BLEEDING DEFINITION, INCIDENCE AND PREDICTORS

The incidence of ischaemic complications during percutaneous intervention (PCI) has reduced over recent years.1 There is no consensus between investigators as to how to define major and minor bleeding (table 1).2 The TIMI investigators defined bleeding during studies assessing thrombolysis and these criteria were used in early adjunctive pharmacology trials such as EPILOGUE, ESPRIT and EPISTENT to measure periprocedural events.3 However, these criteria do not specifically relate to PCI, and are also restrictive in their definition of major bleeding, excluding bleeds that many clinicians would consider as significant. For example, a retroperitoneal bleed with a drop in haemoglobin of 4 g/dl would be classed as minor by the TIMI criteria.4 More recent studies such as ACUITY and REPLACE-2 have developed criteria which are much more inclusive than the TIMI criteria.5^–7 These criteria are also more targeted to PCI-related bleeding including complications such as haematomas >5 cm, access site surgical repair and retroperitoneal bleeds. These differences in the definition of bleeding make interstudy comparisons very difficult. However, major bleeding in the more contemporary studies such as ACUITY roughly equates to the combination of major and minor bleeding as defined by the TIMI criteria in older studies. Additionally, an awareness of the key differences in bleeding definitions is critical when comparing bleeding rates between studies.8

Accepting the difficulties in comparison, protocol-defined major bleeding is reported in 3.2–9.1% of patients recruited to randomised trials of adjunctive pharmacology during PCI.7 9^–14 In two of the most recent trials—ACUITY and ISAR-REACT 2—protocol-defined major bleeding was seen in 5.5% and 5.6% of patients, respectively, suggesting that bleeding rates remain high in contemporary practice.9 10 All randomised trials of adjunctive treatment exclude patients at increased risk of bleeding and thus major bleeding in “real-world” PCI may be even higher. In a study of 11 074 non-selected consecutive patients undergoing PCI, major bleeding was seen in 5.4% of cases.11 The site of bleeding is most commonly related to femoral access, with 50–70% of events either being a large haematoma or a retroperitoneal collection. Although non-access site bleeding such as gastrointestinal bleeds (accounting for 10% of protocol-defined major bleeds) and intracerebral/intraocular bleeds (3% of bleeds) occur less frequently, their clinical ramifications are often profound.

Several studies examining major bleeding during PCI have also examined which patients are most at risk.9 11 12 Predictors reliably identified in multivariate analyses include increased age, female sex, chronic renal impairment, history of hypertension, baseline anaemia, glycoprotein inhibitor use, intra-aortic balloon pump use and complex procedures. The impact on bleeding rates when these risk factors are present together has not been formally studied as yet, although it seems likely that a cumulative effect would occur. Other factors traditionally associated with bleeding such as low body weight, diabetes and cardiogenic shock, although univariate associates, are not independent predictors. The adverse outcome after bleeding may relate in part to the associated risk factors—such as diabetes, age or chronic renal failure—but multivariate modelling assessing the independent impact of bleeding on mortality identifies periprocedural bleeding as an important predictor. Although these models may not be sufficiently powerful fully to exclude the influence of risk factors, it is certainly striking to observe the consistent effect of bleeding on outcome demonstrated in several observational and randomised series of different patients using different statistical models.

Importantly, most of the predictors of major bleeding are evident preprocedurally and thus should allow identification of patients at risk. Recently, a formal bleeding risk score—based on the REPLACE study data—was developed.15 Five major preprocedural characteristics were identified as independent predictors of protocol-defined major bleeding: advancing age, female gender, estimated glomerular filtration rate <60 ml/min, baseline anaemia and low molecular weight heparin within 48 h of the PCI. Figure 1 shows that low and intermediate risk scores are associated with a low risk of protocol-defined major bleeding, but a score >10 led to a marked increase in the number of bleeds seen. This association was present in both access site and non-access site bleeds. The application of this risk score and subsequent tailoring of the procedure and adjuvant pharmacotherapy is discussed below.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Summary of bleeding score proposed by Nikolsky et al.15 The score takes into account five preprocedural variables to calculate the likelihood of major bleeding during percutaneous coronary intervention: age, sex, estimated glomerular filtration rate (eGFR), baseline haematocrit (Hct) adn low molecular weight heparin (LMWH) use.

