Strong antithrombotic treatments may decrease the risk of recurrent ischaemic events in patients with acute coronary syndromes (ACS) and, as adjunctive therapy to percutaneous coronary intervention (PCI), they have shown a strong clinical benefit. However, a significant decrease in the risk of ischaemic complications is invariably associated with an increased risk of bleeding and there is growing evidence that bleeding in patients with ACS has a strong impact on survival.1 It is not easy to ascertain the true impact of bleeding on clinical outcome in patients undergoing PCI due to the need of large sample of patients to avoid the confounding effect of the worse baseline risk profile of patients who suffer bleeding, the need for an appropriate definition of the severity of bleeding complications, the confounding effect of many variables that are linked to bleeding and that are unexplored in most studies, such as the dose and duration of an antithrombotic treatment, the maintenance of femoral sheath in PCI patients after the procedure, the cut off of haematological parameters used for blood transfusion, the criteria used for discontinuation of antithrombotic treatment. Again, most of deaths are not directly due to a fatal bleeding, and the link between bleeding and mortality is not easily or fully explained in most studies, and the mechanism(s) of death remains largely speculative.
The definition of major bleeding
The need for an appropriate definition of the severity of bleeding complications is obvious when considering the criteria used for bleeding severity defined by the TIMI investigators2 and the GUSTO investigators3 originated from fibrinolytic trials that were antecedent to the diffusion of PCI in all clinical settings of coronary disease, from stable coronary disease to ST-segment elevation acute myocardial infarction (AMI). In the fibrinolytic era, the fibrinolytic agent played the central role both in the efficacy of the treatment and the safety of the treatment being fibrinolysis the main direct cause of bleeding, whereas in the current era of percutaneous mechanical revascularization the antithrombotic treatment is adjunctive to PCI and used mainly for the prevention of procedural and periprocedural ischaemic complications. Again, most of the major bleedings that may complicate a PCI procedure are the result of a systemic effect of the antithrombotic treatment associated with an anatomical source of bleeding specifically related to the procedure such as the femoral entry site or the iliac arteries. Alternative entry site approaches for PCI, such as the radial approach, are associated with a decrease in bleeding complications compared with the femoral approach that is several fold greater than the differences among different adjunctive pharmacological strategies revealed by refined and sophisticated analysis in randomized trials including thousands of patients.
New criteria for the definition of the severity of bleeding complications during or after PCI should satisfy the prerequisite of the significant clinical impact of bleeding definition on patient outcome with the use of hard clinical safety endpoints such as death, myocardial infarction, haemorrhagic stroke, or a major surgery for the treatment of bleeding. Other safety endpoints such as minor surgical or endoluminal intervention for arterial repair, haematoma, or length of hospital stay should be considered at a lower degree of bleeding severity due to the lack of univocal evidence of their significant clinical impact. The definition of severe bleeding complication should also include the amount of blood loss and the need for blood transfusion, with the recognition of universally accepted cut-off of haemoglobin level and indications for blood transfusion. The latter should be considered as a very important issue considering the demonstrated detrimental impact of blood products transfusion on clinical outcome and their frequent use in patients with non-harmful haemoglobin levels or in patients with pre-existing anaemia.
Bleeding as a predictor of poor outcome
The association between major bleeding and mortality was assessed by Eikelboom et al.1 in a cohort of 34 146 patients with acute coronary syndromes enrolled in the OASIS Registry, in the OASIS-2 Trial and in the CURE trial.1 In this sufficiently powered study, major bleeding occurred in 783 patients (2.3%) and was associated with increased mortality. The 30 day mortality rate was 12.8% in patients who suffered bleeding and 2.5% in patients without bleeding (HR 5.37, 95% CI 3.97–7.26, P < 0.0001). After 30 days, relation between bleeding and mortality was weaker (HR 1.54, 95% CI 1.01–2.36, P = 0.047). The predictors of major bleeding were the use of fibrinolytics, glycoprotein IIb/IIIa inhibitors, unfractionated heparin, low-molecular-weight heparin, oral anticoagulant agents, intraaortic balloon pump, coronary angiography, and coronary surgery. PCI was not associated with an increased risk of major bleeding. Major bleeding was also associated with an increased risk of ischaemic events including death, myocardial infarction, and stroke. The incidence of major ischaemic events was 20% in patients with bleeding and 5% in patients without bleeding. The increased risk of ischaemic events in patients who suffer bleeding may link bleeding to mortality through a fatal ischaemic event that could be precipitated by platelet activation secondary to bleeding and, more likely, by inappropriate antithrombotic treatment discontinuation.
