Abstract

Aims

To evaluate the association of bleeding with mortality in ST-elevation myocardial infarction (STEMI).

Methods and results

We studied 20 323 patients with STEMI receiving fibrinolytic therapy and an antithrombin in ExTRACT-TIMI 25. Relationships between in-hospital bleeding, patient characteristics, treatments, and in-hospital cardiovascular complications with mortality were evaluated using Cox models. Likelihood ratios estimated each variable's model contribution. High 30-day mortality after major bleeding (n = 309, 37.6% mortality) was driven by the poor prognosis of intracranial haemorrhage (ICH; n = 143, 65.4% mortality, model contribution 7.8%). The adjusted hazard ratios (HRs) for 30-day death for any major bleeding and for ICH were 2.9 [2.4–3.6] and 10.3 [8.2–12.8], respectively. Neither non-ICH major nor minor bleeding was associated with 30-day death after adjustment. Cardiogenic shock (HR 13.5, 61% contribution) and age (HR 1.6/decade, 17% contribution) were most strongly correlated with 30-day death. Among 30-day survivors, age (HR 1.6/decade, contribution 43%) and heart rate (HR 1.2 per 10 b.p.m., contribution 18%) were most strongly associated with mortality between Days 31 and 365.

Conclusion

Cardiogenic shock, age, and ICH were important independent correlates of 30-day and 1-year mortality in STEMI patients receiving fibrinolytic therapy. In-hospital non-ICH major and minor bleeding were not independently associated with increased mortality at 30 days or 1 year.

See page 2077 for the editorial comment on this article (doi:10.1093/eurheartj/ehq173)

The ExTRACT-TIMI 25 trial1 compared adjunctive anticoagulation with enoxaparin throughout the index hospitalization with unfractionated heparin (UFH) administered for 48 h in patients with ST-elevation myocardial infarction (STEMI) treated with fibrinolytic therapy. Enoxaparin was associated with a reduction in the primary composite of death and non-fatal MI (9.9 vs. 12.0%, P < 0.001), but an increase in major bleeding (2.1 vs. 1.4%, P < 0.001). There were no significant differences between enoxaparin and UFH in intracranial haemorrhage (ICH) (0.83 vs. 0.65%, P = 0.14) or in mortality at 30 days (6.9 vs. 7.5%, P = 0.11) or 1 year (10.5 vs. 10.6%, P = 0.72).2

Several studies have demonstrated an association between bleeding and mortality in patients with non-STE acute coronary syndromes (ACS)3–5 and STEMI.6,7 Efforts to reduce bleeding in patients with ACS are reflected in the development of safer antithrombotics,8 alternative methods of arterial access,9 enhancing haemostasis post-coronary angiography,10 and the use of preventive therapies (e.g. proton-pump inhibitors to prevent gastrointestinal haemorrhage). However, the quantitative relationship,6 causality,11,12 and impact of bleeding on late mortality13,14 remain uncertain and controversial. We undertook this analysis to ascertain the independent contribution of bleeding to mortality among patients enrolled in ExTRACT-TIMI 25, taking into account patient characteristics, treatments, and other cardiovascular complications.

Methods

Details of the ExTRACT-TIMI 25 trial have been published previously.1,15 In brief, this trial randomized 20 506 STEMI patients up to 6 h after the onset of symptoms for which fibrinolytic therapy was planned. Major exclusion criteria were: presence of cardiogenic shock, pericarditis, aortic dissection, contraindications to fibrinolysis, treatment with low-molecular weight heparin in the prior 8 h, known renal insufficiency [serum creatinine level >220 µmol/L (2.5 mg/dL) for men and >175 µmol/liter (2.0 mg/dL) for women], or life expectancy <12 months. Patients were randomized 1:1 to either a strategy using enoxaparin or UFH in a double-dummy, double-blind fashion beginning between 15 min before and 30 min after the start of fibrinolysis.

Enoxaparin was to be given as a 30 mg intravenous bolus followed by a subcutaneous injection of 1.0 mg/kg, with injections administered every 12 h. In patients aged 75 or older, the bolus dose was omitted and the maintenance dose reduced to 0.75 mg/kg. For patients with a creatinine clearance <30 mL/min, the dose was 1.0 mg/kg every 24 h. The dose of UFH was 60 U/kg IV (maximum 4000 U) followed by an infusion of 12 U/kg/h (initial maximum 1000 U/h) with titration to achieve a target activated partial thromboplastin time of 1.5–2.0 times the control. The enoxaparin strategy was continued for 8 days or until hospital discharge (whichever came first), and UFH was continued for at least 48 h. We excluded patients who did not receive any anticoagulant study drug or fibrinolytic therapy, leaving 20 323 patients for analysis.

Bleeding definitions

Bleeding was defined using the TIMI scale as either major (ICH or overt bleeding resulting in a loss of haemoglobin >5 gm/dL adjusted for transfusion) or minor (overt bleeding resulting in a loss of haemoglobin of 3–5 gm/dL adjusted for transfusion)15 and assessed throughout the index hospitalization or Day 8 (which was also the maximum duration of the administration of study drug), whichever came first. If a patient had multiple bleeding events, we assigned them the bleeding event of highest severity in the following order: (1) ICH; (2) non-ICH major; and (3) TIMI minor bleeding.

Endpoints

The primary endpoint of interest in this analysis was all-cause mortality through 30 days. The secondary endpoint was 1-year mortality. A landmark analysis explored mortality between Days 30 and 365 among patients who survived to Day 30, by calculating the conditional probability of an event given no events were observed in the first 30 days. The cause of death, bleeding severity, and location were adjudicated by a blinded, independent clinical endpoint committee.

Statistical models

Multivariate Cox models predicting mortality were constructed in a stepwise fashion adding covariates from the five groups below to construct a final model that included all statistically significant variables from each of the five steps. In each step, we used backward elimination, requiring a multivariate P < 0.05 for retention in the model. Variables with >5% missing data were not considered as candidate variables.