BLEEDING INDEPENDENTLY PREDICTS ADVERSE OUTCOMES

As alluded to above, evidence is now accumulating relating the impact of major bleeding during PCI to clinical outcomes. In the first study to examine this question, in-hospital mortality in patients with protocol-defined major bleeding was 7.5% (compared with 0.6% in patients with no bleeding).11 Bleeding was an independent predictor of in-hospital death with an odds ratio (OR) of 3.5. Although there is a paucity of studies examining this relationship, there is a remarkable consistency in the available data. In the ACUITY study a profound impact of protocol-defined major bleeding on 30-day mortality was seen.9 Mortality in patients with bleeding was 7.3% compared with 1.2% in patients without bleeding. Even more striking was the observed OR of 7.55 for mortality. This observed OR was twofold higher than any of the traditional risk factors for death after PCI, including age, renal impairment, diabetes or troponin elevation. As well as impacting on short-term outcomes, more recent data suggest that bleeding may adversely affect 1-year mortality. In the REPLACE-2 study the occurrence of bleeding was associated with an adjusted odds ratio of 3.5 for 1-year mortality.16 This relationship was stronger than the effect on mortality of in-hospital myocardial infarction (MI). The negative impact of bleeding on intermediate-term outcomes may not only be restricted to larger bleeds. In a pooled analysis of GUSTO-IIb, PURSUIT and PARAGON, even minor bleeding (as defined by the GUSTO criteria) had a significant impact on 6-month mortality, with a stepwise relationship between bleeding severity and mortality.17

These data therefore serve to illustrate the importance of major bleeding and focus operator attention on minimising risk as far as possible. Why should bleeding have such an impact on mortality? One explanation is the direct result of the bleed itself and an intracerebral haemorrhage is an obvious example of this.18 Indirect complications relating to a bleeding event, such as the need for invasive monitoring, reparative surgery, endoscopy, anaesthesia, all increase the likelihood of adverse outcomes. However, there are other more subtle explanations for the observed relationship between bleeding and outcome. Significant bleeding often necessitates early discontinuation of antithrombotic/antiplatelet agents, leading to ischaemia, infarction, repeat procedures and stent thrombosis. Indeed, rates of Q-wave and non-Q-wave MI in published studies are three- to fivefold higher in patients with bleeding than in controls. In the ACUITY study, PCI stent thrombosis rates were 3.4% in patients with bleeding and 0.6% in patients without bleeding.9

Higher rate of blood transfusion in patients with bleeding may also paradoxically be associated with adverse outcomes. Contrary to what might be expected, an association between transfusion and adverse outcomes has been demonstrated in patients recruited to both observational and randomised studies in various patient groups. In the largest randomised trial to date, a liberal blood transfusion policy (in an intensive care setting at least) has been shown to be associated with poorer outcomes than a restrictive strategy.19 This association between transfusion and survival was also seen in patients undergoing PCI. Registry data showed that in-hospital mortality in patients with major bleeding who underwent transfusion was 10.6% compared with 5.1% in those who did not receive blood.11 One-year mortality was also higher among patients who received blood (major bleeding with transfusion 22.7% vs 13.6% without transfusion). Additionally, there was a significant association between the number of units transfused and 1-year mortality rates (odd ratio 1.47 per unit transfused), with transfusion being a powerful independent predictor of not only in-hospital mortality but also 1-year mortality. A similar association was also seen in unselected patients with acute coronary syndrome (ACS) followed up by the GRACE registry.20 Transfusion again was an independent predictor of 1-year mortality in that study.

