Objective To investigate the relationship between inhospital bleeding as defined by Bleeding Academic Research Consortium (BARC) consensus classification and short-term and long-term mortality in unselected patients admitted for primary percutaneous coronary intervention (PCI).
Methods We analysed data of all consecutive patients with ST segment elevation myocardial infarction (STEMI) admitted for primary PCI, enrolled in a prospective registry of a high volume centre. The BARC-defined bleeding events were reconstructed from the detailed, prospectively collected clinical data. The primary outcome was mortality at 1 year.
Results Of the 1808 patients with STEMI admitted for primary PCI, 115 (6.4%) experienced a BARC type ≥2 bleeding. As the BARC bleeding severity worsened, there was a gradient of increasing rates of 1-year death. The 1-year mortality rate increased from 11.5% with BARC 0+1 type to 43.5% with BARC type 3b bleeding. After multivariable adjustment for demographic and clinical characteristics of patients, the independent predictors of 1-year death were BARC type 3a (HR 1.99; 95% CI 1.16 to 3.40, p=0.012) and BARC type 3b bleeding (HR 3.22; 95% CI 1.67 to 6.20, p<0.0001).
Conclusions The present study demonstrated that bleeding events defined according to the BARC classification hierarchically correlate with 1-year mortality after admission for primary PCI. The strongest predictor of 1-year mortality is the BARC type 3b bleeding.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Several randomised trials have demonstrated that patients with major bleeding complicating acute coronary syndromes (ACS) and percutaneous coronary intervention (PCI) have worse clinical outcomes than those without bleeding.1–7 However, interpreting and comparing bleeding data is difficult due to differences in bleeding definitions currently in use.8
In order to standardise multiple bleeding definitions for patients receiving antithrombotic therapy, the Bleeding Academic Research Consortium (BARC) has recently developed a consensus classification for bleeding by using a hierarchical approach to describe bleeding severity grade.9 A recent study by Ndrepepa et al10 examined the prognostic value of bleeding events according to the BARC classification in patients with various forms of coronary artery disease treated with PCI, from stable angina to non-ST segment elevation ACS, excluding patients with ST segment elevation myocardial infarction (STEMI). They validated the new BARC bleeding definition compared with TIMI (Thrombolysis In Myocardial Infarction) or Randomised Evaluation of PCI Linking Angiomax to Reduced Clinical Events-2 criteria.
Patients with evolving STEMI admitted for primary PCI are treated with increasing numbers and higher doses of antiplatelet agents and anticoagulants compared with patients with other forms of coronary artery disease and therefore are at higher risk of bleeding complications. The aim of this study was to investigate the relationship between inhospital bleeding defined by BARC consensus document and 1-year mortality as primary, and 30 days as well as between 30 days and 1-year mortality as secondary objective in unselected ‘real world’ patient population who underwent emergency coronary angiography and primary PCI for STEMI in a single large volume centre.
We analysed 1808 consecutive patients with STEMI admitted for primary PCI between August 2009 and January 2011 at the Department of Cardiology, Clinical Center of Serbia, Belgrade. Data of patients who experienced inhospital bleeding were collected prospectively and entered into the dedicated electronic database.
Acute STEMI definition was based on the history of chest pain/discomfort lasting for at least 20 min attributed to myocardial ischaemia, accompanied by persistent ST segment elevation of ≥1 mm in ≥ two contiguous limb leads or ≥2 mm in precordial leads; or presumable new left bundle branch block; or true posterior MI with ST depression of ≥ 1 mm in ≥ two contiguous anterior leads.11
For the purpose of this study the BARC classification was reconstructed from the detailed clinical data of patients with inhospital bleeding. Authors reviewed the bleeding events to see which BARC category they would fall into. The BARC definition contains the six types of bleeding that were hierarchically defined from type 0 in which there is no bleeding to type 5 in which there is fatal bleeding.9 Bleeding definition according to BARC criteria is provided in online supplementary appendix table 1. In this study, BARC type 0 (no evidence of bleeding) and BARC type 1 bleeding (the patients do not seek medical attention) were analysed together. In the current study patients with BARC type 4 (coronary artery bypass graft surgery (CABG)-related) bleeding were not taken into consideration because CABG was generally performed more than 4 weeks after the acute phase of STEMI.
