Risk Factors for In-hospital Nonhemorrhagic Stroke in Patients With Acute Myocardial Infarction Treated With Thrombolysis
Results From GUSTO-I
Background—Nonhemorrhagic stroke occurs in 0.1% to 1.3% of patients with acute myocardial infarction who are treated with thrombolysis, with substantial associated mortality and morbidity. Little is known about the risk factors for its occurrence.
Methods and Results—We studied the 247 patients with nonhemorrhagic stroke who were randomly assigned to one of four thrombolytic regimens within 6 hours of symptom onset in the GUSTO-I trial. We assessed the univariable and multivariable baseline risk factors for nonhemorrhagic stroke and created a scoring nomogram from the baseline multivariable modeling. We used time-dependent Cox modeling to determine multivariable in-hospital predictors of nonhemorrhagic stroke. Baseline and in-hospital predictors were then combined to determine the overall predictors of nonhemorrhagic stroke. Of the 247 patients, 42 (17%) died and another 98 (40%) were disabled by 30-day follow-up. Older age was the most important baseline clinical predictor of nonhemorrhagic stroke, followed by higher heart rate, history of stroke or transient ischemic attack, diabetes, previous angina, and history of hypertension. These factors remained statistically significant predictors in the combined model, along with worse Killip class, coronary angiography, bypass surgery, and atrial fibrillation/flutter.
Conclusions—Nonhemorrhagic stroke is a serious event in patients with acute myocardial infarction who are treated with thrombolytic, antithrombin, and antiplatelet therapy. We developed a simple nomogram that can predict the risk of nonhemorrhagic stroke on the basis of baseline clinical characteristics. Prophylactic anticoagulation may be an important treatment strategy for patients with high probability for nonhemorrhagic stroke, but further study is needed.
Stroke is one of the most feared complications in patients with acute myocardial infarction. In the era before the routine use of thrombolytic therapy and anticoagulation, stroke was observed in 1.7% to 3.2% of patients.1 2 3 4 5 Intracranial hemorrhage was exceedingly rare. In the thrombolytic era, the overall incidence of stroke in large clinical trials is lower, but the types of strokes have changed. Nonhemorrhagic stroke now occurs in 0.1% to 1.3% of patients, and intracranial hemorrhage occurs in 0.07% to 1.5% of patients.6 7 8 9 10 11 12
Recently, multivariable regression models have been developed to help identify patients at increased risk for intracranial hemorrhage after thrombolysis, including a model based on the GUSTO-I trial experience.7 8 9 13 14 15 16 Less is known about the risk factors for nonhemorrhagic stroke in patients with acute myocardial infarction treated with thrombolytic therapy, although several investigators have reported clinical and echocardiographic factors associated with an increased risk for nonhemorrhagic stroke.7 8 10 11 12 17 18 19 20 21 22 23 24 25
We used prospectively collected information from the GUSTO-I trial for patients with and without nonhemorrhagic stroke to determine independent risk factors associated with in-hospital nonhemorrhagic stroke in patients with acute myocardial infarction treated with thrombolytic therapy.
The study population was the 41 021 patients enrolled in the GUSTO-I trial.26 In brief, patients presenting with acute myocardial infarction within 6 hours of symptom onset were randomly assigned to one of four thrombolytic strategies: (1) 1.5 million U streptokinase IV over 1 hour and 12 500 heparin U SC BID; (2) 1.5 million U streptokinase IV over 1 hour and 5000 U heparin IV bolus followed by 1000 U/h; (3) alteplase in an accelerated regimen (15-mg IV bolus followed by 0.75 mg/kg over 30 minutes and then 0.5 mg/kg over the next hour) and 5000 U heparin IV bolus followed by 1000 U/h; or (4) both 1 million U streptokinase and 1.0 mg/kg alteplase IV over 1 hour and 5000 U heparin IV bolus followed by 1000 U/h. All patients received aspirin (160 to 325 mg) daily. The overall experience with stroke in the GUSTO-I has been previously reported.9
Nonhemorrhagic stroke was defined as an acute new neurological deficit resulting in death or lasting for >24 hours without hemorrhage on computed tomography or magnetic resonance imaging, as classified by a physician with supporting documentation from discharge summaries, progress notes, brain images, neurological/neurosurgical evaluation, or autopsy reports.
