Advances in the catheterizations techniques, stent designs, and pharmacotherapeutics have reduced adverse ischemic event rates such as cardiovascular death, ST (stent thrombosis), recurrent myocardial infarctions, and revascularizations in patients with coronary artery disease undergoing elective or acute percutaneous coronary intervention (PCI). However, the same antithrombotic regimens also increase bleeding risk, and a cornerstone of contemporary cardiovascular interventions is balancing ischemic and bleeding events in order to optimize the net benefit for the individual patient.

Bleeding complications in patients undergoing cardiovascular interventions have previously been underappreciated. Over the decade, however, awareness has gradually raised that adverse bleeding outcome after PCI carries substantial hazard comparable to that of post-PCI myocardial infarction associated with mortality. A growing number of studies showing the impact of bleeding on short- and long-term mortality have promoted bleeding end points to a pivotal metric, increasingly applied as a single and combined primary end point in randomized controlled trials (RCTs). The composite end point of net adverse clinical events (NACE), assessing both ischemia and bleeding in the same outcome measure, was conceived to investigate the balanced effect of novel therapeutic agents or clinical strategies in antithrombotic treatment regimens.

However, understanding the full impact of bleeding on outcome is still challenging. While ischemic adverse events are mostly narrowly adjudicated by widely accepted consensus definitions such as ST by the academic research consortium or myocardial infarction (MI) by the global definition of MI, bleeding assessment has been based on a large and very heterogeneous palette of definitions, often arbitrarily modeled as per-protocol criteria for specific clinical trials.

In the following, we will summarize the existing bleeding risk scores, their comparability, and their effect on clinical outcomes, as well as highlight their importance in contemporary clinical decision making and clinical trial design.

Bleeding is the most common noncardiac complication after PCI and leads to incremental increase in costs of health care. Rao et al¹ recently published an updated bleeding model based on the US National Cardiovascular Data Registry (NCDR) in order to predict the risk of postprocedure major bleeding complications among patients undergoing PCI. Bleeding was defined as adverse events occurring within 72 hours after PCI or before hospital discharge, and the criteria are described in Table 59-1. In contemporary clinical practice describing more than 1 million procedures from over 1000 invasive US centers, this report identified over 60,000 PCI procedures, with postprocedure bleeding comprising an incidence of 5.8% in this cohort. Among the bleeding events, 32% related to a specific anatomic location; 44.6% were detected due to a pre- to postprocedure hemoglobin decrease, 21.8% by a blood transfusion, and 1% by cardiac tamponade; and 0.6% were intracranial hemorrhage events.¹ Approximately half of bleeding events occur at the arterial site and may cover a large spectrum of clinical importance from uncomplicated subcutaneous access site hematoma to fatal retroperitoneal bleeding.

Subherwal et al² has described the temporal trends of the incidence of bleeding among 1.7 million patients undergoing PCI between 2005 and 2009. The finding of a nearly 20% reduction in post-PCI bleeding over time was largely due to temporal changes in antithrombotic strategies, as bivalirudin use increased from 17% to 30% and any heparin plus glycoprotein IIb/IIIa inhibitor decreased from 41% to 28%, while PCI by radial access and the use of vascular closure devices remained similar among all-comer PCI.²

Bleeding varies according to the clinical setting of PCI. In ST-segment elevation MI (STEMI) patients, the usage of fibrinolysis and the hectic preprocedural phase with the associated risk of misdosing anticoagulant regimen and a potentially less thorough assessment of individual bleeding risk, as well as the urgent need to gain vascular access, are all likely contributors to the well described increased incidence of major bleeding complications. Several registries have reported elevated bleeding rates in STEMI patients ranging from 6.5% to 11%.^3,4 Intensified antiplatelet and antithrombotic pharmacotherapy in non–ST-segment elevation MI (NSTEMI) patients also makes this subset of patients particularly susceptible to bleeding complications, and the CRUSADE (Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes With Early Implementation of the American College of Cardiology [ACC]/American Heart Association [AHA] Guidelines) registry has reported major bleeding rates of almost 12%.

However, the reported bleeding incidence and the associated mortality risk are greatly dependent on the circumstance (RCT vs observational registry with either retrospective or prospective data collection), patient risk presentation (eg, elderly, acute coronary syndrome [ACS], vs non-ACS), and the bleeding criteria applied.