The discontinuation of antithrombotic treatment as the main link between bleeding and ischaemic mortality was clearly shown by the GRACE investigators.4 In a registry including 40 087 patients with AMI (53% ST-elevation AMI, 47% non-ST-elevation AMI) the incidence of major bleeding was 2.8%. Patients with major bleeding were older, more severely ill, and more likely to undergo invasive procedures. The in-hospital mortality was 20.9% in patients with major bleeding and 5.6% in patients without major bleeding (P < 0.001). After controlling for confounding baseline variables, the risk for increased mortality in patients with major bleeding was lower than the one reported from controlled randomized trials and limited to the in-hospital mortality (HR for in hospital mortality 1.9, 95% CI 1.6–2.2; after hospital discharge HR 0.8, 95% CI 0.6–1).
Among patients who suffered a major bleeding, the in-hospital mortality was considerably higher in those who had antithrombotic treatment discontinued as compared with patients with bleeding and who continued antithrombotic treatment. The mortality rate was 52% in patients who discontinued aspirin and 13% in patients who continued the drug (P < 0.001), 58% in patients who discontinued thienopyridine and 13% in patients who continued the drug (P < 0.001), and 26% in patients who discontinued heparin and 16% in patients who continued anticoagulant treatment (P = 0.03).
Thus, the findings of the GRACE registry suggest that the association between major bleeding and increased death is lower than that reported by controlled randomized trials, limited to the in-hospital stay, and no more evident after hospital discharge, and more evident in patients who had discontinuation of at least one antithrombotic drug.
The REPLACE-2 trial compared a strategy of bivalirudin plus provisional IIb/IIIa inhibitors with a strategy of unfractionated heparin plus routine IIb/IIIa inhibitors in 6001 patients undergoing PCI.5 Major bleeding definition included: intra-cranial haemorrhage, intra-ocular haemorrhage, retroperitoneal haemorrhage, overt haemoglobin loss >3 g/dL, any haemoglobin decrease >4 g/dL, or transfusion >2 U of blood products.
Major bleeding occurred in 195 patients (3.2%) and was more frequent in the heparin plus IIb/IIIa inhibitor arm. Major bleeding was associated with increased mortality at 1 month (5.1% vs. 0.2%, P < 0.001) and 1 year (8.7% vs. 1.9%, P < 0.001). The hazard risk of death for patients who experienced bleeding was 2.66 (95% CI 1.44–4.92, P = 0.002). Overall, 19 deaths occurred, 10 deaths occurred in 195 patients with major bleeding, and six out of the 10 deaths were due to a fatal bleeding. This study identified randomization to heparin plus IIb/IIIa inhibitors, provisional use of IIb/IIIa inhibitors in the bivalirudin arm, a time to sheath pull >6 h, a procedural length >1 h, and the use of intra-aortic balloon as predictors of major bleeding. Major bleeding definition included a decrease in haemoglobin of at least 3 g/dL, whereas TIMI criteria for major bleeding are more stringent (haemoglobin decrease >5 g/dL). If TIMI criteria for major bleeding were adopted, there were no more differences in major bleeding between the bivalirudin arm and the heparin plus IIb/IIIa inhibitor arm (0.9% and 0.6%, P = 0.30) and eventually major bleeding was no more independently related to mortality.
The ACUITY trial is a three-arm study comparing three antithrombotic strategies in patients with acute coronary syndromes: bivalirudin plus provisional IIb/IIIa inhibitors, bivalirudin plus routine IIb/IIIa inhibitors, and heparin plus routine IIb/IIIa inhibitors.6,7 Predefined major bleeding criteria included: intra-cranial or intra-ocular haemorrhage, access site bleeding requiring intervention, >5 cm diameter haematoma, haemoglobin reduction >4 g/dL without or >3 g/dL with an overt source, re-operation for bleeding or blood product transfusion.
