the protocol was approved by the Institution's human research ethic committee.

Bleeding score

To create the bleeding score, published data regarding predictors of bleeding in the GRACE registry were utilized⁹. Based on that registry, the independent predictors of bleeding were included in the score, according to the following rule: 1 point if the odds ratio was < 1.5 in the Registry, 2 points if the odds ratio was between 1.5 and 2.4, and 3 points when the odds ratio was > 2.5. Since it was intended to be an admission score, only variables typically available at admission were considered. Thus, the score consisted of the following variables: history of bleeding (3 points), creatinine clearance < 30 ml/min (2 points), creatinine clearance = 30 - 59 ml/min (1 point), female gender (1 point), peripheral artery disease (1 point), smoking (1 point) and ST-segment deviation (1 point) and each decade of age older than 30 years old (1 point). Creatinine clearance was calculated according to the Cockcroft-Gault formula¹⁰.

Endpoint sefinition

Clinically manifested hemorrhages were defined as a bleeding event for the present analysis when at least one of the major criteria was present: hematocrit fall > 10%, transfusion requirement > 2 units, cerebral hemorrhage or fatal hemorrhage⁹. Thus, only major bleeding events were taken into consideration. Recurrent ischemic events during hospitalization were defined as the composite of nonfatal acute myocardial infarction or recurrent unstable angina. Myocardial infarction as an outcome endpoint was defined as either a new Q-wave or troponin elevation during hospitalization despite normal values during the first 24 hours. For patients with infarction at admission, a new peak of mass CK-MB (> 50% the previous value and above the normal value) was required for diagnosis of reinfarction. An episode of rest angina was recorded as recurrent unstable angina if one of the following criteria was present: physician decision to administer sublingual nitrates or initiate intravenous nitrates, dynamic electrocardiographic changes and physician decision to perform emergency coronary angiography. In-hospital death was also recorded as a cardiovascular endpoint.

Data analysis

The discriminatory ability of the bleeding score was tested by C-statistics, defined as the area under the receiver operator characteristic (ROC) curve, and calibration (observed vs. predicted risk) was assessed by Hosmer-Lemeshow statistic. Based on the ROC curve, the most accurate cut-off point for bleeding prediction was identified. Then, individuals were dichotomized according to this cut-off and incidence of bleeding compared between the two groups, by Chi-square test. In addition, the incidence of bleeding was compared among the three tertiles of the bleeding score and p for trend was obtained through linear-by-linear Chi-square test. In order to evaluate whether the association between the score and bleeding events depended on treatment strategies acting as confounding variables, pharmacologic and invasive treatments were compared between the groups dichotomized by the score, using Chi-square test. Moreover, logistic regression analysis was used to evaluate interaction between the bleeding score and treatment.

The GRACE score for in-hospital mortality was calculated for all patients¹¹. Spearman's correlation coefficient was utilized to test the association between the GRACE score and the bleeding score. Then, the GRACE score was dichotomized using the cutoff of 141, which defines high-risk individuals according to previous literature¹¹. Finally, in-hospital death and recurrent ischemic events were compared between individual with bleeding events and those free of hemorrhage by Chi-square test.

Ordinal variables (scores) were expressed as medians (interquartile range) and continuous variables as means and standard deviation. For baseline characteristics analysis, continuous variables were compared between two groups by Student's t test and categorical variables by Chi-square test. When expected values in the association of two categorical values were < 1, Fisher's exact test was used instead of Chi-square test. The SPSS software package release 10.0 was used for data analysis and a p value < 0.05 was considered statistically significant.

Results

Sample population

Three hundred and eighty-three patients were studied, aged 66 ± 12 years old, 42% females, 21% admitted with ST-segment elevation infarction, 29% with non-ST elevation infarction and the remaining with unstable angina. The incidence of major bleeding during hospitalization was 3.1%. The GRACE Score for in-hospital mortality had a median of 111 (interquartile range: 92 - 133), indicating that half of the patients were considered at least intermediate risk for ischemic complications (GRACE Score > 109). Clinical characteristics are depicted on Table 1.