Why should transfusion be so bad for patients? Intuitively, raising the haemoglobin by transfusion should increase oxygen delivery, but studies show that measures of tissue oxygenation either decrease or do not change.21 Prolonged storage of red blood cells depletes their NO stores and after transfusion they may function as NO “sinks,” promoting vasoconstriction and platelet aggregation.22 The oxygen affinity of stored haemoglobin is also increased, reducing tissue oxygen delivery.23 Other potentially deleterious consequences of transfusion on cardiac performance include increased blood viscosity, free radical production, fluid overload, increased levels of inflammatory markers, electrolyte imbalance and hypothermia. Finally, transfusion results in circulation of platelets not previously exposed to aspirin or clopidogrel and the potential drawbacks of this in the acute setting are obvious.

As might be expected, hospital stays are significantly longer in patients experiencing a major bleed. In the Washington Hospital Centre series duration of hospital stay was 8.9 days for patients with protocol-defined major bleeding compared with 3.1 days for those without bleeding.11 Consistent with prolonged hospital stays, blood transfusion and higher rates of repeat or reparative procedures, studies suggest that bleeding is also an expensive complication. Economic analyses estimate the incremental cost of bleeding to range from US$6022 to US$11 767 for each event. Accordingly, any economic studies of interventions designed to reduce bleeding must not only consider the cost of the intervention itself but also the potential saving of reduced bleeding events.

TAILORING THE INTERVENTIONAL APPROACH TO MINIMISE RISK

In view of the accumulating evidence demonstrating the impact of bleeding on short- and medium-term outcomes, careful case selection is crucial in minimising bleeding complications. As reported recently, the COURAGE trial demonstrates excellent outcomes with medical management in certain patients with stable angina.24 Thus this trial (and others previously) provides justification for operators to recommend conservative management of patients at high-risk of bleeding who have low- and moderate-risk stable angina.

If angioplasty is to be undertaken, a preprocedural bleeding assessment will allow identification of patients at particular risk. A formalised scoring system based on the REPLACE studies may help clinicians in risk assessment.15 As discussed, low and moderate pre-test risk scores are associated with protocol-derived major bleeding rates of around 2%. However, high scores (>10) are associated with greater than double the risk of major bleeding (5%) compared with low and intermediate risk. These data should alert interventionalists to tailor procedures accordingly.

Choice of access site is of crucial importance in minimising risk. As discussed previously, up to three-quarters of major bleeding results from femoral artery access. Additionally, although femoral artery closure devices reduce time to haemostasis, there is little evidence that vascular complications are reduced with their use.25 In fact, there is some evidence to the contrary, with an association between retroperitoneal bleeds and closure devices reported.26 Lowest bleeding rates, therefore, might be expected with routine wrist access. The safety of radial (and indeed ulnar) access appears to be well established in patients presenting with both stable angina and ACS and should be the default access site in patients at risk of bleeding.27^–29 Of course, the radial route is not the panacea for complete abolition of bleeding. Gastrointestinal bleeding requiring transfusion was seen in 2.9% of patients undergoing rescue PCI from the radial route.30 Additionally, patients at highest risk of bleeding—such as the elderly—are technically the most demanding for intervention from the radial artery. Finally, complex procedures—in particular those necessitating the use of a balloon pump—can in some cases only be completed from the groin. As discussed previously, data demonstrate that such cases are at high risk of bleeding.

Careful choice of adjunctive pharmacology is also important in minimising bleeding risk. The ISAR-1 and ISAR-2 studies demonstrate that heparin is sufficient antithrombotic therapy in many low-/intermediate-risk patients with stable angina and patients with troponin-negative ACS.10 31 However, appropriate heparin dosing is important in minimising bleeding, with careful attention paid to avoidance of overprolongation of the activated clotting time (ACT). A clear relationship exists between bleeding frequency and excess ACT and paradoxically—perhaps due to platelet activation—excess ACT also seems to increase ischaemic complications.32^–34 Although low molecular weight heparins seem equivalent to unfractionated heparin in preventing ischaemic complications, their use may be associated with increases in bleeding rates. In the SYNERGY study, TIMI major bleeding occurred in 9.1% of patients randomised to unfractionated heparin compared with 7.6% of those receiving enoxaparin (p = 0.008).35