Since bleeding was the focus of this analysis, bleeding variables were maintained in the models regardless of their level of statistical significance. Major bleeding events were also analysed as either ICH or non-ICH major bleeding separately, since these events are known to have different effects on short-term mortality. Other variables of interest (e.g. cardiogenic shock) were removed or forced into the model to evaluate for confounding. Since the relative impact of MI and bleeding on mortality is of clinical interest, we also explored models that included recurrent MI between randomization and discharge. Likelihood ratios were used to estimate the contribution of each variable to the final model, by calculating the % of the variable's likelihood ratio divided by that for the full model.

  • (1) Most severe type of bleeding during the index hospitalization (up to 8 days maximum).

  • (2) Patient characteristics at randomization, including age, sex, race, diabetes; history of hypertension, angina, MI, or percutaneous coronary intervention (PCI); current smoking, Killip class, systolic blood pressure, heart rate, location of MI, presence of left bundle branch block, time from symptom onset to fibrinolytic.

  • (3) Medical therapies including: use of aspirin or non-steroidal anti-inflammatory agents, UFH, or low-molecular weight heparin within 7 days prior to randomization; treatment with enoxaparin or UFH post-randomization; administration of a fibrin-specific lytic vs. streptokinase for the qualifying MI.

  • (4) Invasive procedures prior to discharge (or Day 8) including: diagnostic coronary angiography, PCI, coronary artery bypass grafting (CABG), intra-aortic balloon counterpulsation, and permanent pacemaker insertion during the index hospitalization.

  • (5) Cardiovascular events between randomization and hospital discharge (or Day 8) including: cardiogenic shock, severe congestive heart failure, non-haemorrhagic stroke, recurrent MI, and recurrent ischaemia.

The Wilcoxon rank-sum test (for continuous variables) and χ2 test (for categorical variables) were used to compare patient baseline characteristics. Cox proportional hazards models were constructed to assess the risk of mortality for the different types of bleeding events. We used the Schoenfeld16 residuals test to assess the proportionality of the hazards. For models that violated the proportional hazards assumption, we introduced a time-dependent covariate and split the time interval at the time the mortality rates crossed. If the piecewise models no longer violated the proportionality hazards assumption, then two hazard ratios (HRs) were reported, one for each time interval. Otherwise, we reported the HR for the midpoint of the time interval. Analyses were performed with Stata/IC version 10.1 (Stata Statistical Software, StataCorp, College Station, TX, USA).

Results

Among 20 323 subjects in the ExTRACT-TIMI 25 trial, 309 (1.5%) had a TIMI major bleeding event during the index hospitalization. Of these, 143 (0.70%) suffered an ICH and 166 (0.82%) a non-ICH TIMI major bleeding event. There were 386 (1.9%) patients who had TIMI minor bleeding without TIMI major bleeding, and 19 628 (96.6%) with no TIMI major or minor bleeding during the index hospitalization.

Patient characteristics according to the type of in-hospital bleeding are summarized in Table 1. Patients with bleeding (compared with those with no bleeding) during the index hospitalization were more likely to be older, female, a non-smoker, with lower body weight, and higher TIMI risk score.17 These same factors were also more prevalent in patients with ICH compared with patients with non-ICH major bleeding, with the exception of low body weight. Prior use of aspirin or non-steroidal anti-inflammatory drugs and post-randomization enoxaparin were more prevalent among patients with in-hospital bleeding events, whereas patients with ICH had higher systolic blood pressure and were more likely to have been treated with a fibrin-specific lytic.

Table 1

Patient baseline characteristics

 Most severe bleeding event during hospitalization

 
 ICH Non-ICH major Minor Any None P-values

 
      Any vs. none ICH vs. non-ICH major 
Number of patients 143 166 386 695 19,628   
Median age (years) [IQR] 68 [58–75] 64 [54–72] 68 [57–74] 67 [56–74] 59 [51–69] <0.001 0.006 
Age > 75 years 28 19 23 23 12 <0.001 0.053 
Female sex 34 22 41 35 23 <0.001 0.014 
White race 87 89 89 89 87 0.288 0.509 
Median weight (kg) [IQR] 73 [66–80] 75 [68–82] 70 [64–81] 72 [65–81] 76 [68–85] <0.001 0.379 
TIMI risk score >3 59 45 51 51 35 <0.001 0.020 
Hypertension 57 43 48 49 44 0.012 0.014 
Hyperlipidaemia 21 18 16 18 18 0.710 0.599 
Current smoker 27 46 42 40 48 <0.001 <0.001 
Diabetes mellitus 25 13 16 17 15 0.102 0.011 
Prior MI 13 11 11 12 13 0.292 0.607 
Prior angina pectoris 27 25 28 27 28 0.620 0.651 
Anterior MI 50 40 37 40 44 0.045 0.109 
Left bundle branch block 1.4 1.2 1.0 1.2 0.9 0.496 0.881 
Median heart rate (min−1) [IQR] 75 [66–85] 75 [62–90] 75 [64–88] 75 [64–88] 75 [64–86] 0.596 0.987 
Median systolic blood pressure (mmHg) [IQR] 140 [127–157] 130 [115–150] 131 [120–150] 134 [120–150] 130 [120–150] 0.378 0.001 
Killip class ≥II 16 14 13 11 0.074 0.052 
Median time from symptom onset to fibrinolytic (h) [IQR] 3.6 [2.5–4.6] 2.9 [2.0–4.5] 3.3 [2.4–4.3] 3.3 [2.3–4.4] 3.1 [2.2–4.3] 0.003 0.011 

 
Medical therapies        
 Within the prior 7 days        
  Aspirin or NSAIDs 17 16 16 17 13 0.016 0.702 
  Unfractionated heparin 15 16 16 16 16 0.987 0.702 
  LMWH 1.2 0.3 0.4 0.4 0.972 0.188 
 Randomized to enoxaparin 57 63 60 60 50 <0.001 0.283 
 Fibrin-specific lytic 83 65 69 71 80 <0.001 <0.001 
  Alteplase 53 39 48 47 55   
  Tenecteplase 27 22 17 20 19   
  Reteplase   
  Streptokinase 17 35 31 29 20   
 Most severe bleeding event during hospitalization