Predictive ability of the bleeding score

The bleeding score had a median of 5 (interquartile range: 4 - 7). Considering the components of the score, creatinine clearance between 30 and 60 ml/min, female gender and ST-segment deviation were the most prevalent, peripheral artery disease and smoking were less frequent (< 20% each), while previous bleeding and clearance < 30 ml/min were rare (< 3%) - Table 1. The score's C-statistics was 0.66 (95% CI: 0.52 - 0.80), indicating a significant prediction of bleeding events during hospitalization - Figure 1. The point presenting the best predictive accuracy was 7, and the incidence of in-hospital major bleeding in individuals with a score > 7 was 6%, compared with 1.9% in those with a score < 7 (relative risk = 3.2; 95% CI = 1.04 - 9.9; p = 0.03). When the sample population was divided into tertiles of the bleeding score, the incidences in the first, second and third tertiles were progressively higher (1.6%, 2.7% and 6%, respectively; p for trend = 0.03). The Hosmer-Lemeshow p value for the score was 0.49, indicating good calibration.

Major bleeding and treatment approach according to the score

Pharmacological and interventional treatments were similar between the groups of score > 7 and < 7, indicating that the predictive ability of the score was independent from the established treatment - Table 2. To evaluate whether the association between bleeding events and antithrombotic or invasive treatment was modified (or predicted) by the score, interaction terms were calculated by logistic regression. Interaction between treatment and score > 7 was significant for Clopidogrel (p = 0.02) and surgical revascularization (p < 0.001) and a trend was observed with GP IIb/IIIa receptor blockers (p = 0.06). Indeed, only individuals with a score > 7 experienced bleeding increase with these drugs - Figure 3. Seemingly, compared with non-surgical patients, surgery was associated with a 15-fold increase in bleeding events in individuals with a score > 7, while the increase was only 3-fold if the patient had a score < 7 - Figure 2. Conversely, there was no score interaction with thrombolytic therapy (p = 0.16) and angioplasty (p = 0.23).

Relationship between bleeding prediction and ischemic death

When the bleeding risk is different from the ischemic risk, there is a clear-cut definition of the best approach in relation to anti-ischemic therapies that increase bleeding events. Thus, it is important to evaluate the bleeding score in relation to the GRACE score. There was a significant positive correlation between the bleeding score and the GRACE score for in-hospital mortality (R = 0.59; p < 0.001) - Figure 3. Sixty-one percent of patients were defined as presenting both low risk of bleeding (score > 7) and low risk of ischemic death (GRACE score > 140), while 11.5% were classified as high-risk by both scores. Conversely, the two scores disagreed in 28% of individuals, 18% were defined as presenting high risk of bleeding and low risk of ischemic death and 9.5% as presenting low risk of bleeding and high risk of ischemic death - Figure 2.

Bleeding as a risk factor for ischemic events

Individuals who experienced major bleeding events during hospitalization had a higher incidence of recurrent ischemic events, in relation to those free of hemorrhage (50% vs 23%, p = 0.03). In accordance, in-hospital death was more frequent in individuals with major bleeding (25% vs 3.5%, p = 0.01).

Discussion

The present study proposed a score to estimate the risk of bleeding in individuals hospitalized with ACS. The score presented a reasonable performance regarding discrimination and a good calibration, as individuals with a score > 7 had a 3-fold higher incidence of bleeding events. Moreover, the interaction analysis showed that the score was able to discriminate individuals that were vulnerable to therapy-related bleeding.

Compared to well-validated risk scores, the discriminatory ability of the present score, measured by a C-statistics of 0.66, should be considered acceptable. For instance, a recently validated score for the prediction of bleeding during percutaneous coronary intervention was considered useful based on a C-statistics of 0.62¹²; the TIMI-Risk Score has a C-statistics of 0.68 for mortality in ACS; and the C-statistics of the Framingham Score varies between 0.65 and 0.85 according to the population studied^13,14. On the other hand, further studies are necessary to identify new predictors capable of improving the current ability of the bleeding score.

A direct correlation between the bleeding score and the GRACE ischemic mortality score was observed, probably due to common variables between the two, such as age, renal function and ST-deviation. Regardless of that, according to these scores, about one-third of the patients presented a high risk of bleeding with a low risk of ischemia or vice-versa. The scores present a clear-cut definition of risk-benefit for these patients. For the remaining patients, more elaborate decision trees should be used based on the scores' risk estimation.

