Risk for Intracranial Hemorrhage after Tissue Plasminogen Activator Treatment for Acute Myocardial Infarction

Methods

Data Sources

The National Registry of Myocardial Infarction 2 (NRMI 2) was initiated in June 1994 as an ongoing registry of patients who received therapy for acute myocardial infarction at selected U.S. hospitals. The registry is supported by Genentech, Inc. (South San Francisco, California). From 1 June 1994 to 30 September 1996, 1484 U.S. hospitals contributed patients to NRMI 2. Participation in the registry is voluntary. Registry hospitals are substantially larger than nonparticipating U.S. hospitals: Twenty-seven percent of registry hospitals have more than 350 beds compared with 8% of nonregistry hospitals. In addition, registry hospitals are more likely than nonregistry hospitals to be certified by the Joint Commission on Accreditation of Health Care Organizations (99% compared with 77%); be affiliated with a medical school (36% compared with 17%); and have a coronary care unit (73% compared with 31%), a cardiac catheterization laboratory (72% compared with 23%), and a cardiac surgery program (39% compared with 11%).

Registry hospitals are encouraged to enter consecutive patients who have had acute myocardial infarction, regardless of treatment or outcome. Approval for hospital participation in the registry may include review by the local institutional review board or human research subjects committee as dictated by local policy. A study coordinator at each participating hospital completes individual data collection forms for each study patient; these forms are forwarded to an independent central data collection center (ClinTrials Research, Inc., Lexington, Kentucky) for processing. Data on individual hospitals are confidential and are available only to the contributing hospital.

Patients

Patients in our study had had acute myocardial infarction documented according to local hospital criteria (usually cardiac enzyme levels or results of electrocardiography or coronary angiography). For the purpose of our study, patients were those enrolled in NRMI 2 who received tPA as the initial reperfusion strategy from 1 June 1994 to 30 September 1996. To be eligible for study inclusion, patients could not have been transferred to a participating registry hospital from any other hospital (registry or otherwise) in the context of management of the acute myocardial infarction event. In addition, study patients could not have received a second dose of any thrombolytic agent.

As of 30 September 1996, NRMI 2 included 389 130 patients. Of these, 99 694 had received tPA as the initial reperfusion strategy; 26 370 of these patients had been transferred from another hospital for treatment of acute myocardial infarction. Of the remaining 73 324 patients, 2115 had received a second dose of a thrombolytic agent and 136 patients had missing information on age or sex. This left 71 073 patients in the study sample. The number of study patients contributed per registry hospital ranged from 1 to 311.

Definitions

The occurrence of primary intracranial hemorrhage was indicated on the registry data collection form, along with the date and time of onset of neurologic symptoms and whether computed tomography or magnetic resonance imaging (MRI) was performed to confirm the event. Events reported to have been confirmed by computed tomography or MRI were of principal interest in our study. The time between administration of tPA and intracranial hemorrhage was calculated. Sequelae of intracranial hemorrhage were characterized as death during hospitalization, residual deficit at discharge, or no residual deficit at discharge. The magnitude of the residual deficit was not classified. The reported intracranial hemorrhages and the circumstances surrounding them were not independently verified; information on these events was limited to that available on the data collection form provided by the participating hospitals.

We characterized patients according to age (<65 years, 65 to 74 years, or ≥ 75 years), sex, and ethnicity (white, black, or other). Clinical characteristics included history of myocardial infarction, angina, congestive heart failure, coronary artery bypass graft or percutaneous transluminal coronary angioplasty, stroke, diabetes mellitus, hypertension, hypercholesterolemia, and smoking. Systolic and diastolic blood pressure were characterized according to the first measurement recorded at hospital presentation. Likewise, Killip class was measured at presentation (no evidence of congestive heart failure, presence of rales or jugular venous distention, pulmonary edema, or cardiogenic shock) [10]. The study sample was stratified into quartiles according to body weight (measured in kg). The dose of tPA administered was categorized as less than 1.5 mg/kg or 1.5 mg/kg or more. These categories were based on manufacturer recommendations for tPA dosing [11]. (For patients weighing >67 kg, the maximum recommended total tPA dose is 100 mg [≤ 1.49 mg/kg]. Dosage adjustments based on weight are advised for patients weighing 67 kg or less.) Duration of tPA infusion was categorized as 90 minutes or less (accelerated tPA) or more than 90 minutes. We also characterized patients according to use of aspirin or intravenous heparin, which may be relevant to risk for bleeding.