 
 ICH Non-ICH major Minor Any None P-values

 
      Any vs. none ICH vs. non-ICH major 
Number of patients 143 166 386 695 19,628   
Median age (years) [IQR] 68 [58–75] 64 [54–72] 68 [57–74] 67 [56–74] 59 [51–69] <0.001 0.006 
Age > 75 years 28 19 23 23 12 <0.001 0.053 
Female sex 34 22 41 35 23 <0.001 0.014 
White race 87 89 89 89 87 0.288 0.509 
Median weight (kg) [IQR] 73 [66–80] 75 [68–82] 70 [64–81] 72 [65–81] 76 [68–85] <0.001 0.379 
TIMI risk score >3 59 45 51 51 35 <0.001 0.020 
Hypertension 57 43 48 49 44 0.012 0.014 
Hyperlipidaemia 21 18 16 18 18 0.710 0.599 
Current smoker 27 46 42 40 48 <0.001 <0.001 
Diabetes mellitus 25 13 16 17 15 0.102 0.011 
Prior MI 13 11 11 12 13 0.292 0.607 
Prior angina pectoris 27 25 28 27 28 0.620 0.651 
Anterior MI 50 40 37 40 44 0.045 0.109 
Left bundle branch block 1.4 1.2 1.0 1.2 0.9 0.496 0.881 
Median heart rate (min−1) [IQR] 75 [66–85] 75 [62–90] 75 [64–88] 75 [64–88] 75 [64–86] 0.596 0.987 
Median systolic blood pressure (mmHg) [IQR] 140 [127–157] 130 [115–150] 131 [120–150] 134 [120–150] 130 [120–150] 0.378 0.001 
Killip class ≥II 16 14 13 11 0.074 0.052 
Median time from symptom onset to fibrinolytic (h) [IQR] 3.6 [2.5–4.6] 2.9 [2.0–4.5] 3.3 [2.4–4.3] 3.3 [2.3–4.4] 3.1 [2.2–4.3] 0.003 0.011 

 
Medical therapies        
 Within the prior 7 days        
  Aspirin or NSAIDs 17 16 16 17 13 0.016 0.702 
  Unfractionated heparin 15 16 16 16 16 0.987 0.702 
  LMWH 1.2 0.3 0.4 0.4 0.972 0.188 
 Randomized to enoxaparin 57 63 60 60 50 <0.001 0.283 
 Fibrin-specific lytic 83 65 69 71 80 <0.001 <0.001 
  Alteplase 53 39 48 47 55   
  Tenecteplase 27 22 17 20 19   
  Reteplase   
  Streptokinase 17 35 31 29 20   

Data shown are per cent of patients unless otherwise indicated. P-values were calculated using the χ2 test for categorical variables and Wilcoxon rank-sum test for continuous variables. ICH, intracranial hemorrhage; TIMI, thrombolysis in myocardial infarction; MI, myocardial infarction; NSAID, non-steroidal anti-inflammatory drug; LMWH, low-molecular weight heparin.

Mortality through 30 days

The all-cause mortality rates through 30 days were 37.6% among patients with TIMI major bleeding during the index hospitalization compared with 9.8% for those with TIMI minor bleeding and 6.6% for those without TIMI major or minor bleeding (P < 0.001) (Figure 1A). The 30-day mortality rates were 65.4 and 13.9% among those with ICH and non-ICH major bleeds, respectively (P < 0.001) (Figure 1B). Sequential adjustment for baseline characteristics, medical therapies, and invasive procedures reduced the HR for patients with ICH and for patients with non-ICH major bleeding, but both remained statistically significant (both P < 0.001, Table 2). Additional adjustment for cardiovascular events during the index hospitalization revealed an important confounding relationship between the development of cardiogenic shock during hospitalization and non-ICH major bleeding. This was further explored in models that grouped patients according to the presence or absence of cardiogenic shock and non-ICH major bleeding and the order of these two events if both were present. Patients who had a non-ICH major bleed either in the absence of cardiogenic shock, or prior to cardiogenic shock, did not exhibit a significantly increased risk of mortality at 30 days (HR 1.3, P = 0.68; HR 2.5, P = 0.24, respectively). In contrast, patients who developed cardiogenic shock prior to discharge but did not have a major non-ICH bleed (HR 19.6), and patients developing cardiogenic shock before a non-ICH major bleed (HR 16.1) had a substantial increase in risk in 30-day mortality (P < 0.001 for both). In the fully adjusted model predicting 30-day mortality, the HR for ICH was 10.3 (8.2–12.8). However, neither non-ICH major bleeding nor minor bleeding was a statistically significant predictor of death from Days 5 to 30 (Table 2) after adjustment for intercurrent hospital clinical events, particularly cardiogenic shock and MI.

Table 2

Risk of mortality through 30 days among patients with bleeding during the index hospitalization

Type of bleeding HR (95% CI) P-value 
All major bleeding   
 Unadjusted 6.4 (5.3–7.8) <0.001 
 Adjusted   
  Baseline characteristicsa only 5.6 (4.6–6.8) <0.001 
  Above + medicationsb 5.9 (4.9–7.2) <0.001 
  Above + in-hospital proceduresc 5.1 (4.2–6.2) <0.001 
  Above + in-hospital eventsd 2.9 (2.4–3.6) <0.001 

 
Intracranial haemorrhage   
 Unadjusted model 13.0 (10.5, 16.0) <0.001 
 Adjusted models   
  Baseline characteristicsa only 10.4 (8.4–12.9) <0.001 
  Above + medicationsb 10.5 (8.4–13.0) <0.001 
  Above + in-hospital proceduresc 9.2 (7.4–11.5) <0.001 
  Above + in-hospital eventsd 10.3 (8.2–12.8) <0.001 

