Incidence and Predictors of Bleeding After Contemporary Thrombolytic Therapy for Myocardial Infarction
Background Although the benefit of thrombolytic therapy in reducing mortality in acute myocardial infarction is well established, the types of bleeding and risk factors for bleeding are less well described in large trials.
Methods and Results We analyzed the baseline characteristics, outcomes, and incidence of bleeding by location, severity, and treatment assignment among 41 021 patients in the GUSTO-I trial of thrombolysis for acute myocardial infarction. Of the 40 903 patients for whom there were complete data, 1.2% suffered severe bleeding and 11.4% experienced moderate hemorrhage at a variety of sites. The most common sources of bleeding were procedure related. The thrombolytic regimen was strongly related to the incidence of bleeding; comparatively more bleeding was seen with the therapies of streptokinase plus intravenous heparin and the streptokinase and tissue plasminogen activator plus intravenous heparin combination. In multivariate analysis, the four most powerful independent predictors of hemorrhage were older age, lighter body weight, female sex, and African ancestry; they remained the most important predictors of bleeding when multivariate analysis was performed on patients who did not undergo invasive procedures. The presence of serious hemorrhage was associated with other undesirable outcomes (recurrent events, left ventricular dysfunction, arrhythmia, or stroke).
Conclusions Important predictors of bleeding in this population are increased age, lighter weight, female sex, African ancestry, and experiencing invasive procedures. Other nonhemorrhagic adverse clinical outcomes were associated with moderate and severe bleeding, which was in turn associated with increased length of hospital stay and mortality at 30 days.
Thrombolytic therapy reduces mortality1 across the spectrum of patients with suspected acute myocardial infarction with ST-segment elevation or bundle-branch block, but some of the benefit of thrombolytic therapy is offset by the hazards of hemorrhagic stroke and noncerebral bleeding. Although both clinical characteristics and hemostatic variables have been studied in an attempt to define pertinent risk factors for hemorrhage,2 3 4 5 6 few studies have focused on the detailed relation of clinical characteristics to hemorrhagic risk,4 5 nor has a practically useful risk algorithm been developed. The Global Utilization of Streptokinase and t-PA for Occluded Coronary Arteries (GUSTO-I) trial7 presented an opportunity to analyze the location and severity of hemorrhage and associated outcomes after treatment with thrombolytic therapy in a large patient population.
Patients and Treatment
Details of the study design, inclusion and exclusion criteria, trial medications, and guidelines for patient management have been described in detail elsewhere.7 Briefly, patients presenting to the hospital within 6 hours of the onset of symptoms, with chest pain lasting ≥20 minutes and accompanied by ECG signs of ≥0.1 mV of ST-segment elevation in two or more leads, were eligible for enrollment. Exclusion criteria included active bleeding, recent trauma or major surgery, history of stroke, noncompressible vascular punctures, and previous treatment with streptokinase or anistreplase.
Patients were randomly allocated to one of four thrombolytic strategies: streptokinase 1.5 million U over 60 minutes plus subcutaneous heparin 12 500 IU twice daily beginning 4 hours after initiation of thrombolytic therapy; streptokinase 1.5 million U over 60 minutes plus intravenous heparin bolus of 5000 U followed by 1000 U/h, with dose adjustment to maintain an activated partial thromboplastin time (aPTT) of 60 to 85 seconds; accelerated tissue plasminogen activator (TPA) bolus of 15 mg immediately followed by infusion of 0.75 mg/kg (up to 50 mg) over 30 minutes and then 0.5 mg/kg (up to 35 mg) over the next 60 minutes, accompanied by the same intravenous heparin regimen; or the combination of intravenous TPA (1.0 mg/kg over 60 minutes, not to exceed 90 mg, with 10% given as a bolus) plus streptokinase (1.0 million U for 60 minutes) given concurrently but through separate catheters, followed by the same intravenous heparin regimen.