Statistical Analysis

To assess comparability with a large clinical trial population, selected characteristics of our study sample were compared with those of patients who received accelerated tPA plus intravenous heparin in the GUSTO-I trial [5]. For our study sample, we evaluated the bivariate association between intracranial hemorrhage and selected demographic and clinical patient characteristics. These variables were used to develop stepwise multivariable logistic regression models with the occurrence of intracranial hemorrhage (confirmed by computed tomography or MRI) as the dependent variable. Patients with unknown values for any variable were excluded from multivariable analyses. The models were constructed with an entry significance level of P = 0.01 and an exit significance level of P = 0.05. Estimated odds ratios for the risk for intracranial hemorrhage, adjusted for all remaining variables, were obtained by using this model. Interactions between patient age and all other variables remaining in the final regression model were assessed. The goodness-of-fit criteria of Hosmer and Lemeshow were assessed for all models [12]. In addition, we calculated an area under the receiver-operating characteristic curve for each model [13]. All tests of statistical significance were two-tailed; a P value less than 0.05 was considered statistically significant. Multiple logistic regression analyses were performed by using the SAS PROC LOGISTIC procedure in SAS, version 6 (SAS Institute, Inc., Cary, North Carolina). The main-effects models were refit by using SAS PROC PHREG to control for potential interhospital variation.

Role of the Study Sponsor

Three of the investigators (Drs. Barron and Breen and Ms. Rundle) worked on this project as part of their duties as employees of Genentech, Inc. Dr. Breen and Ms. Rundle were principally responsible for all statistical analyses relating to the study, which were done by using data sets provided by ClinTrials, Inc. Before this paper was submitted for publication, it was reviewed and approved by the NRMI 2 Advisory Committee (Appendix). After approval of the manuscript by this committee, it was provided to Genentech, Inc., for internal review. No modifications of the study analyses or text were suggested.

Results

Patient Characteristics

Compared with patients who participated in the GUSTO-I trial and received accelerated tPA plus heparin, patients in our study were more likely to be female (Table 1). A history of diabetes mellitus or hypertension was more common and history of smoking was less frequently reported in our study than in the GUSTO-I trial. Median systolic blood pressure at the time of hospital admission was substantially higher in our study sample (140 mm Hg compared with 130 mm Hg in the GUSTO-I trial).

In the GUSTO-I trial, a history of stroke was explicitly listed among the exclusion criteria; 2.9% of the study sample from the registry had a history of stroke. In addition, according to study protocol, GUSTO-I trial participants received tPA over 90 minutes (accelerated tPA) and received weight-adjusted doses according to manufacturer recommendations.

Intracranial Hemorrhage Events

A total of 673 patients (0.95%) were reported to have had intracranial hemorrhage. The number of intracranial hemorrhages per participating hospital ranged from 0 (in 1019 hospitals) to 6. Among the 673 patients reported to have had intracranial hemorrhage, the event was confirmed by computed tomography or MRI in 625 patients (0.88%). Among the 673 patients with reported intracranial hemorrhage, older patients were less likely to have the event confirmed by computed tomography or MRI. Computed tomography or MRI was used to confirm 168 of 174 (96.6%) reported events among patients younger than 65 years of age, 231 of 247 (93.5%) reported events among those 65 to 74 years of age, and 226 of 252 (89.7%) reported events among those 75 years of age or older.

Information on the timing of the onset of neurologic symptoms was available for 96% of the 625 patients with events confirmed by computed tomography or MRI. Among the 625 patients, 331 (53%) died during the hospitalization because of the neurologic event. An additional 158 (25.3%) patients who survived to hospital discharge were reported to have some residual neurologic deficit. Among 600 patients for whom information on the timing of the event was available, 441 events (73.5%) occurred within 24 hours of tPA treatment, 78 (13%) occurred 24 to 48 hours after treatment, and 81 (13.5%) occurred 48 or more hours after treatment. The risk for death increased substantially with increasing patient age; 41.7% of 168 patients younger than 65 years of age, 48.9% of 231 patients 65 to 74 years of age, and 65.5% of 226 patients 75 years of age or older died during hospitalization.