All inhospital events were also assessed using TIMI and Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO) criteria.
Major adverse cardiac events (MACE) were defined as a composite of death, reinfarction, target vessel revascularisation for ischaemia or stroke.
PCI procedure and subsequent antithrombotic medications
All patients with STEMI underwent emergency coronary angiography using standard percutaneous techniques via the femoral or radial artery, followed by primary PCI (angioplasty or intracoronary stent implantation), CABG or medical management. Unfractionated heparin was administrated as an intravenous bolus of 100 IU per kilogram of body weight or 50–60 IU/kg, if glycoprotein IIb/IIIa inhibitor (GPI) is given.12 Aspirin (300 mg orally) was preloaded in all patients, after which 100–300 mg was given orally every day during the first 30 days and 100 mg daily thereafter indefinitely. Clopidogrel was given as a loading dose of 600 mg before insertion of the catheter, and 75 mg orally every day for 1 year.12 Use of GPI during primary PCI was left to the physician's discretion. The only GPI used in our study was tirofiban.13
End points and follow-up data
The primary end points were all-cause mortality at 1 year and within two separate time intervals: at 30 days and between 30 days and 1 year follow-up. Secondary end points were MACE at the same time intervals (cumulative at 1 year, at 30 days and between 30 days and 1 year). Out of hospital clinical outcomes were obtained by telephone interviews of patients and families conducted by educated medical doctors or in the outpatient clinic at 30 days and 1 year follow-up. The follow-up rate for mortality was 100%, as it was supported by the National Institute of Public Health. Patients were followed up for MACE as per standard local practice, as part of the registry for primary PCI. The achieved rate of follow-up for MACE was 94%.
The study protocol was approved by the ethics committee of our institution and was conducted according to the principals of the Declaration of Helsinki. Informed consent was obtained from all patients.
Descriptive statistics was computed as mean values and SD for continuous variables (or median values and IQR if skewed) and as absolute frequencies and per cent values for categorical variables.
Patients were grouped according to the BARC categorisation of bleeding (type 0+1, type 2, type 3a, type 3b, type 3c, type 5a and type 5b).
After excluding patients with BARC type 5 (fatal bleeding), we performed statistical analysis to assess early and late risk of death in patients who experienced non-CABG-related inhospital bleeding that ranged from BARC type 2 to BARC type 3b. The BARC type 3c bleeding (intracranial) occurred only in two patients and was excluded from statistical analysis.
Rates of clinical end points were compared between patients with various types of BARC bleeding definition, cumulative to 1 year and within two separate time intervals (within 30 days and between 30 days and 1 year) using χ2 test for trend. Kaplan-Meier analysis was used to illustrate 1 year survival for patients who had different degrees of bleeding severity. Survival curves were compared using the logrank statistic.
Univariable and multivariable Cox proportional hazards model was used to test the association between BARC bleed severity and mortality at 30 days, between 30 days and 1 year, and at 1 year. We compared adjusted rates of the primary and secondary end points among patients with BARC type 0+1 bleeding and those within the other types of BARC bleeding. Adjustment variables (risk factors) included in this model were age, gender and weight; history of diabetes mellitus, hypertension and hypercholesterolaemia; previous MI, previous PCI, previous stroke and current smoking. Variables entered into the model were prespecified. The proportional hazard assumption was tested by using the procedure ‘stphtest’ in STATA which has been based on methods described by Grambsch and Therneau.14 ,15 Discriminatory power of multivariable models without and with bleeding according to BARC categorisation was assessed by performing receiver operating characteristic (ROC) curve analysis. Multivariable models were used to construct adjusted ROC curves and compute areas under curves (AUCs). The same variables as for Cox proportional hazards model adjusting for the association between bleeding and mortality were entered into the model of adjusted ROC curves. Evaluating AUC with the same data that were used to estimate HRs is the limitation of the analysis. All statistical analyses were performed using IBM SPSS statistics V.19.1 and STATA V.11.2.