Bleeding was defined as severe if it caused substantial hemodynamic compromise that required intervention or treatment and moderate if transfusion was required but did not lead to hemodynamic compromise requiring intervention. Intracranial hemorrhages were by definition severe bleeding events but were not included in this analysis. The description of the intracranial bleeding observed in GUSTO-I has been reported previously.8
Baseline characteristics and clinical outcomes that are discrete factors are described in terms of frequencies and percentages of patients with the characteristic for categorical variables; continuous variables are described using the median and 25th and 75th percentiles. Odds ratios and 95% CIs were calculated by use of standard methods.
Logistic regression modeling techniques were used to evaluate the univariate relationship between baseline or procedural characteristics and the likelihood of having moderate or severe bleeding. Assumptions of linearity were tested by use of cubic spline functions as well as a variety of graphic techniques. These techniques were used to most appropriately describe any nonlinear relationships that existed, as previously described.9
Three multivariate models of moderate or severe bleeding versus mild or no bleeding were created. The first two models were developed in the entire GUSTO population. One model determined the relationship of the baseline clinical factors in the prediction of a bleeding complication. The other determined the relationship of both baseline clinical factors and subsequent predischarge procedures (pacemaker implantation, Swan-Ganz catheter placement, ventilator use, coronary bypass surgery, intra-aortic balloon pump insertion, cardioversion/defibrillation, angiography, or percutaneous coronary angioplasty) in the prediction of a bleeding complication. Recognizing that the incidence of bleeding was greatly influenced by the use of procedures, a third model was developed to determine the effect of the baseline clinical factors on bleeding without the procedural bias. This model was similar to the first but was applied only in the subset of patients who had no in-hospital procedures.
Logistic regression modeling techniques were used to develop each of the three multivariate models. A stepwise variable-reduction technique was used to find the combination of variables that contributed independent information. Once final models were developed, bootstrapping was used for internal validation. The quality of the final models based on the original as well as the bootstrapped samples is described with the use of the concordance index. The concordance index is a description of the discriminant power of the model to reliably predict an outcome.
On the basis of the coefficients of the clinical model of bleeding, a probability chart was developed (Table 9⇓). Each variable in the model received a certain score based on the value of that variable. The total points were then easily transformed into predictive values.
Bleeding complications occurred at a variety of sites (Table 1⇓). The most common sources of moderate and severe bleeding were procedure related; almost 4% of the overall population experienced hemorrhage related to coronary artery bypass grafting, and 2.0% had hemorrhage at the groin site. The most common site of spontaneous bleeding was the gastrointestinal tract. More than 5% of patients had diminutions in the hematocrit identified as representing moderate or severe bleeding, with no obvious source of bleeding reported.
The thrombolytic regimen was strongly related to the incidence of bleeding (Table 2⇓). Less moderate or severe hemorrhage was seen with streptokinase plus subcutaneous heparin treatment compared with streptokinase plus intravenous heparin (P<.0001) and the combination therapies (P<.0007). Less moderate or severe hemorrhage was seen with TPA plus intravenous heparin compared with either streptokinase plus intravenous heparin (P<.0001) or the combination therapies (P<.0001). The bleeding incidences were comparable for streptokinase plus subcutaneous heparin and TPA plus intravenous heparin, and for streptokinase plus intravenous heparin and the combination therapies.
Table 3⇓ illustrates the distribution of the key baseline characteristics for patients who had serious hemorrhage. Bleeding occurred more commonly in patients of older age, female sex, lighter body weight, shorter stature, and African ancestry and in patients in the United States. Less bleeding was seen in current smokers and in patients with a history of smoking.
Regardless of thrombolytic assignment, higher bleeding rates were associated with higher aPTTs (Fig 1⇓, top and bottom). The relationship of aPTT with bleeding risk, however, was different for different thrombolytic agents. That is, for a given aPTT, the risk of bleeding was higher for streptokinase-assigned patients than for TPA-assigned patients.