A large number of bleeding definitions exist (see Table 59-1). This is a challenge as bleeding end points are given an increasingly pivotal role as safety end points in clinical randomized trials, but the comparability is low because the criteria may be quite different. For instance, The TIMI (Thrombolysis in Myocardial Infarction) bleeding classification is a laboratory-based scale, while the GUSTO (Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries) bleeding classification is a clinically based scale that does not consider clinical chemistry data. More than 15 different bleeding criteria are currently applied to various degrees and originate from datasets highly differing in setting and inclusion period (see Table 59-1). They are based on RCTs or observational registries or defined by consensus in research consortiums (Table 59-2), originate from treatment periods in the early 1990s to present day, and assess highly varying clinical entities (eg, intracranial hemorrhage or hemoglobin drop) in order to define major bleeding. It is well known that variations in bleeding criteria used to define major bleeding have led to differences in reported rates.⁵

Intuitively, major bleeding criteria such as the GUSTO criteria, which only consider clinical variables such as intracranial hemorrhage, hemodynamic compromise, or intervention/clinically significant disability, identify much less major bleeding but convey a much higher mortality prediction than the more recent and sensitive per-protocol bleeding criteria, which in some cases adjudicate major bleeding already at a liberal use of any blood transfusion with overt bleeding, such as the ACUITY (Acute Catheterization and Urgent Intervention Triage Strategy), REPLACE-2 (The Randomized Evaluation in PCI Linking Angiomax to Reduced Clinical Events), or STEEPLE (Safety and Efficacy of Enoxaparin in PCI Patients, an International Randomized Evaluation) criteria. Logically, the latter scores will entail higher estimates of major bleeding, but with less associated mortality hazard (see Table 59-1).

Another example of large differences between major bleeding criteria comes from the REPLACE-2 study investigating bivalirudin with provisional glycoprotein IIb/IIIa inhibitor (GPI) versus heparin with planned GPI in patients undergoing elective or urgent PCI.⁶ Bleeding complications were evaluated by the TIMI criteria and the per-protocol REPLACE-2 criteria. When the TIMI major bleeding criteria were used, no difference in bleeding was found; however, when the REPLACE-2 major bleeding criteria were applied, significantly less bleeding was seen in the bivalirudin group. A crucial difference between the TIMI bleeding criteria and the REPLACE-2 bleeding criteria is that a drop in hemoglobin concentration ≥3 g/dL is considered major bleeding in the REPLACE-2 bleeding criteria, but only minor bleeding in the TIMI bleeding criteria (see Table 59-1).

A similar example is found in the STEEPLE trial evaluating 2 different doses of enoxaparin with unfractionated heparin in patients undergoing elective PCI.⁷ Again, the TIMI major bleeding criteria were used without finding a difference between any of the groups. Conversely, when the per-protocol STEEPLE major bleeding criteria were used, there was significantly more bleeding in the unfractionated heparin group than in any of the enoxaparin groups. Also in this case, the STEEPLE criteria adjudicate major bleeding when a drop in hemoglobin concentration ≥3 g/dL occurs in the presence of overt bleeding, a finding that only qualifies for minor bleeding according to the TIMI criteria. Furthermore, the STEEPLE criteria consider any transfusion of ≥1 unit of packed red blood cells in the scenario of overt bleeding as pathognomonic for major bleeding, which adds additional variability to the bleeding adjudication as transfusion standards vary considerably across institutions and regions.

An obvious reason for this large number of per-protocol bleeding criteria is that the earliest major bleeding definitions, such as GUSTO and TIMI, only consider clinically major bleeding and are not very sensitive to other bleeding that requires clinical action and increases treatment items for the patient as well as the length of hospital stay and cost and thus constitutes a sensible end point to measure. Less sensitive bleeding criteria decrease statistical power, thereby augmenting the need for large sample sizes in RCTs. This is especially an issue in the constantly improving environment of the pharmacotherapeutic and technical interventional advances in PCI with consequently decreasing adverse event rates. Less sensitive bleeding criteria have therefore become increasingly less attractive to clinical trial investigators.

Adding to the lack of generalizability of bleeding criteria is the circumstance of their origin (ie, whether they were defined or evaluated in clinical trials with rigorous prospective data collection by clinical coordinators and individual event adjudication by clinical event committees, or whether the outcome was evaluated retrospectively by chart review in observational registries, possibly resulting in underreporting of adverse events). In contrast, bleeding criteria and outcomes from clinical trials can be harder to extrapolate to different clinical settings due to the often narrow inclusion criteria of the trial, whereas registries tend to provide a more accurate image of outcomes in real-world clinical practice.

Finally, large differences in concomitant antiplatelet and anticoagulant regimens, as well as highly variable interventional techniques, have been applied over the years and in publications reporting bleeding outcome from the different criteria.