Major bleeding rates were higher in patients randomized to heparin plus II/b/IIIa inhibitors compared with bivalirudin monotherapy (5.7% vs. 3.0%, P < 0.001) and similar to the bivalirudin plus IIb/IIIa inhibitor arm (5.7% vs. 5.3%, P = 0.38). Randomization to heparin plus IIb/IIIa inhibitors was associated with an increased risk of major bleeding, and major bleeding was associated with an increased 1 month mortality compared with patients without major bleeding: 7.3% vs. 1.2%, P < 0.0001 (HR 7.55, 95% CI 4.68–12.18, P < 0.0001). The incidence of fatal or life-threatening bleeding was very low (<1%), but patients with major bleeding experienced a higher incidence of thrombotic/ischaemic events, including death, myocardial infarction, urgent target vessel revascularization, and stent thrombosis for patients who underwent PCI, confirming that the ultimate cause of death in most patients who die due to major bleeding is a coronary thrombotic complication. The latter may be due platelet activation subsequent to haemorrhage or to blood transfusion, but, it is more likely that discontinuation of antithrombotic treatment play a major role in the genesis of ischaemic complications, as shown by the GRACE investigators.4 The definition of the criteria for major bleeding is not a marginal point. The hypothesis that an antithrombotic treatment that is non-inferior to standard treatment in terms of recurrent ischaemic events and that is superior in terms of safety due to a minor incidence of bleeding and subsequent adverse events is not proved. In the ACUITY-PCI,6 a better short-term clinical outcome (the quadruple end point including death, infarction, TVR and bleeding) provided by bivalirudin monotherapy is not associated with a better 1 year survival despite a decrease in major bleeding: 1 year mortality rate is 3.8 in the bivalirudin alone arm, 3.9% in the bivalirudin plus IIb/IIIa inhibitor arm, and 3.9% in the heparin plus IIb/IIIa inhibitor arm. On the other hand, it is has been demonstrated that the increase in the risk of death parallels the severity of bleeding.8 Thus, the adoption of broad criteria for major bleeding definition may not show any significant impact of antithrombotic treatment related bleeding on long-term survival.
The decrease in bleeding complications with an adjunctive antithrombotic treatment that include bivalirudin alone compared with heparin plus routine IIb/IIIa inhibitors was confirmed by the results of the PROTECT-TIMI 30 Trial.9 This trial, based on a sample of 754 patients with non-S-segment elevation ACS, compared three antithrombotic strategies as adjunctive to PCI: bivalirudin alone, eptifibatide plus unfractionated heparin, and eptifibatide plus enoxaparin. The primary endpoint of the study was the coronary flow reserve as assessed angiographically, while bleeding complications were assessed using the TIMI criteria. Patients randomized to bivalirudin had a greater coronary flow reserve compared with patients randomized to eptifibatide (1.43 vs. 1.33, P = 0.036), and a decrease in minor TIMI bleeding complications (0.4% vs. 2.5%, P = 0.027) and blood transfusion (0.4% vs. 4.4%, P < 0.01). There were no differences in major bleeding complications (0% vs. 0.7%). However, patients randomized to eptifibatide compared with bivalirudin had a 21% decrease in ischaemic complications that included the length of ischaemia as revealed by Holter monitoring (8.77% vs. 11.6%), myocardial infarction (6.6% vs. 8.5%), and death (0% vs. 0.4%). The net clinical outcome considering both ischaemic and bleeding complications favoured the eptifibatide plus heparin or enoxaparin treatments. Differently from the REPLACE 2 trial,5 that showed heparin plus IIb/IIIa inhibitor treatment to be a strong predictor of bleeding, in the PROTECT-TIMI 30 trial the only predictor of major and minor bleeding was age.10 It should be strengthened that the study provided unexpected results with a better effect on microcirculation of bivalirudin compared with eptifibatide, while the protective effect of bivalirudin against bleeding complication was seen at a lower degree compared with the REPLACE-2 and ACUITY trials and at the price of a non-significant but harmful increase in ischaemic complications.