 
Non-ICHa major bleeding   
 HRe Days 5–30 (95% CI)  
 Unadjusted 4.1 (2.5–6.8) <0.001 
 Adjusted models   
  Baseline characteristicsa only 3.8 (2.3–6.2) <0.001 
  Above + medicationsb 3.9 (2.4–6.4) <0.001 
  Above + in-hospital proceduresc 3.1 (1.9–5.2) <0.001 
  Above + in-hospital eventsd 1.0 (0.6–1.6) 0.89 

 
Minor bleeding   
 HRe Days 5–30 (95% CI)  
 Unadjusted 2.9 (2.0–4.3) <0.001 
 Adjusted models   
  Baseline characteristicsa only 1.9 (1.3–2.8) 0.001 
  Above + medicationsb 1.9 (1.3–2.9) 0.001 
  Above + in-hospital proceduresc 1.9 (1.3–2.8) 0.001 
  Above + in-hospital eventsd 1.3 (0.9–2.0) 0.16 
Type of bleeding HR (95% CI) P-value 
All major bleeding   
 Unadjusted 6.4 (5.3–7.8) <0.001 
 Adjusted   
  Baseline characteristicsa only 5.6 (4.6–6.8) <0.001 
  Above + medicationsb 5.9 (4.9–7.2) <0.001 
  Above + in-hospital proceduresc 5.1 (4.2–6.2) <0.001 
  Above + in-hospital eventsd 2.9 (2.4–3.6) <0.001 

 
Intracranial haemorrhage   
 Unadjusted model 13.0 (10.5, 16.0) <0.001 
 Adjusted models   
  Baseline characteristicsa only 10.4 (8.4–12.9) <0.001 
  Above + medicationsb 10.5 (8.4–13.0) <0.001 
  Above + in-hospital proceduresc 9.2 (7.4–11.5) <0.001 
  Above + in-hospital eventsd 10.3 (8.2–12.8) <0.001 

 
Non-ICHa major bleeding   
 HRe Days 5–30 (95% CI)  
 Unadjusted 4.1 (2.5–6.8) <0.001 
 Adjusted models   
  Baseline characteristicsa only 3.8 (2.3–6.2) <0.001 
  Above + medicationsb 3.9 (2.4–6.4) <0.001 
  Above + in-hospital proceduresc 3.1 (1.9–5.2) <0.001 
  Above + in-hospital eventsd 1.0 (0.6–1.6) 0.89 

 
Minor bleeding   
 HRe Days 5–30 (95% CI)  
 Unadjusted 2.9 (2.0–4.3) <0.001 
 Adjusted models   
  Baseline characteristicsa only 1.9 (1.3–2.8) 0.001 
  Above + medicationsb 1.9 (1.3–2.9) 0.001 
  Above + in-hospital proceduresc 1.9 (1.3–2.8) 0.001 
  Above + in-hospital eventsd 1.3 (0.9–2.0) 0.16 

The comparator group consisted of 19,628 patients with no major or minor bleeding during the index hospitalization. Variables retained in the model included an indicator variable for the type of bleeding and the following covariates (multivariate P < 0.05 included):

aAge, sex, race, diabetes, hypertension, prior angina, prior MI, prior PCI, heart rate, systolic blood pressure, Killip class, MI location, and time from symptom onset to fibrinolysis. CABG, coronary artery bypass graft surgery; CI, confidence interval; HR, hazard ratio; IABP, intra-aortic balloon counterpulsation; ICH, intracranial haemorrhage; PCI, percutaneous coronary intervention.

bTreatment with UFH vs. enoxaparin at randomization.

cCoronary angiography, PCI, IABP, and pacemaker.

dCardiogenic shock, ischaemic stroke, and severe heart failure.

eThe HRs shown for non-ICH major and for minor bleeding are from Days 5 to 30 due to non-proportionality of the hazards over the first 30 days. The corresponding HRs, 95% CI, and P-values for Days 0–5 for non-ICH major bleeding were 0.7 (0.3–1.5), P = 0.40 for the unadjusted and first three adjusted models, and 0.8 (0.4–1.6), P = 0.47 for the fully adjusted model. For minor bleeding, the HRs, 95% CI, and P-values at Day 3 (midpoint of the 5-day interval was selected due to continued non-proportionality of the HR during Days 0–5) were 0.5 (0.3–1.1), P = 0.07 for the unadjusted model, and 0.3 (0.2–0.6), 0.3 (0.1–0.6), 0.3 (0.2–0.6), and 0.2 (0.1–0.4), P ≤ 0.001 for each of the sequentially adjusted models.

Figure 1

The 30-day all-cause mortality in patients with and without bleeding (A) and by the type of major bleeding (B). Bleeding events during the index hospitalization were used to classify patients as major, minor, or no bleeding (A) and as intracranial haemorrhage, non-intracranial haemorrhage, and no bleeding (B). See text and Table 2 for details regarding the adjusted analyses. ICH, intracranial haemorrhage.

Figure 1

The 30-day all-cause mortality in patients with and without bleeding (A) and by the type of major bleeding (B). Bleeding events during the index hospitalization were used to classify patients as major, minor, or no bleeding (A) and as intracranial haemorrhage, non-intracranial haemorrhage, and no bleeding (B). See text and Table 2 for details regarding the adjusted analyses. ICH, intracranial haemorrhage.

In a Cox model, the three most important variables contributing to 30-day mortality were development of cardiogenic shock prior to discharge, increased age, and ICH (Table 3). The per cent of information content that each variable contributed to the model estimated by the likelihood ratios were 61, 17 (per decade), and 7.8%, for these three variables, respectively. Neither non-ICH major bleeding nor minor bleeding contributed significantly to the model (contributions 0.22 and 0.03%, respectively).

Table 3

Major contributors to 30-day mortality

Variable % Model contribution 
Cardiogenic shock 61 
Age (per decade older) 17 
Intracranial haemorrhage 7.8 
Heart rate (per 10 b.p.m. increase) 2.9 
No coronary angiogram pre-discharge 2.2 
Severe heart failure 2.0 
Female gender 1.5 
Systolic blood pressure (per 10 mmHg decrease) 1.2 
Variable % Model contribution 
Cardiogenic shock 61 
Age (per decade older) 17 
Intracranial haemorrhage 7.8 
Heart rate (per 10 b.p.m. increase) 2.9 
No coronary angiogram pre-discharge 2.2 
Severe heart failure 2.0 
Female gender 1.5 
Systolic blood pressure (per 10 mmHg decrease) 1.2 

Ten additional independent covariates had small (<1%) individual contributions to the model. The model contributions for non-ICH major bleeding and minor bleeding were 0.22 and 0.03%, respectively.