For all of these reasons, in an effort to harmonize bleeding criteria, standardized bleeding definitions for cardiovascular clinical trials were designed and resulted in a consensus report from the Bleeding Academic Research Consortium (BARC), defining the BARC bleeding criteria, which have recently been validated in an independent population (see Table 59-2).⁸ Ndrepepa et al⁸ found in this patient-level pooled analysis of 12,459 patients recruited in 6 randomized trials of patients undergoing PCI that BARC class ≥3 was associated with a significant and similar adjusted 1-year mortality hazard of 3.19 (95% confidence interval [CI], 2.34-4.35), compared to TIMI (major + minor; 3.64; 95% CI, 2.62-5.07) and REPLACE-2 (major; 3.14; 95% CI, 2.30-4.29). The European Society of Cardiology Working Group on Thrombosis has stated in their latest issued position paper that bleeding should be reported using at least 2 bleeding scales and 1 of these should be the BARC bleeding criteria.⁹

The best clinical evidence of the association between bleeding and mortality in ACSs comes from 2 prominent examples in large-scale randomized trials, in which the ischemic end points between the treatment groups were very similar but major bleeding differed significantly, resulting in a mortality benefit associated with the reduced bleeding. This effect was evident in the HORIZONS-AMI (The Harmonizing Outcomes with Revascularization and Stents in Acute Myocardial Infarction) trial at 30 days and 1 year of follow-up in STEMI patients treated with bivalirudin compared to patients treated with unfractionated heparin and GPIs.¹⁰ Patients undergoing primary PCI and treated with bivalirudin had a significantly reduced primary end point defined by the non–coronary artery bypass graft (CABG)–related major bleeding rate of 4.9% versus 8.3% in patients treated with heparin plus a provisional GPI, resulting in a significant decrease in death as a result of cardiac cause from 2.8% to 1.8%. At the same time, ischemic outcome in terms of reinfarction, revascularization stroke, and all ST (acute and subacute) did not differ, indicating that the survival benefit is attributable to the improved bleeding profile of bivalirudin at 30 days of follow-up. This difference between groups widened even more in the time period between 30 days and 1 year after PCI. Cardiac death and reinfarction occurred less frequently in patients treated with bivalirudin. This long-term reduction might have been attributable to the prevention of iatrogenic hemorrhagic complications. The rates of all-cause mortality, cardiac mortality, and stroke were each 5-fold higher in patients with major bleeding compared to those without major bleeding, while the rate of reinfarction was 2 times higher.¹¹

A similar effect of decreased major bleeding associated with improved survival has also been reported in patients without STEMI. While the rates of death, MI, or refractory ischemia through day 9 after hospitalization were equal between groups, non–ST-segment elevation ACS patients randomized to fondaparinux treatment had a lower rate of major bleeding at 9 days and significantly improved 6-month survival compared with patients assigned to enoxaparin.¹² The OASIS-5 (Organization for the Assessment of Strategies for Ischemic Syndromes) trial therefore constitutes another example of a survival benefit attributable to an improved bleeding profile. Since then, several other substudies in cohorts originating from RCTs have confirmed that major bleeding is a powerful independent predictor in ACS populations of 30-day¹³ and 1-year mortality¹⁴ regardless of the bleeding criteria used. Major bleeding has also been associated with ischemia end points and ST.¹³

In observational data from over 3 million all-comer procedures in the NCDR CathPCI Registry performed in the United States between 2004 and 2011, a propensity-matched analysis revealed that postprocedural bleeding events were associated with increased risk of in-hospital mortality and that approximately 1 out of 8 deaths was related to bleeding complications. Major bleeding was associated with increased in-hospital mortality with a number needed to harm (NNH) of 29. The association between major bleeding and in-hospital mortality was observed in all strata of preprocedural bleeding risk, although NNH decreased substantially across NCDR bleeding risk strata (high: NNH = 21; intermediate: NNH = 39; and low: NNH = 69). Although both access site (NNH = 117) and non–access site bleeding (NNH = 16) were associated with increased in-hospital mortality, the mortality risk associated with access site bleeding was substantially lower than for non–access site bleeding.¹⁵ Both access and non–access site bleeding events occurring within 30 days of PCI were also independently associated with an increased long-term risk of mortality at 1-year follow-up. Non–access site bleeding has recently been confirmed to be a stronger correlate of mortality than access site bleeding, and it improves the discriminatory power of models for mortality prediction.¹⁶

In patients with MI, nuisance bleeding (BARC 1) is common in the following year, related with ongoing use of dual antiplatelet therapy (DAPT) and independently associated with worse patient perception of quality of life.¹⁷ Whether or not premature DAPT cessation after PCI due to nuisance bleeding is associated with increased rates of ST or MI is not well described. Improved selection of patients for prolonged DAPT may help minimize the incidence and adverse consequences of nuisance bleeding. Bleeding outcomes in relation to criteria, patient presentation, and circumstance of data collection are summarized in Table 59-3. The clinical variables defining the various bleeding criteria are summarized in Table 59-4.