Sustained hypotension was defined as systolic blood pressure of <90 mm Hg for >1 hour despite fluid replacement.
Cardiogenic shock was defined as systolic blood pressure of <90 mm Hg for ≥1 hour, not responsive to fluid replacement alone, thought to be secondary to cardiac dysfunction, and associated with signs of hypoperfusion (cool, clammy skin, oliguria, or altered sensorium or cardiac index of ≤2.2 L · min−1 · /m−2). If the systolic blood pressure increased to >90 mm Hg as a result of positive inotropic agents alone within 1 hour, this was still classified as cardiogenic shock.
Congestive Heart Failure
Congestive heart failure was considered to be signs or symptoms of congestion (rales above the lung base, dyspnea, pulmonary edema on chest radiograph, or peripheral edema) or low cardiac output (weakness, fatigue) thought to be secondary to cardiac dysfunction.
Killip class I was defined as the absence of rales over the lung fields and the absence of S3 gallop. Class II included patients who had rales across ≤50% of the lung fields or the presence of an S3 gallop. Class III was defined as rales across >50% of the lung fields, and class IV was defined as pulmonary edema with hypoperfusion (cardiogenic shock).
Ejection fraction was determined by the site investigator at the time of cardiac catheterization and, therefore, is available only for patients undergoing angiography.
The evaluation of patients with stroke and the collection of specific data about stroke treatment and clinical outcomes in the trial were prospectively planned. The protocol specified that all patients with a new focal neurological deficit undergo a complete evaluation including brain imaging. Patients with nonhemorrhagic stroke were identified from the responses to specific questions on the Case Report Form or from a two-page supplemental Stroke Details Form, which was completed for each patient with a suspected stroke. In addition, the study collected available supporting documents, such as physicians’ notes, operative reports, discharge summaries, autopsy reports, and results of computed tomographic and magnetic resonance imaging studies.
Nonhemorrhagic Stroke Adjudication
A Stroke Review Committee (see “Appendix”) was established to adjudicate and categorize all suspected strokes. For each suspected nonhemorrhagic stroke, the medical records and brain imaging studies were independently reviewed by two teams, each consisting of a cardiologist and neurologist. Each team was responsible for determining whether a nonhemorrhagic stroke had occurred. Disagreements were resolved by a third team.
Patient Functional Assessment
Functional status was determined for all patients with nonhemorrhagic stroke at hospital discharge or 30 days, whichever came first. Patients were classified as not disabled if they had no deficit (no sequelae) or minor deficits (functional status unchanged) and as disabled if they had moderate deficits (significant limitations of activity) or severe deficits (unable to live independently or work).9
Continuous variables are shown as medians with 25th, and 75th percentiles; discrete variables are shown as frequencies and percentages. Cox proportional hazards modeling was used to determine the univariable predictors of nonhemorrhagic stroke and determine the multivariable baseline risk factors for nonhemorrhagic stroke. When examining the effect of interventions on nonhemorrhagic stroke, we considered that some patients may have had a nonhemorrhagic stroke before they could receive an intervention or may have had a planned intervention canceled. Crediting the nonintervention group with all of these early events would unfairly inflate the lack of association of the stroke and the intervention. As a result, we used time-dependent Cox modeling to determine multivariable in-hospital predictors of nonhemorrhagic stroke. The interventions included as time-dependent covariates were cardiac catheterization, percutaneous transluminal coronary angioplasty, coronary artery bypass graft surgery, and intra-aortic balloon counterpulsation. In-hospital congestive heart failure and cardiogenic shock were similarly treated as time-dependent covariates. A backward-elimination method was used to determine the significant predictors from the baseline and in-hospital analyses (elimination criterion, P<.05). Baseline and in-hospital predictors were then combined to determine the overall predictors of nonhemorrhagic stroke. Predictors in each analysis were tested using the likelihood ratio χ2 test. Results are also presented as hazard ratios and 95% confidence intervals.