These findings were supported by a comparison of the causes of death among patients who died by Day 30, stratified by the type of bleeding during the index hospitalization. Intracranial haemorrhage was listed as the cause in 80/93 of deaths at 30 days among patients with an ICH prior to discharge. However, among the 23 patients with a non-ICH major bleed who died by Day 30, the most common cause of death was congestive heart failure/shock (n = 9), and only 6 deaths were attributed to bleeding. In the 38 patients with a minor bleed who died by Day 30, non-haemorrhagic cardiovascular causes accounted for 32 deaths, whereas non-ICH bleeding was classified as the cause of death in only 3 patients.

Mortality at 1 year

Mortality at 1 year was 42.9% among patients with major bleeding and 9.9% among those without bleeding (P < 0.001). However, the mortality rate was most marked among those with ICH (71.8%), whereas mortality rates were 18.1 and 15.8% among patients with non-ICH major and minor bleeding, respectively (unadjusted P-values vs. no bleeding ≤0.001 for each). After multivariate adjustment, only ICH remained independently associated with increased mortality at 1 year (HR 8.7, P < 0.001), contributing 5.9% to the fully adjusted model. Cardiogenic shock during the index hospitalization (HR 10.6, contribution 52%) and age (HR 1.6 per decade, contribution 23%) were the most important correlates of 1-year mortality. Non-ICH major bleeding and minor bleeding each contributed ≤0.2% to the model.

Landmark analyses of deaths between Days 31 and 365

Among the patients who survived to Day 30, there were 192 patients (1.0%) with a TIMI major bleeding event [including 49 (0.26%) with ICH and 143 (0.76%) with non-ICH major bleeding] and 348 (1.8%) with a minor bleeding event during the index hospitalization. The 18 318 patients (97.1%) with no major bleeding during the index hospitalization and who survived to Day 30 served as the comparator group. Mortality rates between Days 31 and 365 among survivors to Day 30 were 8.5% among patients with major bleeding, 6.6% in those with minor bleeding, and 3.5% among those without bleeding (P < 0.001 for major bleeding and P = 0.002 for minor bleeding compared with no bleeding) (Figure 2A). However, the excess in mortality between Days 31 and 365 among patients with major bleeding during the index hospitalization compared with those with no bleeding was driven by a high mortality rate between Days 30 and 365 among patients with an initial ICH who had survived the first 30 days (18.5%, P < 0.001 compared with patients with no bleeding). There were no significant differences in mortality at 31–365 days in patients with an in-hospital non-ICH major bleed (5.0%) compared with those with no bleeding (3.5%, unadjusted P = 0.36) (Figure 2B).

Figure 2

Landmark analyses of all-cause mortality between Days 31 and 365 among patients who survived to Day 30 by the presence (A) and type (B) of bleeding during the index hospitalization. See text and Table 3 for details regarding the adjusted analyses.

Figure 2

Landmark analyses of all-cause mortality between Days 31 and 365 among patients who survived to Day 30 by the presence (A) and type (B) of bleeding during the index hospitalization. See text and Table 3 for details regarding the adjusted analyses.

In analyses adjusted for baseline characteristics (Table 4), the HR for ICH remained statistically significant (4.6, P < 0.001), but was attenuated compared with the unadjusted analysis (HR 5.8). Moreover, after adjustment for baseline characteristics, there were no significant associations between either non-ICH major bleeding or minor bleeding and mortality between Days 31 and 365 (Table 4). The most important baseline characteristics that confounded the relationship between non-ICH bleeding and later mortality included age (HR 1.6 per decade), heart rate (HR 1.2 per 10 b.p.m. increase), Killip class > I (HR 1.5), time from symptom onset to fibrinolytic (HR 1.1 per additional hour delay), systolic blood pressure (HR 1.1 per each 10 mmHg lower), prior angina (HR 1.4), prior MI (HR 1.5), and diabetes (HR 1.4) (P ≤ 0.001 for each). In the fully adjusted Cox model, the variables that most strongly contributed to mortality between Days 31 and 365 were age (HR 1.6 per each decade, contribution 43%) and heart rate (HR 1.2 per 10 b.p.m. increase, contribution 18%).

Table 4

Risk of mortality between Days 31 and 365 among patients who had a bleeding event during the index hospitalization and survived to Day 30

Type of bleeding HR (95% CI) P-value 
All major bleeding   
 Unadjusted 2.5 (1.5, 4.0) <0.001 
 Adjusted for baseline characteristicsa 2.1 (1.2, 3.6) 0.006 

 
Intracranial haemorrhage   
 Unadjusted 5.8 (3.0, 11.3) <0.001 
 Adjusted for baseline characteristicsa 4.6 (2.4, 8.8) <0.001 

 
Non-ICH major bleeding   
 Unadjusted 1.4 (0.7, 3.0) 0.36 
 Adjusted for baseline characteristicsa 1.1 (0.4, 2.6) 0.89 

 
Minor bleeding   
 Unadjusted 1.9 (1.3, 2.9) 0.002 
 Adjusted for baseline characteristicsa 1.3 (0.9, 2.0) 0.21 
Type of bleeding HR (95% CI) P-value 
All major bleeding   
 Unadjusted 2.5 (1.5, 4.0) <0.001 
 Adjusted for baseline characteristicsa 2.1 (1.2, 3.6) 0.006 

 
Intracranial haemorrhage   
 Unadjusted 5.8 (3.0, 11.3) <0.001 
 Adjusted for baseline characteristicsa 4.6 (2.4, 8.8) <0.001 

 
Non-ICH major bleeding   
 Unadjusted 1.4 (0.7, 3.0) 0.36 
 Adjusted for baseline characteristicsa 1.1 (0.4, 2.6) 0.89 

 
Minor bleeding   
 Unadjusted 1.9 (1.3, 2.9) 0.002 
 Adjusted for baseline characteristicsa 1.3 (0.9, 2.0) 0.21 

aSee Table 1 footnote for a list of covariates. ICH, intracranial haemorrhage.