In multivariate analysis, the three most powerful independent predictors of hemorrhage were older age, lighter body weight, and female sex (Table 4⇓; Fig 2⇓). Several variables added significantly to the model but exhibited a different effect in US versus non-US patients. Worsening in Killip class was related to a significant increase in the incidence of bleeding complications for US but not for non-US patients. For patients randomized to streptokinase plus subcutaneous heparin who were in Killip class I, there was a 76% higher likelihood of having a serious hemorrhage for US than for non-US patients. With the streptokinase plus subcutaneous heparin as the comparative group in the non-US patients, those who received streptokinase plus intravenous heparin or the combination (streptokinase plus TPA plus heparin) treatments were associated with 63% and 66% more bleeding, respectively. Hemorrhage in US patients was much less with streptokinase plus intravenous heparin and not significantly different for combination therapy. US patients assigned to TPA had significantly less bleeding than US patients given streptokinase plus subcutaneous heparin; non-US patients assigned to TPA had more bleeding.A higher incidence of bleeding events was strongly related to invasive procedures (Table 5⇓). Fifty-eight percent of patients (73% US, 27% non-US) had at least one procedure (Swan-Ganz catheter placement, insertion of pacemaker or balloon pump, angiography, angioplasty, or bypass surgery). Moderate or severe bleeding occurred in 50% of patients having coronary artery bypass grafting or intra-aortic balloon pump insertion, five times the rate in patients without these procedures. The only procedure not associated with more bleeding was coronary angioplasty. Moderate or severe bleeding was found in only 6% of patients who did not undergo any of these procedures.
Table 6⇓ summarizes a second multivariate model to predict bleeding, which adjusts for the baseline characteristics of the patients and highlights the effect of the procedures on bleeding according to country of origin and treatment assignment. The prognostically important baseline characteristics remained fairly consistent with those seen in the first model, except that previous angina and infarct location were no longer important contributors, but previous bypass surgery was. Bypass surgery was the procedure most strongly associated with bleeding risk, but this risk was greater for US patients than for non-US patients, whereas the opposite was true for angiography and Swan-Ganz catheter placement. There was a significantly increased risk of bleeding in the use of an intra-aortic balloon pump with drug regimens other than combination therapy. The implantation of a pacemaker was related to an increased risk of bleeding regardless of which thrombolytic therapy was given or country of origin. Angioplasty was significantly related to an increased risk of bleeding for patients not randomized to TPA.
Table 7⇓ summarizes the effect of the baseline characteristics in the 42% of patients who received no procedure during their hospitalization. Age, weight, and sex remained the most important predictors of bleeding. Treatment assignment was the next most important predictor. Race, pulse, and Killip class remained important but had strong differential effects according to treatments received. Of the other baseline characteristics, the risk factors of hypertension and current smoking, the presentation variables of diastolic blood pressure and infarct location, and a history of previous angina were no longer significant independent predictors. Of most interest is the absence of an effect from US versus non-US status. Once differences in the use of procedures were accounted for, patients in the US were no longer at significantly increased risk of bleeding compared with other patients. There was no evidence of a lower threshold for transfusion (and therefore for identification of hemorrhage as moderate) in US patients: the median nadir hematocrit among patients transfused was 25.0% in US versus 26.9% in non-US patients.
The presence of serious hemorrhage was associated with other undesirable outcomes (recurrent events, left ventricular dysfunction, arrhythmia, or stroke) (Table 8⇓). Twenty percent to 30% of patients with any one of these adverse events had moderate or severe bleeding. Length of hospitalization for patients with bleeding complications was longer (P=.0001). The predicted value for the likelihood of experiencing a moderate or severe bleeding complication in this patient population can be obtained from the probability chart we developed (Table 9⇓). Because the decision to use invasive therapy was made after treatment assignment, a separate probability chart based on the model in patients treated without catheterization was developed to calculate the predicted value for their likelihood of experiencing a moderate or severe bleeding complication (Table 10⇓).
As expected from prior studies, older age,4 10 lower body weight,10 and female sex10 11 13 were the three most powerful independent predictors of hemorrhage that could be determined at the time of presentation with acute myocardial infarction. Coronary artery bypass surgery and invasive procedures were strongly related to bleeding. Regardless of whether use of procedures was included in the multivariate analyses, age, weight, and sex (in order of importance) remained the most powerful baseline predictors of bleeding. An additional finding that has not been well established from prior studies is the higher likelihood of bleeding in patients of African descent, especially when treated with TPA.