We created a scoring nomogram on the basis of coefficients from the baseline multivariable regression modeling. Each independent predictor was assigned a particular score according to its predictive value. The sum of the scores indicates the probability of a nonhemorrhagic stroke based on baseline predictors for individual patients (F. E. Harrell: Design S plus functions for biostatistical/epidemiological modeling, testing, estimation, validation, graphics, prediction, and typesetting by storing enhanced model design attributes in the fit. UNIX version available from firstname.lastname@example.org; 1996).
There were 247 patients with nonhemorrhagic stroke in the GUSTO-I population (0.6%), of whom 42 (17%) died. In all, 56.7% of patients with nonhemorrhagic stroke died or were disabled at the time of hospital discharge (Table 1⇓).
The baseline clinical characteristics for the 246 patients with in-hospital nonhemorrhagic stroke (1 patient had a nonhemorrhagic stroke after hospital discharge and 30-day follow-up and is not included in the analyses) and the 40 688 patients without nonhemorrhagic stroke are shown in Table 2⇓. Patients with nonhemorrhagic stroke were older and more often female and had a higher incidence of previous coronary disease, hypertension, and diabetes. They also tended to have a worse baseline Killip class and more often had atrial fibrillation/flutter and anterior infarcts at enrollment.
Patients with nonhemorrhagic stroke were more likely to have adverse in-hospital events and bypass surgery (Table 3⇓). The univariable χ2 values and hazard ratios for the baseline and in-hospital factors associated with in-hospital nonhemorrhagic stroke are shown in Tables 4⇓ and 5⇓.
Baseline Multivariable Predictors
Six baseline clinical and historical characteristics were significant independent predictors of in-hospital nonhemorrhagic stroke (Table 6⇓). Age was the most important baseline clinical predictor. Higher heart rate, history of stroke or transient ischemic attack, diabetes, previous angina, and history of hypertension were the other statistically significant baseline predictors. Atrial fibrillation/flutter on the baseline ECG and Killip class at the time of enrollment were not statistically significant independent predictors of nonhemorrhagic stroke (χ2=3.13, P=.08; and χ2=5.23, P=.16, respectively). The predicted probability of nonhemorrhagic stroke based on baseline clinical and historical data can be calculated for individual patients with the nomogram shown in Fig. 1⇓.
In-hospital Multivariable Predictors
Worse Killip class was the most potent predictor for nonhemorrhagic stroke, followed by atrial fibrillation/flutter and performance of coronary artery bypass surgery and cardiac catheterization (Table 7⇓). In the analysis of in-hospital factors, patients with atrial fibrillation/flutter at baseline were compared with those who did not have atrial fibrillation/flutter at baseline and did not develop the arrhythmia during the hospital stay. In addition, patients who develop atrial fibrillation/flutter during hospitalization were compared with patients without atrial fibrillation at baseline or in-hospital. Patients developing atrial fibrillation/flutter after enrollment had a greater than twofold higher incidence of nonhemorrhagic stroke than patients who did not have atrial fibrillation/flutter at baseline or during hospitalization (hazard ratio, 2.08; 95% confidence interval, 1.52 to 2.86). Patients with atrial fibrillation/flutter at baseline also had a higher likelihood of nonhemorrhagic stroke than patients who never had atrial fibrillation/flutter during the hospital stay (hazard ratio, 2.44; 95% confidence interval, 1.43 to 4.16).
Combined Multivariable Predictors
All 10 independent predictors from the baseline and in-hospital multivariable analyses except history of hypertension remained statistically significant, independent predictors of nonhemorrhagic stroke in the combined analysis (Table 8⇓). Worse Killip class and older age were the two most important predictors, followed by performance of cardiac catheterization and bypass surgery. A second combined analysis included only the 15 220 (37.1%) patients undergoing angiography who had a documented ejection fraction (Table 9⇓). Age, performance of cardiac catheterization, and bypass surgery were significant and potent independent predictors of nonhemorrhagic stroke. Lower ejection fraction, higher heart rate, diabetes, and worse Killip class had borderline statistical significance as predictors of nonhemorrhagic stroke in the combined model of patients with known ejection fraction. Atrial fibrillation/flutter, previous angina, previous hypertension, and stroke or transient ischemic attack were no longer statistically significant independent predictors.