Relationship of recurrent myocardial infarction and non-intracranial haemorrhage bleeding with mortality

Recurrent MI (HR 2.8 [2.0–4.0]), non-ICH major bleeding (HR 3.0 [1.8–5.0], and minor bleeding (HR 1.9 (1.3–2.8)) were each independently correlated with mortality between Days 5 and 30 adjusted for baseline patient characteristics, medications, and pre-discharge invasive procedures (P ≤ 0.001). Re-infarction (HR 1.9 (1.3–2.7), P < 0.001), but neither non-ICH major bleeding (HR 1.0 (0.6–1.7), P = 0.99) nor minor bleeding (HR 1.2 (0.8–1.7), P = 0.46), remained independently associated with mortality between Days 5 and 30 after adjustment for cardiovascular events, including cardiogenic shock. In the landmark analysis exploring death between 31 and 365 days among survivors to Day 30, the HRs for re-infarction were 1.5 (P = 0.087) with and 1.5 (P = 0.055) without cardiogenic shock in the model.

Discussion

The major findings of these analyses of the relationship between major bleeding and mortality during the index hospitalization of 20 323 patients with STEMI enrolled in a worldwide fibrinolytic trial include the following. Our findings underscore the poor prognosis associated with ICH, a complication of fibrinolytic therapy in patients with STEMI that persists despite the development of drugs with increased fibrin specificity and a reduction in the dosing of concomitant anticoagulant.18 Our observations with regard to non-ICH bleeding are consistent with the report by Spencer et al.6 which demonstrated that patient comorbidity accounts for an important part of the excess mortality in patients with major bleeding following MI. In fact, in the landmark analysis, the unadjusted mortality rates from Days 31 to 365 among survivors to Day 30 shows a higher mortality for patients with an in-hospital minor bleeding (6.6%) compared with non-ICH major bleeding (5.0%), an observation that is no longer present once adjustment is made for differences in baseline characteristics. Our analyses suggest that early deaths following non-ICH major bleeding are related to antecedent cardiovascular events (particularly cardiogenic shock). Early deaths following minor bleeding and deaths after 30 days in patients with either non-ICH major or minor bleeding are largely explained by the higher risk profile of these patients as determined by their baseline characteristics.

  • (1) ICH was independently associated with mortality at 30 days, 1 year, and among survivors to Day 30, between Days 31 and 365.

  • (2) Neither non-ICH TIMI major bleeding nor TIMI minor bleeding was independently associated with increased mortality at 30 days, at 1 year, nor from Days 31 to 365 among survivors to Day 30. The major confounder of the relationship between non-ICH bleeding and short-term mortality was the development of cardiogenic shock preceding the bleeding event.

  • (3) The most important factors associated with death at 30 days were the development of cardiogenic shock during the index hospitalization, increased age, and ICH.

  • (4) Patient baseline characteristics, such as age, history of MI and angina, initial haemodynamics, Killip class, and delay to fibrinolytic, identify patients who are more likely to die between Days 31 and 365. These confounders explain the apparent associations observed between non-ICH bleeding (major, minor, and their combination) and mortality between Days 31 and 365.

Our results differ from some prior analyses3,5,7 of the relationship between bleeding and death due to important differences in the analytic techniques used, study populations (STEMI vs. non-STE ACS), and treatments administered (fibrinolytic vs. primary PCI or no reperfusion therapy). To minimize the biases and confounding that can arise in such analyses, we evaluated ICH and non-ICH bleeding separately, included invasive procedures, in-hospital cardiovascular events, and the timing of these events in our adjusted models, and performed landmark analyses to provide greater insight into the relationship between in-hospital bleeding and short- and long-term mortalities.

A number of studies, particularly from contemporary trials of newer anti-thrombotic agents in patients with non-STE ACS, have advanced the hypothesis that less bleeding translates into lower mortality.3,5,7 This hypothesis is attractive since reducing bleeding is ispso facto a desirable outcome. Thus, a reduction in bleeding, if it truly were causally related to a reduction in mortality, would represent an important additional benefit to patients. These notions have led to new recommendations in the European Society of Cardiology Clinical Practice Guidelines in Patients with Non-STE-ACS highlighting the importance of assessing the risk of bleeding and initiating steps (and therapies) to minimize bleeding.11 In addition, bleeding often results in blood transfusion, which itself has been associated with poorer outcomes19,20 in some, but not all,21,22 analyses. Therefore, reducing bleeding is an important goal in the development of future therapies.

The relationship between bleeding and ischaemic events is not straightforward.23 First, bleeding complications represent a post-randomization event, and by nature, any comparison of patients with vs. without bleeding (the so-called ‘endpoint–endpoint’ analysis) runs a substantial risk of confounding due to imbalance between the groups (i.e. one cannot randomize patients to ‘bleeding’ vs. ‘no bleeding’).24 Differences in baseline characteristics, therapies (before and after randomization), and timing of events among patients with and without a bleeding event may be marked.6 Second, despite attempts to adjust for various differences between groups of patients with vs. without bleeding, retrospective analyses are not able to capture all relevant covariates, including many that may have led physicians to recommend for or against specific procedures and treatments which could have affected outcomes, including bleeding. Third, determining the causal relationship between post-randomization events, such as bleeding and mortality, may be confounded by other events that may occur before (e.g. cardiogenic shock), coincident (e.g. surgery), or after (e.g. transfusion) the bleeding event. Modelling the effects of individual events that are part of a complex, inter-related web is difficult to accomplish using standard multivariate techniques. Furthermore, assigning the cause of death to a single (or the most proximate) event may not adequately describe the subtleties involved. For example, bleeding may cause ischaemia due to a mismatch in oxygen supply demand, which may then lead to a fatal arrhythmia. Additional research and alternative methods to assess the possible causal relationship between bleeding, concurrent events, and mortality are needed. We have recently adopted efforts in our adjudication of death in ongoing clinical trials to provide greater granularity regarding the presence and possible causal relationship between bleeding and death by asking the clinical endpoint committee to determine, in each patient who dies and had a bleeding event, whether the bleeding was fatal, contributed to the cause of death, or was unrelated to death.