Bleeding After Thrombolysis in Prior Trials
The Fibrinolytic Therapy Trialists' (FTT) Collaborative Group evaluated nine trials including 58 600 patients randomized to thrombolytic therapy versus control therapies in suspected acute myocardial infarction.1 An excess of 0.7% in major noncerebral bleeding (life-threatening or requiring blood transfusion) was detected, with a rate of 1.1% in those assigned to thrombolytic therapy versus 0.4% in control subjects (P<.0001). The combined data from GISSI-214 and ISIS-315 included 48 294 patients randomized to standard-dose TPA or streptokinase and 62 067 patients randomized to aspirin plus subcutaneous heparin or aspirin. Major bleeding rates were higher for streptokinase than for TPA (0.9% versus 0.7%; P<.05) and for subcutaneous heparin than for no heparin (1.0% versus 0.7%; P<.00001).
Relationship of Thrombolytic and Heparin Regimens to Bleeding in GUSTO-I
As in previous large trials comparing streptokinase and TPA, the risk of noncerebral bleeding was greater after streptokinase than after accelerated TPA. This may be due in part to the greater fibrin specificity of TPA, with less resultant depletion of circulating fibrinogen.3 5 The differential effect of streptokinase on the coagulation system is reflected in higher aPTTs after streptokinase than after accelerated TPA, which may be explained by greater depletion of fibrinogen and factors V and VIII and higher levels of fibrin-degradation products.3 16 However, even after accounting for different effects of the thrombolytic regimens on aPTT, the risk of bleeding was greater after streptokinase than after accelerated TPA. The effect of thrombolytic therapy on hemostasis is complex, and other possible contributors to bleeding risk include cleavage of large-molecular-weight von Willibrand factor multimers17 18 and of glycoprotein Ib and IIb/IIIa platelet receptors.17 19
In contrast to the greater risk of noncerebral bleeding with streptokinase, the risk of intracranial hemorrhage was greater after accelerated TPA both in the GUSTO-I study8 and in previous studies.20 The reason for the differential effect on the risk of cerebral versus noncerebral bleeding is unclear. To the extent that noncerebral bleeding is more related to vascular injury after administration of thrombolytic therapy (ie, vascular puncture for cardiac catheterization or development of new gastric ulcer) and therefore to an intact ability to form a new fibrin clot, compared with cerebral bleeding as a result of preexisting vascular injury, one would expect the differential effect seen.
Predictors of Bleeding
Older age has been found to be a risk factor for bleeding after thrombolysis in some4 10 but not all11 previous studies. The largest previous experience, reported by the FTT Collaborative Group, did not find a relationship between older age and higher incidence of noncerebral bleeding. In GUSTO-I, however, age was the strongest predictor of moderate or severe bleeding. Even among patients who did not undergo interventional procedures or bypass surgery in GUSTO-I, age was the most important baseline characteristic in predicting hemorrhage.
Patients with lighter weight had a higher incidence of bleeding in prior studies of thrombolytic therapy for acute myocardial infarction. We have confirmed that more bleeding occurred in lighter-weight patients (Table 3⇑), whether or not an invasive management strategy was used.
Female sex, even after adjustment for the presence of other factors, was a potent predictor of bleeding both in prior trials10 12 and in GUSTO I. Weaver et al21 reported on the characteristics and treatment outcome of women in the GUSTO-I trial.
The finding of increased bleeding in patients of African ancestry is intriguing. In the Thrombolysis and Angioplasty in Myocardial Infarction (Phase I) study of 324 white patients and 24 patients of African ancestry, Sane et al6 noted an enhanced sensitivity among African Americans to TPA, which produced increased thrombolytic efficacy and more pronounced systemic fibrinogenolysis and resulted in an increase in bleeding that required transfusions. Lower fibrinogen levels and elevated fibrin(ogen)-degradation products in the circulation after treatment with TPA in this population is postulated to cause excessive bleeding due to poor clot formation and the antithrombotic properties of the fibrin(ogen)-degradation–product fragments.