In the present study, the bleeding incidence of 3.1% was similar to what has been reported in different cohorts^6-9,15,16, indicating the external validity of our sample population. Studies differ in relation to major bleeding definitions. Some are conservative and define major bleeding depending only on clinical manifestation, such as the GUSTO criteria. Others are more liberal and use laboratory criteria together with clinical criteria¹⁷. We chose to define major bleeding by both clinical (fatal, cerebral hemorrhage or need for transfusion) and laboratory criteria (hematocrit fall > 10%), as this definition of bleeding is more sensitive to detect this undesirable complication.

Several studies have reported independent predictors of bleeding^6-9,15,16. We chose to use the data provided by the GRACE registry⁹, because it was the largest cohort to report such results and it is a pure observational registry, as opposed to some cohorts originated from clinical trials^6,7. Despite some variation among studies, in general the cohorts provide similar results regarding predictors. These studies have reported antithrombotic and invasive treatments as predictors of bleeding. However, we chose not to include treatment variables in score calculation, because as a clinical tool, it is more useful to determine baseline risk at admission, before treatment choices have been made.

Study limitations should be recognized. Our sample was large enough to provide a sufficient certainty that the results were not driven by chance, but the frequency of bleeding events in each subgroup defined by the score should not be seen as a precise estimation, because of limited sample size. For instance, we reported that a high-risk score implies a chance of 6% in bleeding. However, the 95% confidence interval for this incidence is 2.7% to 12%. In this regard, larger studies are necessary to better quantify absolute risk related to score results. Secondly, due to the small sample size, we could not stratify individuals into more than three groups of risk. The score will be more useful if it stratifies patients in several groups of bleeding risk, with a wider range of risk variation and much greater risk in the extremities. In other words, when larger samples are studied, subgroups of higher risk than 6% will be identified. Thirdly, although the beneficial effect of the score is very plausible, the efficacy of the score in clinical decision-making remains to be proved by randomized clinical trials.

Conclusion

The present study validates a score to determine bleeding risk during hospitalization for ACS and suggests the use of this tool as a means to optimize therapeutic decisions targeted to reduce ischemic events.

Potential Conflict of Interest

No potential conflict of interest relevant to this article was reported.

Sources of Funding

There were no external funding sources for this study.

Study Association

This study is not associated with any post-graduation program.