Limitations

Our findings represent a post-hoc analysis of patients classified by an event (bleeding) that occurred after randomization through discharge or Day 8 in a clinical trial of patients with STEMI treated with fibrinolytic and antithrombin. The ExTRACT-TIMI 25 trial was not prospectively designed to analyse the relationship between bleeding and mortality, and the frequency of some clinical events were low, thus limiting our power to detect differences between subgroups of patients. Thus, our observations are exploratory in nature, and as such should be considered to be hypothesis-generating. We utilized multivariate techniques to adjust for potential confounders, but cannot exclude the possibility that additional unmeasured confounders influenced the outcomes. Strict enrollment criteria were used to identify patients eligible for enrollment in ExTRACT-TIMI 25, patient management was guided by a study protocol, and the TIMI criteria were used to classify bleeding. We and others have implemented efforts to include multiple bleeding scales in ongoing trials to facilitate future analyses of bleeding within and across studies. All of our patients were treated with a fibrinolytic and either enoxaparin or UFH during the index hospitalization for a maximum of 8 days, and bleeding data were not systematically reported beyond discharge/Day 8. Therefore, the findings described above may not apply to all patients with STEMI or those with NSTE-ACS in clinical practice.

Conclusions

Intracranial haemorrhage, but not non-ICH TIMI major or minor bleeding, is strongly associated with short (30-day) and longer-term (1-year) mortality in patients undergoing fibrinolysis for STEMI. In addition to ICH, development of cardiogenic shock during the index hospitalization and advanced age are important contributors to mortality both at 30 days and 1 year. Thus, efforts to reduce mortality after STEMI should focus on prevention of pump failure and intracranial bleeding.

Non-ICH bleeding following fibrinolytic therapy is associated with excess mortality, and further efforts to reduce these bleeding events are desirable. However, the relationship between non-ICH bleeding and mortality is complex and highly confounded. Neither non-ICH TIMI major nor minor bleeding was independently associated with an increase in either short- or longer-term mortality once patient baseline characteristics, treatments, and concurrent cardiovascular events were taken into consideration.

Funding

The ExTRACT-TIMI 25 trial was supported by a research grant from Sanofi-Aventis to the TIMI Study Group. However, no funding or other support was provided for this manuscript. All co-authors are/were members of the TIMI Study Group.

Conflict of interest: D.A.M., E.M.A., C.M.G. and E.B. report having received research grant support from Sanofi-Aventis. R.P.G., D.A.M., E.M.A., C.M.G. and E.B. have received lecture fees and/or have served on paid advisory boards for Sanofi-Aventis and Bristol-Myers Squibb. C.H.M. and S.A.M. have received research grant support from Sanofi-Aventis. R.R.G. and S.M. report no conflicts of interest relevant to this article.

References

1
Antman
EM
Morrow
DA
McCabe
CH
Murphy
SA
Ruda
M
Sadowski
Z
Budaj
A
Lopez-Sendon
JL
Guneri
S
Jiang
F
White
HD
Fox
KA
Braunwald
E
Enoxaparin versus unfractionated heparin with fibrinolysis for ST-elevation myocardial infarction
N Engl J Med
 , 
2006
, vol. 
354
 (pg. 
1477
-
1488
)
[PubMed]
2
Morrow
DA
Antman
EM
Fox
KAA
White
HD
Giugliano
R
Murphy
SA
McCabe
CH
One-year outcomes after a strategy using enoxaparin vs. unfractionated heparin in patients undergoing fibrinolysis for ST-segment elevation myocardial infarction: 1-year results of the ExTRACT-TIMI 25 Trial
Eur Heart J
 , 
2010
 
