Improving clinical outcomes by reducing bleeding in patients with non-ST-elevation acute coronary syndromes

Abstract

Introduction

The primary goal of antithrombotic therapy in patients with acute coronary syndromes (ACS) is to prevent recurrent ischaemic events and death. Antiplatelet and anticoagulant drugs are effective for preventing myocardial infarction (MI) and death in patients with ACS but also cause bleeding.^1,2 In the past, clinicians have considered bleeding to be an inevitable consequence of effective antithrombotic therapy and the avoidance of bleeding therefore received little attention. More recently, there has been increasing awareness that bleeding is associated with an increased risk of adverse outcomes, including MI, stroke, and death.^3–6 Approximately 5% of patients presenting with ACS experience major bleeding during the next 30 days,^3,7 and the incidence of major bleeding is higher in patients with ACS undergoing early invasive procedures.^8–11 Strategies that reduce the risk of bleeding while preserving or enhancing efficacy are needed.

The Organization to Assess Ischemic Syndromes (OASIS)-5 trial demonstrated that fondaparinux compared with enoxaparin reduced major bleeding by about one-half during treatment.¹² The objective of this paper was to examine the effect of fondaparinux compared with enoxaparin on different types of bleeding in the OASIS-5 trial and to examine the impact of bleeding on death and recurrent ischaemic events.

Methods

The Organization to Assess Ischemic Syndromes (OASIS)-5 trial was an international, double-blind, randomized controlled trial of 20 078 patients with non-ST-elevation ACS. The inclusion and exclusion criteria and primary efficacy and safety outcomes have been published previously.¹³ Briefly, patients were eligible for inclusion if they presented within 24 h after the onset of symptoms and met at least two of the three following criteria: an age of at least 60 years, an elevated level of troponin or creatine kinase MB isoenzyme, or electrocardiographic changes indicative of ischaemia. Patients with contraindications to low-molecular weight heparin, recent haemorrhagic stroke, indications for anticoagulation other than an ACS, or a serum creatinine level of at least 3 mg/dL (265 µmol/L) were excluded.

Patients were randomized to receive subcutaneous fondaparinux 2.5 mg o.d. or subcutaneous enoxaparin, 1 mg/kg b.i.d. In patients whose creatinine clearance <30 mL/min, the dose of enoxaparin was reduced to 1 mg/kg o.d.

Fondaparinux was given until hospital discharge or for up to 8 days (whichever occurred first). Enoxaparin was given for 2–8 days or until the patient was in clinically stable condition, which is in accordance with its approved North American labelling for the management of patients with unstable angina and MI without ST-segment elevation.

The primary efficacy outcome was the composite, death, MI, or refractory ischaemia. Information on strokes was also systematically collected. The primary safety outcome was major bleeding. The primary efficacy and safety outcomes were adjudicated in a blinded fashion by a committee. Patients were followed for up to 6 months (180 days).

Definition of major bleeding

Bleeding was defined according to the criteria used in the Efficacy and Safety of Subcutaneous Enoxaparin in Non-Q-Wave Coronary Events (ESSENCE) trial.^14,15 The use of this definition was mandated by the Food and Drug Administration because enoxaparin was approved for the management of ACS based on the results of the ESSENCE trial. Thus, major bleeding was defined as clinically overt bleeding that was fatal (bleeding reported to cause death), symptomatic intracranial haemorrhage, retroperitoneal haemorrhage, intraocular haemorrhage leading to significant vision loss, a decrease in haemoglobin of at least 3.0 g/dL (with each blood transfusion unit counting for 1.0 g/dL of haemoglobin), or bleeding requiring transfusion of two or more units of red blood cells or equivalent of whole blood. All other clinically significant bleeding not meeting the definition for major bleeding and which led to interruption of the study drug for at least 24 h, surgical intervention, or transfusion of not more than one unit of blood was defined as minor bleeding.^14,15 If a patient experienced both major and minor bleeding we only considered the most serious bleed.

Statistical analysis

All the analyses were performed using SAS software, version 9.1 (SAS Institute Inc., Cary, NC, USA).