References

• 1. Steg PG, Goldberg RJ, Gore JM, Fox KA, Eagle KA, Flather MD, et al. Baseline characteristics, management practices, and in-hospital outcomes of patients hospitalized with acute coronary syndromes in the Global Registry of Acute Coronary Events (GRACE). Am J Cardiol. 2002; 90 (4): 358-63.
• 2. Hochman JS, McCabe CH, Stone PH, Becker RC, Cannon CP, DeFeo-Fraulini T, et al. Outcome and profile of women and men presenting with acute coronary syndromes: a report from TIMI IIIB. TIMI Investigators. Thrombolysis in Myocardial Infarction. J Am Coll Cardiol. 1997; 30 (1): 141-8.
• 3. Eagle KA, Lim MJ, Dabbous OH, Pieper KS, Goldberg RJ, Van de WF, et al. A validated prediction model for all forms of acute coronary syndrome: estimating the risk of 6-month postdischarge death in an international registry. JAMA. 2004; 291 (22): 2727-33.
• 4. Anderson JL, Adams CD, Antman EM, Bridges CR, Califf RM, Casey DE Jr, et al. ACC/AHA 2007 guidelines for the management of patients with unstable angina/non ST-elevation myocardial infarction: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 2002 Guidelines for the Management of Patients With Unstable Angina/Non ST-Elevation Myocardial Infarction): developed in collaboration with the American College of Emergency Physicians, the Society for Cardiovascular Angiography and Interventions, and the Society of Thoracic Surgeons: endorsed by the American Association of Cardiovascular and Pulmonary Rehabilitation and the Society for Academic Emergency Medicine. Circulation. 2007; 116 (7): e148-e304.
• 5. Antman EM, Hand M, Armstrong PW, Bates ER, Green LA, Halasyamani LK, et al. 2007 Focused Update of the ACC/AHA 2004 Guidelines for the Management of Patients With ST-Elevation Myocardial Infarction: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines: developed in collaboration With the Canadian Cardiovascular Society endorsed by the American Academy of Family Physicians: 2007 Writing Group to Review New Evidence and Update the ACC/AHA 2004 Guidelines for the Management of Patients With ST-Elevation Myocardial Infarction, Writing on Behalf of the 2004 Writing Committee. Circulation. 2008; 117 (2): 296-329.
• 6. Eikelboom JW, Mehta SR, Anand SS, Xie C, Fox KA, Yusuf S. Adverse impact of bleeding on prognosis in patients with acute coronary syndromes. Circulation. 2006; 114 (8): 774-82.
• 7. Manoukian SV, Feit F, Mehran R, Voeltz MD, Ebrahimi R, Hamon M, et al. Impact of major bleeding on 30-day mortality and clinical outcomes in patients with acute coronary syndromes: an analysis from the ACUITY Trial. J Am Coll Cardiol. 2007; 49 (12): 1362-8.
• 8. Segev A, Strauss BH, Tan M, Constance C, Langer A, Goodman SG. Predictors and 1-year outcome of major bleeding in patients with non-ST-elevation acute coronary syndromes: insights from the Canadian Acute Coronary Syndrome Registries. Am Heart J. 2005; 150 (4): 690-4.
• 9. Spencer FA, Moscucci M, Granger CB, Gore JM, Goldberg RJ, Steg PG, et al. Does comorbidity account for the excess mortality in patients with major bleeding in acute myocardial infarction? Circulation. 2007; 116 (24): 2793-801.
• 10. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron. 1976; 16 (1): 31-41.
• 11. Granger CB, Goldberg RJ, Dabbous O, Pieper KS, Eagle KA, Cannon CP, et al. Predictors of hospital mortality in the global registry of acute coronary events. Arch Intern Med. 2003; 163 (19): 2345-53.
• 12. Nikolsky E, Mehran R, Dangas G, Fahy M, Na Y, Pocock SJ, et al. Development and validation of a prognostic risk score for major bleeding in patients undergoing percutaneous coronary intervention via the femoral approach. Eur Heart J. 2007; 28 (16): 1936-45.
• 13. Antman EM, Cohen M, Bernink PJ, McCabe CH, Horacek T, Papuchis G, et al. The TIMI risk score for unstable angina/non-ST elevation MI: a method for prognostication and therapeutic decision making. JAMA. 2000; 284 (7): 835-42.
• 14. Prediction of mortality from coronary heart disease among diverse populations: is there a common predictive function? Heart. 2002; 88 (3): 222-8.
• 15. Rao SV, O'Grady K, Pieper KS, Granger CB, Newby LK, Van de WF, et al. Impact of bleeding severity on clinical outcomes among patients with acute coronary syndromes. Am J Cardiol. 2005; 96 (9): 1200-6.
• 16. Moscucci M, Fox KA, Cannon CP, Klein W, Lopez-Sendon J, Montalescot G, et al. Predictors of major bleeding in acute coronary syndromes: the Global Registry of Acute Coronary Events (GRACE). Eur Heart J. 2003; 24 (20): 1815-23.
• 17. Steinhubl SR, Kastrati A, Berger PB. Variation in the definitions of bleeding in clinical trials of patients with acute coronary syndromes and undergoing percutaneous coronary interventions and its impact on the apparent safety of antithrombotic drugs. Am Heart J. 2007; 154 (1): 3-11.

Validation of a score for predicting bleeding events during acute coronary syndromes

Luis C. L. Correia; Saulo Merelles; Ana Vasconcelos; Thais Cerqueira; Tiago Reis; Carolina Esteves; José C. Lima; J. Péricles Esteves

Publication Dates

• Publication in this collection
  13 Aug 2010
• Date of issue
  Oct 2010

History

• Reviewed
  09 July 2009
• Received
  09 July 2009
• Accepted
  29 Oct 2009