Stroke is a well-documented event in patients with acute myocardial infarction. The routine use of anticoagulant and thrombolytic therapies has reduced the occurrence of nonhemorrhagic stroke but increased the risk of intracranial hemorrhage. Nonhemorrhagic stroke, however, is associated with significant morbidity and mortality.9 We identified six easily determined baseline patient clinical and historical factors (older age, higher heart rate, history of stroke or transient ischemic attack, diabetes, previous angina, and history of hypertension) that independently predict nonhemorrhagic stroke in patients with acute myocardial infarction treated with thrombolysis, heparin, and aspirin. In addition, four common in-hospital characteristics (worse Killip class, atrial fibrillation/flutter, bypass surgery, and cardiac catheterization) are independently associated with nonhemorrhagic stroke.
Several investigators have reported that patients with increased age,7 8 10 anterior myocardial infarction,8 10 11 and worse Killip class7 10 are at increased risk for nonhemorrhagic stroke. In our analysis, anterior infarct was not a univariable predictor of nonhemorrhagic stroke, although there was a trend toward a higher incidence of anterior infarction in patients with nonhemorrhagic stroke than in those without stroke (44.5% versus 39.0%; P=.13). In addition, isolated apical infarct and combined anteroapical infarct also were not statistically significant univariable predictors of nonhemorrhagic stroke (χ2=0.75, P=.39; and χ2=3.23, P=.072, respectively).
From 28% to 40% of patients with anterior infarction and 0% to 1.5% of patients with inferior infarction have a left ventricular thrombus, according to small echocardiographic studies performed primarily in the era before the routine use of thrombolysis and aggressive anticoagulation.17 18 19 20 23 24 25 The incidence of systemic embolization was 0% to 33% in patients with documented thrombus, lower in patients receiving anticoagulation, and exceedingly rare in patients with no observed thrombus.17 18 19 20 21 22 23 24 25 Echocardiographic features associated with an increased risk of embolization include older age,19 larger thrombus size,19 thrombus mobility,20 21 22 23 and pendulousness of the clot.19 20 21 22 23 Because echocardiography was not mandated by the GUSTO-I protocol, the incidence of left ventricular thrombus in the GUSTO-I population is unknown. In addition, data were not collected on warfarin therapy, which may have influenced these findings. Further study is required to clarify the relationship among location of infarction, incidence of thrombus, impact of anticoagulation, and risk of nonhemorrhagic stroke.
The incidence of stroke associated with the procedures and events that this study found to correlate with increased risk of nonhemorrhagic stroke—specifically, atrial fibrillation/flutter, cardiac catheterization, and coronary artery bypass surgery—has been well established by previous investigators. The risk of stroke in patients with nonvalvular atrial fibrillation is ≈5% per year, or a fivefold to sevenfold higher incidence compared with patients without the arrhythmia.27 28 Several large registries have reported that stroke occurs in <0.01% of patients undergoing cardiac catheterization.29 30 The incidence of stroke in patients undergoing coronary artery bypass surgery has been reported to occur in 0.4% to 5.4% of patients.31 32 33 One of the largest most comprehensive multicenter series reported nonfatal stroke in 2.6% of patients after coronary artery bypass surgery.34
A unique aspect of this study was the incorporation of in-hospital events in the analysis, to examine not only baseline clinical characteristics but also in-hospital factors common in patients after acute myocardial infarction—such as atrial arrhythmias, coronary interventions, and bypass surgery—that may predict nonhemorrhagic stroke. The finding of the Fibrinolytic Therapy Trialists’ Cooperative Group that 62% of the nonhemorrhagic strokes in patients treated with thrombolysis occurred >24 hours after enrollment emphasizes the importance of this analysis.6 In addition, 60% of nonhemorrhagic strokes in the GUSTO-I trial occurred >48 hours after randomization (Fig. 2⇓).9
Our study has several limitations. An important aspect of the study was the time-dependent analysis of clinical events and procedures in relation to the onset of stroke symptoms. The time of onset for atrial fibrillation or worse Killip class was not known, so only the day of occurrence was used in the time-dependent analysis and compared with the day of stroke symptom onset. Dates and times for cardiac interventions and other in-hospital events were available.