This study included a total of 1808 patients with STEMI who underwent emergency coronary angiography. Of these, 1628 (90.0%) were treated with primary PCI. The proportion of female gender was 30.2% (546 patients). Baseline demographic and procedural characteristics of the patients are shown in table 1. Of the patients who underwent emergency coronary angiography, 115 (6.4%) experienced a BARC bleeding type ≥2. With regards to the BARC bleeding definition, 1.7% of patients experienced a BARC type 2 bleed, 3.1% of patients experienced a BARC type 3a bleed and 1.3% experienced a BARC type 3b bleed. Among BARC type 3b bleeding, overt bleeding plus haemoglobin drop ≥5 g/dL occurred in 21.7% of patients, bleeding requiring intravenous vasoactive agents occurred in 17.4%, whereas the rate of combined previous subtypes and bleeding requiring surgical intervention was 60.9%. The BARC type 3c bleeding (intracranial) occurred in only two patients (0.1%), and BARC type 5 bleeding (fatal) occurred in three (0.2%) patients.
Patients with BARC bleeding type ≥2 were older and more often had hypertension and hypercholesterolaemia compared with those having BARC type 0+1 (table 1). A higher proportion of patients with BARC type 3b bleeding were of lower weight and more often of female gender compared with those with BARC type 3a or type 2 bleeding. Killip classes 2–4 had approximately 20% of patients with BARC type 0+1, but more than half of patients with BARC type 3a and BARC type 3b bleeding. The use of intra-aortic balloon pump was higher in patients with BARC types 3a and 3b. GPI use, rate of PCI procedure and procedural time increased as the BARC bleeding severity worsened. With the worsening of BARC bleeding severity there was a gradient of increasing proportion of patients with stent implantation and increasing number of stents per patient.
Bleeding events and outcomes
The rates of mortality and MACE within 30 days, between 30 days and 1 year and 1-year cumulative rate, among patients who experienced the various classes of BARC bleeding definition, is shown in table 2. The 30-day mortality significantly increased only with BARC type 3b bleeding. Between 30 days and 1 year, there was a stepwise increase in mortality rates until type 3a bleed, staying at the same level in patients with BARC type 3b bleed. The 1-year cumulative mortality rate increased from 11.5% with BARC type 0+1 to 43.5% with BARC type 3b bleeding, demonstrating hierarchical increase with BARC bleeding severity worsening.
As were mortality, MACE rates increased gradually with the higher BARC class.
Figure 1 presents Kaplan-Meier curves for mortality at 1 year among patients in various classes of BARC bleeding definition.
BARC bleeding severity as a predictor of mortality and MACE
The unadjusted and adjusted HRs of mortality are presented in table 3. Of bleeding types, the only independent predictor of 30-day mortality was type 3b bleeding compared with BARC 0+1 bleeding. Between 30 days and 1 year, the independent predictors of death were BARC 3a and 3b types. At 1 year, the severity of bleeding events by BARC hierarchically correlate with the adjusted risk of dying, with BARC type 3b bleeding as a strongest predictor of mortality. Other independent predictors of 1-year mortality included age per 1-year increase, previous MI, and previous stroke.
HRs for MACE for different BARC categories compared with BARC 0+1 as a reference category are similar to the risk of mortality at 1 year and within two separate time intervals (within 30-days and between 30-days and 1 year) (table 4). The only independent predictor of MACE at 30 days was BARC 3b. The severity of bleeding according to BARC hierarchically correlates with the adjusted risk of MACE at 1 year, and between 30 days and 1 year. Among other independent predictors of MACE at 1 year, besides older age, previous MI and previous stroke, was diabetes mellitus.
Similar analysis was performed for prognostic impact of bleeding as defined by TIMI and GUSTO criteria. Adjusted HRs for mortality according to different categories of TIMI and GUSTO bleeding classifications are shown in table 5.
At 30-days only TIMI major and GUSTO severe bleeding categories were significantly associated with the risk of death. TIMI minimal, minor and major, as well as GUSTO moderate and severe bleedings were significantly associated with increased rates of 1-year mortality.
Predictive value of BARC bleeding definition
Predictivity of bleeding by the BARC categorisation with regards to mortality at 1 year and within two separate time intervals was assessed by calculating the adjusted AUCs in multivariable models including bleeding by BARC in each model (table 6). Bleeding events as defined by the BARC consensus document have significant predictive value with regards to mortality at 1 year (p=0.037) and for period between 30 days and 1 year (p=0.047), but not within first 30 days (p=0.093).