Previous publications have noted an increase in bleeding complications with invasive management strategies,5 including balloon pumping22 and coronary angioplasty and bypass surgery.10 A second multivariate model was developed to evaluate the importance of the occurrence of hemorrhage with invasive procedures after adjusting for baseline characteristics and treatment assignment; this model showed that performance of an invasive procedure was independently associated with bleeding (Table 6⇑). The finding of more moderate and severe bleeding in US patients appears to be explained by the higher use of invasive management strategies in the United States. Because invasive procedures occurred after randomization, we cannot conclude whether the increased risk of bleeding was related to greater general severity of illness of the patient, the particular invasive procedure, or other instrumentation deemed necessary for patient care (eg, urinary catheter placement or nasogastric tube), data that were not collected.
When the effect of baseline characteristics and treatment of patients who did not have a procedure during their hospitalization was analyzed, age, weight, female sex, thrombolytic treatment assignment, and African ancestry, in order of importance, remained significant predictors for bleeding (Table 7⇑).
Patients treated with thrombolytic therapy for acute myocardial infarction are more likely to have bleeding complications if they are of increased age, lighter weight, female sex, or African ancestry or if they undergo invasive procedures. Less bleeding was associated with current smoking and with treatment with streptokinase plus subcutaneous heparin than with the other three thrombolytic regimens. Age, weight, treatment regimen, female sex, and African ancestry remained important predictors of bleeding when patients who had invasive procedures were excluded. Bleeding was associated with other adverse outcomes and with increased length of hospital stay and mortality at 30 days. Defining those patients at high risk for hemorrhagic complications should help practicing clinicians to balance the hazards of aggressive systemic thrombolytic therapies in the treatment of patients presenting with acute myocardial infarction.
This study was supported in part by grants from Bayer (New York, NY), CIBA-Corning (Medfield, Mass), Genentech (South San Francisco, Calif), ICI Pharmaceuticals (Wilmington, Del), and Sanofi Pharmaceuticals (Paris, France).
*A complete list of GUSTO-I Investigators has been published previously and can be found in Reference 77 .
Presented in part at the 68th Scientific Sessions of the American Heart Association, Anaheim, Calif, November 13-16, 1995, and previously published in abstract form (Circulation. 1995;92[suppl I]:I-460).
- Received June 18, 1996.
- Revision received December 5, 1996.
- Accepted December 16, 1996.
- Copyright © 1997 by American Heart Association
Fibrinolytic Therapy Trialists' (FTT) Collaborative Group. Indications for fibrinolytic therapy in suspected acute myocardial infarction: collaborative overview of early mortality and major morbidity results from all randomised trials of more than 1000 patients. Lancet. 1994;343:311-322.
Rao AK, Pratt C, Berke A, Jaffe A, Ockene I, Schreiber TL, Bell WR, Knatterud G, Robertson TL, Terrin ML, for the TIMI Investigators. Thrombolysis In Myocardial Infarction (TIMI) Trial phase I: hemorrhagic manifestations and changes in plasma fibrinogen and the fibrinolytic system in patients treated with recombinant tissue plasminogen activator and streptokinase. J Am Coll Cardiol. 1988;11:1-11.
Califf RM, Topol EJ, George BS, Boswick JM, Abbottsmith C, Sigmon KN, Candela R, Masek R, Kereiakes D, O'Neill WW, Stack RS, Stump D, and the Thrombolysis and Angioplasty in Myocardial Infarction Study Group. Hemorrhagic complications associated with the use of intravenous tissue plasminogen activator in treatment of acute myocardial infarction. Am J Med. 1988;885:353-359.
Bovill EG, Terrin ML, Stump DC, Berke AD, Frederick M, Collen D, Feit F, Gore JM, Hillis D, Lambrew CT, Leiboff R, Mann KG, Markis JE, Pratt CM, Sharkey SW, Sopko G, Tracy RP, Chesebro JH, for the TIMI Investigators. Hemorrhagic events during therapy with recombinant tissue-type plasminogen activator, heparin, and aspirin for acute myocardial infarction. Ann Intern Med. 1991;115:256-265.