The incidence, severity [fatal, major (excluding fatal), minor], and sites of bleeding were compared in patients randomized to receive fondaparinux compared with enoxaparin.

Baseline characteristics among patients who developed bleeding compared with those who did not develop bleeding were compared using a χ² test for categorical variables and a t-test for continuous variables.

We developed several statistical models to explore the association between bleeding and outcome in the overall study population and in patients randomized to receive fondaparinux compared with enoxaparin. The first model examined a patient's propensity to major bleeding using logistic regression, incorporating both baseline characteristics and co-interventions associated with major bleeding as independent variables and bleeding as the dependent variable. We included the following variables: treatment allocation, age, sex, clinical diagnosis, heart rate, systolic blood pressure, creatinine, body mass index, ST segment shift on the presenting electrocardiogram, haemoglobin, baseline antithrombotic drug use, interventions prior to bleeding, smoking status, history of diabetes, history of hypertension, history of cancer, prior stroke, prior MI, prior peripheral artery disease, prior percutaneous coronary intervention (PCI), and prior coronary artery bypass graft (CABG) surgery. We included interactions and higher order terms in the model. In the subsequent models, we examined the association between major bleeding and minor bleeding and MI, stroke, and death using Cox regression after graphically testing the assumption of proportionality. The propensity score was included in these models as a linear term. We only included co-interventions that were used prior to bleeding and outcomes that occurred after bleeding. Major and minor bleeding were incorporated as time-dependent covariates to account for survival bias and for the possibility that timing of bleeding relative to an outcome may be important (for example, bleeding that occurs after a non-fatal outcome). A two-sided P-value of 0.05 was considered statistically significant.

Results

These analyses included 20 078 patients enrolled in the OASIS-5 trial, of whom 990 (4.9%) developed major (including fatal) bleeding and 423 (2.1%) developed minor bleeding during follow-up to 180 days.

Bleeding outcomes according to randomized treatment allocation

Bleeding outcomes in the OASIS-5 trial are summarized in Table 1 and the cumulative hazard of different categories of bleeding is presented in Figure 1.

Figure 1

Open in new tabDownload slide

Cumulative hazard of the occurrence of fatal bleeding, non-fatal major bleeding, and minor bleeding.

Fondaparinux compared with enoxaparin reduced fatal bleeding, non-fatal major bleeding, and minor bleeding, as well as the need for transfusion during the first 9 days. Gastrointestinal, retroperitoneal, and puncture site bleeding were also significantly reduced (Table 1).

The incidence of major (excluding fatal) bleeding over the whole study period (up to 180 days) was significantly lower in patients randomized to fondaparinux compared with enoxaparin [400/10 057 (4.0%) vs. 537/10 021 (5.4%), relative risk (RR) 0.73; 95% CI 0.64–0.82, P < 0.0001]. Similar reductions in bleeding were observed throughout the study period with fondaparinux compared with enoxaparin for the categories of fatal bleeding [20/10 057 (0.20%) vs. 33/10 021 (0.4%), RR 0.58, 95% CI: 0.22–1.02, P = 0.053] and minor bleeding [115/10 057 (1.2%) vs. 308/10 021 (3.3%), RR 0.37; 95% CI 0.30–0.45, P < 0.0001], and for TIMI major bleeding (data not shown). There was also a significant reduction in the need for transfusion [336/10 057 (3.3%) vs. 429/10 021 (4.3%), RR 0.78, 95% CI 0.67–0.90] throughout the study period.

Baseline characteristics in patients who developed bleeding and in those who did not develop bleeding

Baseline patient characteristics and co-interventions in patients who developed fatal, major (excluding fatal), or minor bleeding and in those who did not develop bleeding are summarized and compared in Table 2.

Patients who developed bleeding were older and more likely to be female; were more likely to have a history of diabetes, hypertension, stroke, or peripheral arterial disease; and were more likely to have heart failure, have a higher creatinine level, lower baseline haemoglobin level, or ST-segment changes on the electrocardiogram at presentation. Patients with bleeding were also more likely to be treated with a thienopyridine (clopidogrel or ticlopidine), a glycoprotein IIb/IIIa inhibitor, or intravenous unfractionated heparin; and were more likely to undergo coronary angiography or PCI, or to undergo CABG surgery prior to the bleed.