Because information was not collected about the use of warfarin therapy, the impact of anticoagulation in this patient population cannot be assessed. In addition, information was not collected about postdischarge events (other than mortality and procedures that may be associated with nonhemorrhagic stroke). However, only one nonhemorrhagic stroke occurred after hospital discharge and before 30-day follow-up. A proportion of the GUSTO-I population may have had a cerebrovascular event that was pathophysiologically related to the index myocardial infarction >30 days after enrollment.
In the in-hospital model, patients who presented with or developed atrial fibrillation/flutter (compared with those without the arrhythmia) were at higher risk for nonhemorrhagic stroke. However, in the baseline multivariable model, atrial fibrillation/flutter at enrollment did not independently predict nonhemorrhagic stroke. Duration of atrial fibrillation/flutter before enrollment, which we were unable to assess, may have been a contributing factor.
The GUSTO-I stroke dataset, which included 247 nonhemorrhagic stroke patients, is the largest single-trial experience to date. The systematic review of all suspected strokes by the Stroke Review Committee and the availability of computed axial tomographic, magnetic resonance imaging, or autopsy data for 93% of stroke patients resulted in definitive classification of stroke subtypes; thus, the number of “unknowns” was small. The mechanism of nonhemorrhagic stroke was not determined, and some strokes may not have been related to cardiac events or procedures. Despite the relatively large numbers, the sample size still limited the ability to perform more extensive multivariable regression modeling.
Finally, these results are only applicable to the patients with acute myocardial infarction treated with thrombolytic therapy and should not be generalized to all patients with acute myocardial infarction.
Nonhemorrhagic stroke remains a serious event in patients with acute myocardial infarction treated with thrombolytic, antithrombin, and antiplatelet therapy and is associated with significant morbidity and mortality. Older age, higher heart rate, history of stroke or transient ischemic attack, diabetes, previous angina, and history of hypertension are independent clinical and historical predictors of nonhemorrhagic stroke; physicians can use these factors to determine the baseline probability of nonhemorrhagic stroke with a simple scoring nomogram (Fig 1⇑). This information may enhance awareness of the potential for nonhemorrhagic stroke early after myocardial infarction. Beyond the baseline characteristics, adverse in-hospital cardiac events, such as worsening Killip class, atrial fibrillation/flutter, and performance of cardiac catheterization or bypass surgery also are potent, independent predictors of nonhemorrhagic stroke.
Patients with several risk factors or a high probability for nonhemorrhagic stroke may benefit from prophylactic anticoagulation, but further study is needed.
The following are GUSTO-I Stroke Committee Members: Gabriel I. Barbash, MD, MPH, Cardiologist, and Rina Tadmor, MD, Neurologist, Israel; Joel M. Gore, MD, Cardiologist, and Michael A. Sloan, MD, Neurologist, USA; Maarten L. Simoons, MD, Cardiologist, and Peter Koudstaal, MD, Neurologist, The Netherlands; W. Douglas Weaver, MD, Cardiologist, and W.T. Longsteth, Jr, MD, Neurologist, USA; and Harvey D. White, MV, DSc, Cardiologist, and Neil Anderson, Neurologist, New Zealand.
This study was supported by Bayer (New York, NY), CIBA-Corning (Medfield, Mass), Genentech (South San Francisco, Calif), ICI Pharmaceuticals (Wilmington, Del), and Sanofi Pharmaceuticals (Paris, France). The authors thank John Daniel for his help with manuscript preparation and editorial support.
Reprint requests to Kenneth W. Mahaffey, MD, Box 3630, Duke University Medical Center, Durham, NC 27710.
Members of the GUSTO-I Stroke Committee are listed in the “Appendix.”
- Received June 23, 1997.
- Revision received October 9, 1997.
- Accepted October 20, 1997.
- Copyright © 1998 by American Heart Association
Thompson PL, Robinson JS. Stroke after acute myocardial infarction: relation to infarct size. BMJ. 1978;2:457–459.
Komrad MS, Coffey CE, Coffey KS, McKinnis R, Massey EW, Califf RM. Myocardial infarction and stroke. Neurology. 1984;34:1403–1409.