Patient Characteristics and Intracranial Hemorrhage

In bivariate analyses, several patient characteristics were associated with intracranial hemorrhage, including advancing age (hemorrhage occurred in 0.40% of patients younger than 65 years of age, 1.24% of those 65 to 74 years of age, and 2.13% of those 75 years of age or older); female sex: black ethnicity; and history of stroke, hypertension, or smoking (Table 2). Systolic blood pressure of 140 mm Hg or more or diastolic blood pressure of 100 mm Hg or more were significantly associated with intracranial hemorrhage. We found an inconsistent relation between advancing Killip class and intracranial hemorrhage. A tPA dose of 1.5 mg/kg or greater was more commonly associated with intracranial hemorrhage than were lower doses. Increasing body weight and aspirin use were inversely related to intracranial hemorrhage.

Of the 71 073 study patients, 10 793 (15.2%) with unknown values for any of the candidate risk factors were excluded from multivariable analyses. In a model examining the main effects of candidate risk factors (Table 3), older patients were significantly more likely than younger patients to experience intracranial hemorrhage. Women and black persons were significantly more likely to experience this event. Systolic blood pressure of 140 mm Hg or more or diastolic blood pressure of 100 mm Hg or more was associated with the occurrence of intracranial hemorrhage. History of stroke and a tPA dose of 1.5 mg/kg or more continued to be strong risk factors. The P value for the goodness-of-fit test for this model was 0.18, indicating an adequate but less-than-perfect fit, and the area under the receiver-operating characteristic curve was 0.73. Controlling for possible hospital heterogeneity resulted in no meaningful change in the findings of the main-effects model.

We subsequently examined interactions between patient age and all other variables retained in the final regression model. Significant interactions were seen between age and history of stroke and age and sex; no significant interactions were found between age and any of the other variables included in the main-effects model. Table 4 summarizes the results of a multiple logistic regression model that allowed examination of the risk for intracranial hemorrhage for several patient groupings defined according to age, sex, and history of stroke. In this model, men who were younger than 65 years of age and had no history of stroke served as the reference against which all other patient groups were compared. After adjustment for other potential confounders, including ethnicity, systolic blood pressure, diastolic blood pressure, and tPA dose, advancing age was found to be directly related to intracranial hemorrhage in men and women without a history of stroke (Table 4). For men and women with a history of stroke, the effect of advancing age on risk for intracranial hemorrhage was overwhelmed by the strong effect of stroke history (Table 4).

In a separate model, we examined the effect of body weight on intracranial hemorrhage because this factor was shown to be associated with bleeding and intracranial hemorrhage after thrombolytic therapy [2, 6, 14]. This model did not include tPA dose because of the strong correlation between this variable and body weight. The model included all other variables previously described as being independently associated with intracranial hemorrhage. The results of this model revealed an inverse relation between body weight and intracranial hemorrhage (odds ratio, 0.81 per 10-kg increase in body weight [CI, 0.76 to 0.87]). We found no meaningful changes in the effects of any of the other variables in the model compared with those described in Table 3. The P value for the goodness-of-fit test for this model was 0.16, indicating an adequate but less-than-perfect fit, and the area under the receiver-operating characteristic curve was 0.74.

Discussion

Our examination of intracranial hemorrhage in a large national registry of patients with acute myocardial infarction who received tPA suggests that this event occurs more frequently in the general practice setting than in clinical trials. The overall incidence of this complication among the registry population was 0.95% if all reported intracranial hemorrhage events were considered and 0.88% if only events confirmed by computed tomography or MRI were considered.

We observed a substantial increase in the incidence of intracranial hemorrhage in older patients. For patients younger than 65 years of age, the incidence of confirmed intracranial hemorrhage was 0.40%; this incidence increased to 2.13% among persons 75 years of age or older. These estimates of incidence according to age group may underestimate the true effect of age on the risk for this complication because older patients in our study were less likely to have suspected intracranial hemorrhage confirmed by imaging studies. When all reported events are taken into account (including those not confirmed by computed tomography or MRI), the incidence of intracranial hemorrhage was 0.42% among patients younger than 65 years of age and 2.40% among those older than 75 years of age.

The incidence of intracranial hemorrhage observed in this usual-care setting is somewhat higher than that observed in the context of randomized clinical trials [2]. Participants in the GUSTO-I trial received tPA according to a standardized protocol that did not allow excessive dosing (≥ 1.5 mg/kg), and history of stroke was an exclusion criterion. In our study, more than 15% of patients received a tPA dose of 1.5 mg/kg or more; almost 3% of the patients were reported to have a history of stroke. Both factors were significantly associated with the occurrence of intracranial hemorrhage.