The ongoing HORIZONS trial is a prospective randomized trial that compares bivalirudin plus bailout glycoprotein IIb/IIIa inhibitors with unfractioned heparin plus routine glycoprotein IIb/IIIa inhibitors (abciximab or eptifibatide as per investigator discretion) in 3602 patients undergoing primary PCI for ST-segment elevation AMI.11 A second randomization (stent arm) compares the safety and efficacy of a paclitaxel-eluting stent with the equivalent uncoated stent. The two primary endpoints of the pharmacological study are major bleeding and net clinical outcome at 30 days. Net clinical outcome was defined by the sum of major bleeding and major adverse clinical events (death from any cause, infarction, ischaemic target vessel revascularization, and stroke), while major adverse clinical events is a secondary endpoint. Bailout glycoprotein IIb/IIIa inhibitor was recommended in the bivalirudin alone arm in case of giant thrombus or refractory no-reflow. The adjunctive treatment included pre-angiography loading of 300 or 600 mg of clopidogrel in all patients. The 30 day results of the pharmacological arm of the study showed a benefit of the strategy of bivalirudin monotherapy. The two primary endpoints were reached: the major bleeding rate was 8.3% in the glycoprotein IIb/IIIa inhibitor arm and 4.9% in the bivalirudin arm (P < 0.0001), and the net clinical outcome rate was 12.1% and 9.2%, respectively (P = 0.006). The major clinical adverse event rates were similar in the two groups (5.5% and 5.4%), whereas the cardiac mortality rate was lower in the bivalirudin alone arm compared with the glycoprotein IIb/IIIa inhibitor arm: 1.8% and 2.9%, respectively (P = 0.035). Ischaemic target vessel revascularization was more frequent in the bivalirudin arm compared with the glycoprotein IIb/IIIa inhibitor arm (2.6% vs. 1.9%, P = 0.18), as well as acute stent thrombosis (1.3% vs. 0.3%, P = 0.009). The 30 day survival curves show an excess of mortality in the bivalirudin arm in the first 5 days that was related to the increase in stent thrombosis in this group. Several issues suggest caution in the interpretation of these preliminary results. First, the two primary endpoints were reached using the major bleeding definition of the ACUITY trial. The differences in major bleeding between the groups are less relevant or absent when using the corresponding TIMI definition of major bleeding (5.0% in the glycoprotein arm and 3.1% in the bivalirudin arm, P = 0.003) or the GUSTO definition (0.6% and 0.4%, P = 0.65) respectively. Secondly, the study was underpowered for major adverse ischaemic events. Whether these results will translate in improvement in survival will be ascertained by the ongoing long-term follow-up.
Several concerns arose from the preliminary results of the study can be anticipated. The incidence of major bleeding was higher than the one reported by concluded trials on abciximab in patients undergoing primary PCI for AMI.
The incidence of severe bleeding in the abciximab plus stenting arm of the CADILLAC trial was 0.8% and of the composite of moderate and severe bleeding was 3.3%.12 In the same trial, bleeding did not correlate to 1 year mortality.13
Again, in a recent metanalysis of the results from three European trials ISAR-2,14 ADMIRAL,15 and ACE16 that compared abciximab with control in primary PCI, the incidence of major bleeding according to the TIMI classification was only 2.5% in patients randomized to abciximab and 2% in patients randomized to control.17 The analysis showed a strong clinical benefit of abciximab at 3 year follow-up. The estimated mortality rate was 10.9% in patients assigned to abciximab and 14.3% in patients randomized to control, and the composite of death and re-infarction rates were 12.9% and 19% respectively (RR 0.633, 95% CI 0.452–0.887, P = 0.008).
The balance between bleeding and ischaemic complications in percutaneous coronary intervention practice
Owing to the conflicting or incomplete results of randomized controlled trials and registries regarding the safety and efficacy of different pharmacological antithrombotic treatments in the prevention of both bleeding and ischaemic complications, an individualized approach that considers, for each patient undergoing PCI, the risk of ischaemic and bleeding complications in the light also of patient's characteristics that are not considered in clinical trials seems to be the most appropriate. The risk of ischaemic complications may largely overcome the one of bleeding in patients with stable coronary disease but complex coronary anatomy requiring complex and multiple stent implantation such as distal left main disease, multivessel disease, or last patent vessel. Conversely, also in the setting of primary PCI, the risk of bleeding may overcome the one of ischaemic complications in an elderly woman, with low body mass index, renal insufficiency, and single-vessel disease supplying a small area at risk.
In high-risk PCI patients, it seems reasonable the adoption of the more potent antithrombotic treatment that include the routine use of clopidogrel loading and IIb/IIIa inhibitors that is supported by a strong evidence of an improved survival18,19 and the use of all measures that may dramatically decrease the entry site complications, including a radial approach in patients with high risk of bleeding and unfavourable femoral-iliac anatomy, the appropriate use of intra-aortic balloon pump, and other left ventricle support devices, while the discontinuation of the antithrombotic treatment should be avoided in all cases but life-threatening bleeding that cannot be corrected quickly by an endoluminal or surgical approach.
Conflict of interest. None declared.