Although previous randomised trials of glycoprotein inhibitors using lower heparin dosing are reported as showing infrequent bleeding, critical appraisal of the data does not back up these claims. For example in ESPRIT, TIMI major bleeding occurred in 1% of patients receiving eptifibatide compared with 0.4% of controls.14 However, if TIMI minor bleeding is included (which most operators would probably consider clinically significant) then bleeding occurred almost twice as often in the eptifibatide group at rates that are consistent with other studies (3.8% with eptifibatide vs 2.1% in controls). Accordingly, most studies of bleeding predictors demonstrate that glycoprotein inhibitors (regardless of ACT levels) are independently associated with excess bleeding and thus, careful consideration of their benefit versus risk should be made before their use.9 11 In patients at high risk of bleeding who are also deemed at high risk of ischaemic complications, a balanced judgment needs to be made as to whether glycoprotein inhibitors should be used or not. Certainly, in such cases the radial route should be the preferred access route.

If adjunctive treatment over and above heparin alone is required then pharmacology other than glycoprotein inhibitors should be considered. This is particularly relevant in patients at increased risk of bleeding. Bivalirudin, a direct thrombin inhibitor, has been studied in two major trials and significantly reduces protocol-defined major bleeding. In the REPLACE-2 trial, 6010 patients with stable angina or settled ACS were randomised to glycoprotein inhibitor (GPI) plus heparin or bivalirudin alone during PCI.6 36 The composite end point of death, MI or repeat revascularisation was seen in 7.1% of patients in the glycoprotein group compared with 7.6% of the bivalirudin group. However, major bleeding (as defined by the REPLACE investigators) occurred in 4.1% of the glycoprotein group compared with 2.4% of the bivalirudin group (p<0.001). In a study of patients with higher risk ACS, the ACUITY investigators again demonstrated non-inferiority in combined ischaemic end points (8% in glycoprotein group vs 9% in the bivalirudin group, p = NS) but with a significant reduction in protocol-defined major bleeding in the bivalirudin-treated patients (7% vs 4%, p<0.0001).7 Thus these studies demonstrate in stable patients and patients with ACS that bivalirudin significantly reduces the occurrence of protocol-defined major bleeding during PCI. The most notable reduction in bleeding was through reduction in groin haematoma, with significant reductions in retroperitoneal, intracerebral and gastrointestinal bleeds noted in the bivalirudin arm of REPLACE-2 (combined 1.4% vs 0.3%, p<0.05). Retroperitoneal bleeds were also significantly fewer with bivalirudin in the ACUITY study. Although the effects of a retroperitoneal bleed on mortality are well recognised, the significant impact of large femoral haematomas on morbidity, length of stay and hospital costs should not be underestimated.

One caveat to these studies came from the recent meta-analysis of ACUITY, REPLACE-2 and PROTECT-TIMI 30 (a further study of bivalirudin vs GPI with an angiographic primary end point).7 This meta-analysis demonstrated a non-significant trend (p = 0.15) for a reduction in ischaemic end points in the patients receiving GPI treatment. This has led some bivalirudin operators to continue using GPI agents in patients perceived to be at highest risk of an ischaemic end point.

Figure 2 shows an algorithm summarising a tailored approach to minimising bleeding risks.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 An algorithm summarising a tailored approach to minimise the frequency of periprocedural major bleeding. GPI, glycoprotein inhibitor.

CONCLUSIONS

The revised 2007 European Society of Cardiology guidelines for the treatment of non-ST elevation ACS specifically dealt with avoidance and management of bleeding and was the first time that any of the major representative bodies had issued such guidance. The preamble states that: “Among crucial new issues, bleeding is highlighted, considered in the past as a non-event. It has become evident that bleeding is indeed a serious event with a catastrophic impact on outcomes, with a four-fold increase in death and MI at 30 days and long-term.” Thus, operators should be mindful of this emerging problem with careful and appropriate case selection, preprocedural risk assessment and radial/ulnar access used where possible. Evidence-based choice of adjunctive pharmacotherapy is of central importance with carefully managed heparin dosing and bivalirudin administration encouraged rather than GPIs in patients at particular risk of bleeding. The same care should be taken to prevent bleeding as is taken to minimise ischaemic complications to optimise patient outcomes after PCI.