Occurrence of deaths and ischaemic events in relation to bleeding

During the first 30 days, approximately one in six of all deaths occurred in patients who experienced bleeding (minor or major). Two-thirds (66.7%) of the 57 excess deaths in patients treated with enoxaparin compared with fondaparinux occurred in patients who experienced bleeding (Table 3).

During the 180 days of study follow-up, approximately one in eight of all deaths occurred in patients who experienced bleeding during the first 9 days. More than 90% of the 64 excess deaths in patients treated with enoxaparin compared with fondaparinux occurred in patients who experienced bleeding during the first 9 days (Table 3).

Examination of the association between the composite outcome death/MI/stroke and bleeding revealed a similar pattern: During the first 30 days, approximately one in seven of the composite outcome death/MI/stroke occurred in patients who experienced bleeding, and during 180 days of study follow-up approximately one in nine of these events occurred in patients who experienced bleeding during the first 9 days (Table 3).

Association between bleeding and death or recurrent ischaemic events

The associations between major (non-fatal) or minor bleeding and the composite outcome death/MI/stroke or individual components of this outcome at 30 and 180 days are summarized in Table 4.

In the overall patient population, major bleeding during the first 30 days was associated with about four-fold increase in death/MI/stroke [21.8 vs. 6.2%, hazard ratio (HR) 3.00, 95% CI 3.30–4.82] and a similar increase in each of the individual components of this composite outcome, death (8.4 vs. 2.7%, HR 3.46, 95% CI 2.60–4.60), MI (8.3 vs. 3.4%, HR 4.39, 95% CI 3.45–5.59), and stroke (3.0 vs. 0.7%, HR 4.66, 95% CI 2.83–7.65) at 30 days (P < 0.0001 for each analysis, Table 4). All the analyses were adjusted for the propensity to develop bleeding, and the timing of bleeding in relation to non-fatal outcomes. The results were similar irrespective of whether serum creatinine or creatinine clearance was included in the model.

The magnitude of the association between major bleeding and outcome (death/MI/stroke and individual components of the composite outcome) was attenuated at 180 days (about a three-fold increase in risk), but remained highly significant (P < 0.0001 for each analysis, Table 4). Minor bleeding was also independently associated with an increased risk of death/MI/stroke and for death as an individual component at 30 and 180 days (Table 4).

Discussion

Our results indicate that one in 20 patients in the OASIS-5 trial experienced major bleeding and that more than one-quarter of those who developed major bleeding experienced death, MI, or stroke during follow-up till 180 days. One of every six deaths during the first 30 days occurred in patients who experienced bleeding (major or minor), and one of every eight deaths during 180 days of study follow-up occurred in patients who experienced bleeding. The use of fondaparinux compared with enoxaparin for up to 9 days reduced fatal bleeding, non-fatal major bleeding, and minor bleeding by about one-half; reduced puncture site, and retroperitoneal bleeding, and reduced the need for packed red blood cell transfusions. The vast majority of excess deaths in patients treated with enoxaparin occurred in patients who experienced bleeding. Our results also demonstrate an independent, strong, temporal, and dose-related (higher risk of death in those with more severe degrees of bleeding) association between bleeding and death, MI, and stroke.

Major bleeding was more common than the individual outcomes of death or MI in the OASIS-5 trial and was associated with an increased risk of death or non-fatal adverse outcomes irrespective of whether patients were treated with fondaparinux or enoxaparin. The strong and independent relation between bleeding and adverse outcomes and the excess bleeding in patients treated with enoxaparin compared with fondaparinux is consistent with the conclusion that the benefits of fondaparinux for preventing non-fatal ischaemic events and death are mediated primarily by a reduction in bleeding.