The main findings of our study are: (1) severity of bleeding defined according to the BARC classification hierarchically correlate with 1-year mortality/MACE after admission for primary PCI; (2) The BARC type 3b bleeding is an independent predictor of 30-day mortality/MACE and BARC types 3a and type 3b are independent predictors of 1-year mortality/MACE after admission for primary PCI; (3) The inclusion of bleeding according to the BARC definition in models, significantly increase the prediction of 1-year mortality after admission for primary PCI.
It has recently been shown that bleeding complications have been associated with an increased risk of subsequent adverse outcomes, including death, MI, stroke and stent thrombosis.1–3 16–18 Therefore, prevention of bleeding has become equally as important as the prevention of ischaemic events. However, this effort is hampered by the lack of uniformity in bleeding definition currently in use. Two definitions of bleeding that have been most commonly used in published clinical trials and registries are the TIMI, based mainly on laboratory tests,19 and the GUSTO, based mainly on clinical criteria,20 both developed in the fibrinolytic era. Later bleeding definitions have used elements from TIMI and GUSTO criteria, made new combinations with these elements or added new elements.17 The new BARC bleeding definition offers a balanced combination of laboratory based and clinically based metrics in bleeding criteria.
A study by Rao et al21 has shown that use of different bleeding definitions in the same populations of non-ST segment ACS may lead to considerable differences in the reporting of bleeding events. After multivariable adjustment, the stepwise increase in 30-day and 6-month risk for death and reinfarction with worsening GUSTO bleeding persisted while the risk with TIMI bleeding was no longer statistically significant.21 In the Harmonising Outcomes with RevasculariZatiON and Stents in Acute Myocardial Infarction trial, bleeding as a dichotomous characteristic in patients with STEMI treated with primary PCI was associated with the similar increased risk of mortality after 30 days and after 1 year (approximate sixfold increased) compared with those without bleeding.7
A recent study by Ndrepepa et al10 has shown a close association between bleeding events defined by BARC (dichotomised at BARC class ≥2 and BARC class ≥3) and 1-year mortality among patients with various forms of presentation of coronary artery disease undergoing PCI, recruited in six Intracoronary Stenting and Antithrombotic Regimen clinical trials. They confirmed that BARC bleeding definition has a relationship with mortality similar to bleeding defined according to the TIMI or Randomised Evaluation of PCI Linking Angiomax to Reduced Clinical Events-2 criteria.
In the present study, we investigate the relationship between bleeding defined by BARC criteria and mortality/MACE in a real world, unselected population of patients with evolving STEMI admitted for primary PCI. Our study includes a small number of patients who bleed, so the strength of the study is not very large. We found that bleeding events by BARC categorisation were strongly associated with 30-day mortality/MACE and 1-year mortality/MACE. After multivariable adjustment, the BARC type 3b bleeding was the independent predictor of 30-day mortality/MACE, and the strongest predictor of 1-year mortality/MACE. Patients with BARC type 3b bleeding were at more than threefold higher 30-day risk of death and MACE compared with patients with BARC type 0+1.
Our study includes analysis of bleeding complications in a STEMI population, with the limited use of radial access and thrombus aspiration as compared with the contemporary interventional practice. The impact of comorbidities on outcomes cannot be excluded. Our database included unselected, all-comers population with high prevalence of risk factors including 22% of patients presenting with Killip class II–IV. According to available data from other studies and our present work, patients with more severe bleeding have worse prognosis. The possible mechanisms that link bleeding complications with increased mortality rates in patients with ACS are likely to be multifactorial. Acute, inhospital bleeding may be associated with hypovolaemia, anaemia and impaired oxygen carrying capacity. The ensuing haemodynamic compromise could trigger a hyperadrenergic state with deleterious effects on the already ischaemic myocardium.3 The occurrence of inhospital bleeding may lead to discontinuation of dual antiplatelet therapy with the risk of recurrence of ischaemic events and of stent thrombosis. Furthermore, the need for transfusion of preserved blood cells might result in microcirculatory disorders due to depletion in nitric oxide and 2,3-diphosphoglyceric acid.17 ,18 In our study the milder degrees of bleeding (BARC 3a and BARC 2 bleeding) are associated with long-term adverse outcomes. Mechanisms that might lead to late mortality include proinflammatory and prothrombotic effects as a result of bleeding and transfusion which may persist well beyond the immediate bleeding period and set the substrate for future ischaemic events.5 ,6 Higher long-term mortality rates in patients with bleeding complications after primary PCI might be fostered due to a clinician's hesitation to fully reinstitute antiplatelet medication, β blockers and ACE inhibitors, after a bleeding event.6
We confirmed that bleeding events defined by BARC classification hierarchically correlated with 1-year mortality/MACE. Thus, patients with BARC type 3a bleeding were at almost twofold higher risk and those with BARC type 3b were at more than threefold higher risk of 1-year death compared with patients with BARC type 0+1. Patients with BARC type 3b bleed were at more than threefold higher risk of 1-year MACE than patients with BARC type 0+1 bleed.