Report of the Working Party on Anticoagulant Therapy in Coronary Thrombosis to the Medical Research Council. Assessment of short-anticoagulant administration after cardiac infarction. BMJ. 1969;1:335–342.
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.
Maggioni AP, Franzosi MG, Santoro E, White H, Van de Werf F, Tognoni G, for the Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto Miocardico (GISSI-2) and the International Study Group. The risk of stroke in patients with acute myocardial infarction after thrombolytic and antithrombotic treatment. N Engl J Med. 1992;327:1–6.
Gore JM, Sloan M, Price TR, Randall AM, Bovill E, Collen D, Forman S, Knatterud GL, Sopko G, Terrin ML. Intracerebral hemorrhage, cerebral infarction, and subdural hematoma after acute myocardial infarction and thrombolytic therapy in the Thrombolysis in Myocardial Infarction Study: Thrombolysis in Myocardial Infarction, Phase II, pilot and clinical trial. Circulation. 1991;83:448–459.
Gore JM, Granger CB, Simoons ML, Sloan MA, Weaver WD, White HD, Barbash GI, Van de Werf F, Aylward PE, Topol EJ, for the GUSTO-I Investigators. Stroke after thrombolysis: mortality and functional outcomes in the GUSTO-I trial. Circulation. 1995;92:2811–2818.
Maggioni AP, Franzosi MG, Farina ML, Santoro E, Celani MG, Ricci S, Tognoni G, Gruppo Italiano per lo Studio della Streptochinasi nell’Infarto Miocardico (GISSI). Cerebrovascular events after myocardial infarction: analysis of the GISSI trial. BMJ. 1991;302:1428–1431.
Longstreth WT Jr, Litwin PE, Weaver WD, for the MITI Project Group. Myocardial infarction, thrombolytic therapy, and stroke: a community-based study. Stroke. 1993;24:587–590.
Selker HP, Beshansky JR, Schmid CH, Griffith JL, Longstreth WT Jr, O’Connor CM, Caplan LR, Massey EW, D’Agostino RB, Laks MM. Presenting pulse pressure predicts thrombolytic therapy-related intracranial hemorrhage: Thrombolytic Predictive Instrument (TPI) project results. Circulation. 1994;90:1657–1661.
Johannessen K-A, Nordrehaug JE, von der Lippe G. Left ventricular thrombosis and cerebrovascular accident in acute myocardial infarction. Br Heart J. 1984;51:553–556.
Johannessen K-A, Nordrehaug JE, von der Lippe G, Vollset SE. Risk factors for embolisation in patients with left ventricular thrombi and acute myocardial infarction. Br Heart J. 1988;60:104–110.
Stratton JR, Resnick AD. Increased embolic risk in patients with left ventricular thrombi. Circulation. 1987;75:1004–1011.
Weinreich DJ, Burke JF, Pauletto FJ. Left ventricular mural thrombi complicating acute myocardial infarction: long-term follow-up with serial echocardiography. Ann Intern Med. 1984;100:789–794.
Wolf PA, Dawber TR, Thomas HE, Kannel WB. Epidemiologic assessment of chronic atrial fibrillation and risk of stroke: the Framingham study. Neurology. 1978;28:973–977.
Mora CT, Murkin JM. The central nervous system: responses to cardiopulmonary bypass. In: Mora CT, ed. Cardiopulmonary Bypass: Principles and Techniques of Extracorporeal Circulation. New York, NY: Springer-Verlag; 1995:114–146.
Shaw PJ, Bates D, Cartlidge NEF, Heaviside D, Julian DG, Shaw DA. Early neurological complications of coronary artery bypass surgery. BMJ. 1985;291:1384–1387.
Roach GW, Kanchuger M, Mangano CM, Newman M, Nussmeier N, Wolman R, Aggarwal A, Marschall K, Graham SH, Ley C, Ozanne G, Mangano DT, for the Multicenter Study of Perioperative Ischemia Research Group and the Ischemia Research and Education Foundation Investigators. Adverse cerebral outcomes after coronary bypass surgery. N Engl J Med. 1996;335:1857–1863.