Gore JM, Granger CB, Simoons ML, Sloan MA, Weaver WD, White HD, Barbash GI, Van de Werf F, Aylward PE, Topol EJ, Califf RM, for the GUSTO-I Investigators. Stroke after thrombolysis: mortality and functional outcomes in the GUSTO-I trial. Circulation. 1995;92:2811-2818.
Lee KL, Woodlief LH, Topol EJ, Weaver WD, Betriu A, Simoons M, Aylward P, Van de Werf F, Califf RM, and the GUSTO-I Investigators. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction: results from an international trial of 41 021 patients. Circulation. 1995;91:1659-1668.
White HD, Barbash GI, Modan M, Simes J, Diaz R, Hampton JR, Heikkila J, Kristinsson A, Moulopoulos S, Paolasso EAC, Van der Werf T, Pehrsson K, Sandoe E, Wilcox RG, Verstraete M, von der Lippe G, Van de Werf F. After correcting for worse baseline characteristics, women treated with thrombolytic therapy for acute myocardial infarction have the same mortality and morbidity as men except for a higher incidence of hemorrhagic stroke. Circulation. 1993;88:2097-2103.
Kalbfleisch J, Thadani U, LittleJohn JK, Brown G, Magorien R, Kutcher M, Taylor G, Maddox WT, Campbell WB, Perry J Jr, Spann JF, Vetrovec G, Kent R, Armstrong PW. Evaluation of a prolonged infusion of recombinant tissue-type plasminogen activator (duteplase) in preventing reocclusion following successful thrombolysis in acute myocardial infarction. Am J Cardiol. 1992;69:1120-1127.
Stone GW, Grines CL, Browne KF, Marco J, Rothbaum D, O'Keefe J, Hartzler GO, Overlie P, Donohue B, Chelliah N, Vliestra R, Puchrowicz-Ochocki S, O'Neill WW. Comparison of in-hospital outcome in men versus women treated by either thrombolytic therapy or primary coronary angioplasty for acute myocardial infarction. Am J Cardiol. 1995;75:987-992.
Gruppo Italiano per lo Studio della Sopravvivenza nell'Infarto Miocardioco. GISSI-2: in-hospital mortality and clinical course of 20 891 patients with suspected acute myocardial infarction randomised between alteplase and streptokinase with or without heparin. Lancet. 1990;336:71-75.
ISIS-3 (Third International Study of Infarct Survival) Collaborative Group. ISIS-3: a randomised comparison of streptokinase vs tissue plasminogen activator vs anistreplase and of aspirin plus heparin vs aspirin alone among 41 299 cases of suspected acute myocardial infarction. Lancet. 1993;339:753-770.
Collen D. Biological properties of plasminogen activators. In: Sobel BE, Collen D, Grossbard EB, eds. Tissue Plasminogen Activator in Thrombolytic Therapy. New York, NY: Marcel Dekker Inc; 1987:3-24.
Kamat SG, Michelson AD, Benoit SE, Moake JL, Rajasekhar D, Hellums JD, Kroll MH, Schafer AI. Fibrinolysis inhibits shear stress–induced platelet aggregation. Circulation. 1995;92:1399-1407.
Federici AB, Berkowitz SD, Zimmerman TS, Mannucci PM. Proteolysis of von Willebrand factor after thrombolytic therapy in patients with acute myocardial infarction. Blood. 1992;79:38-44.
Stricker RB, Wong D, Shiu DT, Reyes PT, Shuman MA. Activation of plasminogen by tissue plasminogen activator on normal and thromboasthenic platelets: effect of surface proteins and platelet aggregation. Blood. 1986;68:275-280.
Weaver WD, White HD, Wilcox RG, Aylward PE, Morris D, Guerci A, Ohman EM, Barbash GI, Betriu A, Sadowski Z, Topol EJ, Califf RM, for the GUSTO-I Investigators. Comparisons of characteristics and outcomes among women and men with acute myocardial infarction treated with thrombolytic therapy. JAMA. 1996;275:777-782.