. Published online ahead of print 17 April 2010
[PubMed]
3
Budaj
A
Eikelboom
JW
Mehta
SR
Afzal
R
Chrolavicius
S
Bassand
JP
Fox
KA
Wallentin
L
Peters
RJ
Granger
CB
Joyner
CD
Yusuf
S
Improving clinical outcomes by reducing bleeding in patients with non-ST-elevation acute coronary syndromes
Eur Heart J
 , 
2009
, vol. 
30
 (pg. 
655
-
661
)
4
Eikelboom
JW
Mehta
SR
Anand
SS
Xie
C
Fox
KA
Yusuf
S
Adverse impact of bleeding on prognosis in patients with acute coronary syndromes
Circulation
 , 
2006
, vol. 
114
 (pg. 
774
-
782
)
5
Manoukian
SV
Feit
F
Mehran
R
Voeltz
MD
Ebrahimi
R
Hamon
M
Dangas
GD
Lincoff
AM
White
HD
Moses
JW
King
SB
3rd
Ohman
EM
Stone
GW
Impact of major bleeding on 30-day mortality and clinical outcomes in patients with acute coronary syndromes: an analysis from the ACUITY Trial
J Am Coll Cardiol
 , 
2007
, vol. 
49
 (pg. 
1362
-
1368
)
6
Spencer
FA
Moscucci
M
Granger
CB
Gore
JM
Goldberg
RJ
Steg
PG
Goodman
SG
Budaj
A
FitzGerald
G
Fox
KA
Does comorbidity account for the excess mortality in patients with major bleeding in acute myocardial infarction?
Circulation
 , 
2007
, vol. 
116
 (pg. 
2793
-
2801
)
7
Stone
GW
Witzenbichler
B
Guagliumi
G
Peruga
JZ
Brodie
BR
Dudek
D
Kornowski
R
Hartmann
F
Gersh
BJ
Pocock
SJ
Dangas
G
Wong
SC
Kirtane
AJ
Parise
H
Mehran
R
Bivalirudin during primary PCI in acute myocardial infarction
N Engl J Med
 , 
2008
, vol. 
358
 (pg. 
2218
-
2230
)
8
Crowther
MA
Warkentin
TE
Bleeding risk and the management of bleeding complications in patients undergoing anticoagulant therapy: focus on new anticoagulant agents
Blood
 , 
2008
, vol. 
111
 (pg. 
4871
-
4879
)
9
Jolly
SS
Amlani
S
Hamon
M
Yusuf
S
Mehta
SR
Radial versus femoral access for coronary angiography or intervention and the impact on major bleeding and ischemic events: a systematic review and meta-analysis of randomized trials
Am Heart J
 , 
2009
, vol. 
157
 (pg. 
132
-
140
)
10
Madigan
JB
Ratnam
LA
Belli
AM
Arterial closure devices. A review
J Cardiovasc Surg (Torino)
 , 
2007
, vol. 
48
 (pg. 
607
-
624
)
[PubMed]
11
Bassand
JP
Bleeding and transfusion in acute coronary syndromes: a shift in the paradigm
Heart
 , 
2008
, vol. 
94
 (pg. 
661
-
666
)
12
Berger
PB
Manoukian
SV
Bleeding is bad … . Isn't it?
Circulation
 , 
2007
, vol. 
116
 (pg. 
2776
-
2778
)
13
Cohen
M
Predictors of bleeding risk and long-term mortality in patients with acute coronary syndromes
Curr Med Res Opin
 , 
2005
, vol. 
21
 (pg. 
439
-
445
)
14
Segev
A
Strauss
BH
Tan
M
Constance
C
Langer
A
Goodman
SG
Predictors and 1-year outcome of major bleeding in patients with non-ST-elevation acute coronary syndromes: insights from the Canadian Acute Coronary Syndrome Registries
Am Heart J
 , 
2005
, vol. 
150
 (pg. 
690
-
694
)
15
Antman
EM
Morrow
DA
McCabe
CH
Jiang
F
White
HD
Fox
KA
Sharma
D
Chew
P
Braunwald
E
Enoxaparin versus unfractionated heparin as antithrombin therapy in patients receiving fibrinolysis for ST-elevation myocardial infarction. Design and rationale for the Enoxaparin and Thrombolysis Reperfusion for Acute Myocardial Infarction Treatment-Thrombolysis In Myocardial Infarction study 25 (ExTRACT-TIMI 25)
Am Heart J
 , 
2005
, vol. 
149
 (pg. 
217
-
226
)
16
Schoenfeld
D
Partial residuals for the proportional hazards regression model
Biometrika
 , 
1982
, vol. 
69
 (pg. 
239
-
241
)
17
Morrow
DA
Antman
EM
Charlesworth
A
Cairns
R
Murphy
SA
deLemos
JA
Giugliano
RP
McCabe
CH
Braunwald
E
The TIMI risk score for ST elevation myocardial infarction: A convenient, bedside, clinical score for risk assessment at presentation
Circulation
 , 
2000
, vol. 
102
 (pg. 
2031
-
2037
)
[PubMed]
18
Van de Werf
F
Barron
HV
Armstrong
PW
Granger
CB
Berioli
S
Barbash
G
Pehrsson
K
Verheugt
FW
Meyer
J
Betriu
A
Califf
RM
Li
X
Fox
NL
Incidence and predictors of bleeding events after fibrinolytic therapy with fibrin-specific agents: a comparison of TNK-tPA and rt-PA
Eur Heart J
 , 
2001
, vol. 
22
 (pg. 
2253
-
2261
)
19
Rao
SV
Jollis
JG
Harrington
RA
Granger
CB
Newby
LK
Armstrong
PW
Moliterno
DJ
Lindblad
L
Pieper
K
Topol
EJ
Stamler
JS
Califf
RM
Relationship of blood transfusion and clinical outcomes in patients with acute coronary syndromes
JAMA
 , 
2004
, vol. 
292
 (pg. 
1555
-
1562
)
20
Yang
X
Alexander
KP
Chen
AY
Roe
MT
Brindis
RG
Rao
SV
Gibler
WB
Ohman
EM
Peterson
ED
The implications of blood transfusions for patients with non-ST-segment elevation acute coronary syndromes: results from the CRUSADE National Quality Improvement Initiative
J Am Coll Cardiol
 , 
2005
, vol. 
46
 (pg. 
1490
-
1495
)
21
Alexander
KP
Chen
AY
Wang
TY
Rao
SV
Newby
LK
LaPointe
NM
Ohman
EM
Roe
MT
Boden
WE
Harrington
RA
Peterson
ED
Transfusion practice and outcomes in non-ST-segment elevation acute coronary syndromes
Am Heart J
 , 
2008
, vol. 
155
 (pg. 
1047
-
1053
)
22
Wu
WC
Rathore
SS
Wang
Y
Radford
MJ
Krumholz
HM
Blood transfusion in elderly patients with acute myocardial infarction
N Engl J Med
 , 
2001
, vol. 
345
 (pg. 
1230
-
1236
)
23
Yan
AT
Yan
RT
Huynh
T
DeYoung
P
Weeks
A
Fitchett
DH
Langer
A
Goodman
SG
Bleeding and outcome in acute coronary syndrome: insights from continuous electrocardiogram monitoring in the Integrilin and Enoxaparin Randomized Assessment of Acute Coronary Syndrome Treatment (INTERACT) trial
Am Heart J
 , 
2008
, vol. 
156
 (pg. 
769
-
775
)
24
Pieper
KS
Tsiatis
AA
Davidian
M
Hasselblad
V
Kleiman
NS
Boersma
E
Chang
WC
Griffin
J
Armstrong
PW
Califf
RM
Harrington
RA
Differential treatment benefit of platelet glycoprotein IIb/IIIa inhibition with percutaneous coronary intervention versus medical therapy for acute coronary syndromes: exploration of methods
Circulation
 , 
2004
, vol. 
109
 (pg. 
641
-
646
)

Supplementary data

Comments

0 Comments