The fixed 2.5 mg o.d. dose of fondaparinux used in the OASIS-5 trial is the same as the dose that was used in venous thrombo-embolism prevention trials. In contrast, the 1 mg/kg dose of enoxaparin given o.d. or b.i.d. is substantially higher than the 30 mg b.i.d. or 40 mg o.d. dose of enoxaparin used in venous thrombo-embolism prevention trials. Because both of these drugs are renally excreted, enoxaparin is more likely than fondaparinux to have achieved excessive anticoagulant levels in patients with renal impairment in the OASIS-5 trial.¹⁶

There is now an overwhelming body of evidence that bleeding is an independent predictor of adverse outcome across the spectrum of patients with coronary artery disease, including patients with non-ST-elevation ACS, ST-elevation MI, and those undergoing PCI. The mechanism of the association between bleeding and adverse outcome remains to be established but several possibilities should be considered. First, effective antithrombotic treatments are usually stopped in patients who develop bleeding, which could result in an increased incidence of non-fatal or fatal ischaemic events.¹⁷ The ACUITY investigators reported a greater than five-fold increase in stent thrombosis among patients who experienced major bleeding,⁵ which is consistent with previous reports linking premature discontinuation of antiplatelet drugs in coronary stent patients with ischaemic events and death.^18,19 Secondly, bleeding can lead to reduced oxygen supply, caused by reduced tissue perfusion and anaemia, as well as increased oxygen demand, caused by stress and tachycardia, thereby resulting in tissue ischaemia. Thirdly, transfusion of stored packed red blood cells can lead to reduced tissue oxygen delivery, caused by depletion of 2,3-diphosphoglycerate and nitric oxide from red blood cells, which leads to increased affinity of haemoglobin for oxygen and impaired release of oxygen at the capillary level. Packed red blood cell transfusion²⁰ and the duration of storage of packed red blood cells prior to transfusion have been associated with an increased risk of adverse outcomes.²¹ Fourth, interventions, including surgery and anaesthesia, that are performed to treat bleeding complications may carry additional risk, particularly in patients who have recently experienced an ACS and have discontinued their antithrombotic medications.

The strengths of our study are that we analysed a dataset involving more than 20 000 patients with ACS randomized to fondaparinux or enoxaparin, including almost 1000 major bleeds, more than 1000 deaths, more than 1000 MIs, and more than 250 strokes. Detailed information concerning patient demographics, baseline characteristics, and co-interventions were collected, allowing us to obtain reliable estimates of the incidence and prognosis of bleeding.

Our study also has limitations. First, despite the large sample size and substantial number of outcomes, our study was not powered to determine the effect of fondaparinux compared with enoxaparin for uncommon but serious outcomes such as intracranial bleeding. Secondly, we were unable to explore the effect of discontinuation of concomitant antithrombotic therapies on adverse outcomes in the OASIS-5 trial because this information was not routinely collected in all patients, and, when available, was confounded by the severity of the bleeding. Thirdly, the association between bleeding and outcome in the OASIS-5 trial is subject to confounding because older and sicker patients are more likely to experience bleeding, but are also more likely to experience MI, stroke, or death. Despite exhaustive effort to adjust for potential confounders, we are unable to prove a causal relation between bleeding events and death. However, our data add to the growing body of consistent evidence that bleeding is an independent and potentially modifiable determinant of recurrent ischaemic events and death. Fourth, very few minor bleeds were reported after day 9, raising the possibility that minor bleeding after the period of study drug administration was under-reported. Finally, we performed multiple statistical comparisons, which increases the likelihood that some of our findings are due to the play of chance.

In conclusion, our analyses involving more than 20 000 patients in the OASIS-5 trial demonstrate that the routine use of fondaparinux in place of enoxaparin in patients with ACS will reduce the risk of bleeding by up to one-half and that prevention of bleeding translates into substantial reductions in both morbidity and mortality during 180 days of follow-up. Increasing recognition that bleeding is an independent determinant of outcome in patients with ACS should prompt efforts to reduce the risk of bleeding and thereby improve clinical outcomes by appropriate selection of antithrombotic therapies as well as judicious use of invasive revascularization procedures.

Conflict of interest: none declared.

References

Author notes