Our findings regarding other independent predictors of intracranial hemorrhage extend the results of previous studies. We found that elevated systolic (≥ 140 mm Hg) or diastolic (≥ 100 mm Hg) blood pressure at hospital presentation were significantly associated with the occurrence of intracranial hemorrhage. Female sex and black ethnicity were also associated with this complication. In analyzing the experience from the GUSTO-I trial, Gore and coworkers [2] noted a strong bivariate relation between female sex and intracranial hemorrhage that did not persist after controlling for other factors, including patient age, blood pressure, and body weight. In contrast, White and colleagues [15] found that women were at greater risk for intracranial hemorrhage even after adjustment for differences in baseline characteristics of the sexes. More recently, Berkowitz and associates [16] used data from the GUSTO-I trial to identify independent predictors of hemorrhagic events (excluding intracranial hemorrhage). When multivariable analysis was performed with patients who did not undergo invasive procedures, the most powerful independent predictors of hemorrhage were older patient age, lighter body weight, female sex, and black ethnicity.

One of our more interesting findings is the relation between age and intracranial hemorrhage according to history of stroke. To our knowledge, no study has fully examined the interaction between these two factors. We found that among men and women without a history of stroke, the risk for intracranial hemorrhage increased progressively with advancing age. For men and women with a history of stroke, however, the risk for intracranial hemorrhage was consistently high across all age groups.

Our findings emphasize the seriousness of intracranial hemorrhage as a complication of tPA administration in acute myocardial infarction. More than half of all patients who experienced intracranial hemorrhage died during hospitalization, and most patients who survived to hospital discharge experienced residual neurologic deficit. These rates of death and disability are consistent with the findings of Gore and colleagues [2], who examined intracranial hemorrhage in patients treated with various thrombolytic regimens.

The limitations of the registry data deserve mention. Each registry hospital is individually responsible for identifying cases of acute myocardial infarction and collecting data. Nonconsecutive patient enrollment is possible, and data are not independently validated. With regard to occurrence of such complications as intracranial hemorrhage, an expert panel did not independently review these events to develop and apply optimal standard definitions and classifications of neurologic events [2]. It was also not required that adverse cerebrovascular events be confirmed by computed tomography or MRI.

Intracranial hemorrhage is the most feared complication of thrombolytic therapy used in the management of acute myocardial infarction. Our findings emphasize the importance of individual risk assessment for this potential complication. Appropriate weight-adjusted dosing of tPA may limit the risk for this adverse consequence of thrombolytic therapy. The high risk associated with tPA treatment in patients with a history of stroke, regardless of age or sex, suggests that other therapies, such as primary coronary angioplasty, may be preferable [17].

Appendix

The following are members of the National Registry of Myocardial Infarction 2 Advisory Committee: Nisha Chibber Strobos, MD, Johns Hopkins University School of Medicine; William J. French, MD, University of California, Los Angeles, School of Medicine; Joel M. Gore, MD, University of Massachusetts Medical School; Costas Lambrew, MD, University of Vermont School of Medicine; Joseph P. Ornato, MD, Medical College of Virginia; Jan Penney, RN, MSN, CCRN, MidMichigan Regional Medical Center; William Rogers, MD, University of Alabama Medical School; Alan Tiefenbrunn, MD, Washington University School of Medicine; and W. Douglas Weaver, MD, Henry Ford Hospital.

A complete list of hospitals participating in the National Registry of Myocardial Infarction 2 is available from ClinTrials, Inc., 2365 Harrodsburg Road, Lexington, KY 40504.

From University of Massachusetts Medical School and the Fallon Healthcare System, Worcester, Massachusetts; University of California, San Francisco, California; Genentech, Inc., South San Francisco, California; Harbin Clinic, Rome, Georgia; University of California, Los Angeles, School of Medicine, Los Angeles, California; and University of Alabama Medical Center, Birmingham, Alabama.

Drs. Gore and Goldberg: Department of Medicine, University of Massachusetts Medical School, 55 Lake Avenue North, Worcester, MA 01655.

Drs. Barron and Breen and Ms. Rundle: Genentech, Inc., Medical Affairs, 1 DNA Way, South San Francisco, CA 94080.

Dr. Sloan: Department of Neurosciences, Harbin Clinic, 1825 Martha Berry Boulevard, Rome, GA 30165-1698.

Dr. French: Harbor UCLA Medical Center, 1000 West Carson Street, Torrance, CA 90509.

Dr. Rogers: University of Alabama at Birmingham, 334 Lyons-Harrison Research Building, 701 South 19th Street, Birmingham, AL 35294-0007.