We also estimated that inclusion of bleeding events defined by the BARC consensus document in models have significant predictive value with regards to mortality at 1 year and mortality between 30 days and 1 year.
The rate of 1-year mortality for non-bleeders in this study (11.5%) is higher than that of the Harmonising Outcomes with RevasculariZatiON and Stents in Acute Myocardial Infarction trial (3.1%) and the study of Ndrepepa et al (2.3%).7 ,10 It may be due to high prevalence of Killip class II–IV in the described registry population.
The described rate of bleeding in this registry is greater than in majority of randomised controlled clinical trials. One of the possible reasons is that patients with known bleeding history are mostly excluded from randomised trials.
Our study has some limitations. First, bleeding events have been collected prospectively, but needed to be reclassified according to BARC criteria from the collected data. This was due to the fact that data collection began prior to publication of BARC criteria. Second, patients with BARC type 0 and BARC type 1 bleeding were analysed together. We could not determine if BARC type 1 bleeding occurred. This makes that BARC 0–1 bleeding type represent those without documented bleeding. Therefore, data presented in this study need confirmation from studies based on prospectively adjudicated events according to BARC bleeding definition.
The results of our study show that bleeding events defined according to the BARC classification hierarchically correlate with 1-year mortality/MACE after admission for primary PCI. Patients with BARC type 3b bleeding are at the highest risk of 30-day and 1-year mortality/MACE. The inclusion of bleeding defined by BARC criteria in models significantly improved the prediction of 1-year mortality and mortality for a period between 30 days and 1 year.
What is already known on this subject
Randomised trials have demonstrated that patients with major bleeding complicating acute coronary syndromes and percutaneous coronary intervention have worse clinical outcomes than those without bleeding. However, interpreting and comparing bleeding data is difficult due to differences in bleeding definitions currently in use.
What this study adds
This study validated the new BARC bleeding definition and demonstrated that severity of bleeding according to the BARC classification hierarchically correlates with 1-year mortality/major adverse cardiac events after admission for primary PCI.
How might this impact on clinical practice
This study confirmed the prognostic importance of bleeding measured by BARC definition. So, doctors should be encouraged to accept this definition in future trials and registries as an universal bleeding classification.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online appendix
Contributors All authors listed have contributed sufficiently to the project to be included as authors, and all those who are qualified to be authors are listed in the author byline. All authors have read and approved the manuscript. GRS as corresponding author acts as guarantor. DMM was involved in the study conception and design; acquisition, analysis and interpretation of data; drafted the article and had final approval of the manuscript. MRA, IBM and DGM were involved in the study design, model construction and interpretation of results; drafted the article and had final approval of the manuscript. MMM, NMA, VDV and LZS were involved in the study design and interpretation of data; revising the article and had final approval of the manuscript. JMM and NIK were involved in the acquisition and analysis of data; revising the article and had final approval of the manuscript. ZHM, MNZ and VMD were involved in the study design and interpretation of data; revising the article and had final approval of the manuscript. GRS was involved in the study conception and design; interpretation of data; drafting and revising the article and had final approval of the manuscript.
Competing interests None.
Ethics approval Ethics Committee of the Faculty of Medicine, University of Belgrade, Belgrade, Serbia.
Provenance and peer review Not commissioned; externally peer reviewed.