Clinical features and pathogenesis of intracerebral hemorrhage after rt-PA and heparin therapy for acute myocardial infarction

Intracranial hemorrhage is an infrequent though severe complication of thrombolytic therapy for acute myocardial infarction. The reported frequencies of intracranial hemorrhage in previous studies are 0.1% to 0.8% for streptokinase, ^[1-12] 0.3% to 0.9% for recombinant tissue-type plasminogen activator (rt-PA), ^[6-10,12-18] and 0.8% for anisoylated plasminogen-streptokinase activator complex ^[9]. The Global Utilization of Streptokinase and rt-PA for Occluded Coronary Arteries (GUSTO) Trial reported ^[12] that intracranial hemorrhage occurred in 0.9% of patients treated with combined streptokinase plus accelerated rt-PA. Since these trials did not employ routine screening with brain CT following thrombolytic therapy, the frequency of occult or undiagnosed intracranial hemorrhages is unknown.

The Thrombolysis in Myocardial Infarction (TIMI) Phase II Pilot Study and Clinical Trial assessed the effects on mortality, re-infarction, other morbidity, and left ventricular function of an invasive treatment strategy (cardiac catheterization and, if angiographic findings demonstrated appropriate anatomy, percutaneous transluminal coronary angioplasty (PTCA) or coronary artery bypass grafting if coronary anatomy was too complex or hazardous for PTCA) versus a conservative strategy following intravenous rt-PA and heparin therapy for acute myocardial infarction ^[19-21]. As previously reported in TIMI II, ^[16] 23 of 56 cerebrovascular complications were primary parenchymatous intracerebral hemorrhages (ICHs), with a higher rate in patients treated with 150 mg rt-PA (1.3%) than in patients treated with 100 mg rt-PA (0.4%). Primary ICH was associated with increasing age but not female gender. In this report, we delineate the clinical and radiologic features, manner of presentation, and temporal course of primary ICH as well as limited hematologic and pathologic findings.

Methods. TIMI II protocol. Detailed descriptions of the TIMI II methods, cardiovascular and hemostatic findings, and overall results have been previously reported ^[16,20,21]. Patients presenting within the first 4 hours of acute MI with ST segment elevation received rt-PA, heparin, and aspirin. An initial 5,000-IU bolus of IV heparin was given at the start of rt-PA infusion; continuous infusion began within 1 hour at a rate of 1,000 IU/hr, and the dose was then adjusted to maintain the activated partial thromboplastin time (aPTT) between 1.5 and 2.0 times that of control. An unexpected number of intracranial hemorrhages occurred when patients were treated with 150 mg of rt-PA in the early part of TIMI II ^[22,23]. Protocol changes were made, and the dose of rt-PA was reduced to 100 mg for the remaining 3,016 patients (60 mg in the first hour, including a 6-mg bolus; 20 mg in the second hour; and 5 mg in each of the next 4 hours) ^[16,23,24]. Aspirin (80 mg/d) on the same day as thrombolytic therapy was part of the study regimen in the first 627 patients in the TIMI II pilot and clinical trial. For the remaining 3,297 patients, initiation of aspirin was postponed to the next day and then increased to 325 mg/d on day 6, when IV heparin was replaced with subcutaneous heparin.

Clinical evaluation. Detailed information regarding circumstances surrounding the onset of the hemorrhage was recorded, including the date and time of onset of neurologic symptoms and signs. The earliest possible time was defined as the time of onset of first symptoms or signs compatible with CNS dysfunction. This could be a headache, with or without a focal deficit. When a patient was asleep or unconscious and responded or awoke with obvious signs of a focal deficit, the earliest possible time was defined as the time of loss of consciousness or going to sleep. The latest possible time was defined as the time when unequivocal evidence of CNS dysfunction was present.

Neurologic data were recorded by a neurologist or other responsible physician and abstracted by study staff on special data collection forms. Particular attention was paid to the mode of onset (rapid, gradual, or stepwise), time interval to maximal deficit, level of consciousness, and nature and distribution of presenting neurologic symptoms and signs. Diagnostic evaluation consisted of CT. CTs for 22 of 23 patients (96%) were reviewed centrally by two of the investigators (M.A.S. and T.R.P.). One patient with ICH whose CT could not be obtained for review was evaluated on the basis of a local CT report and a neurosurgeon's operative note.

Neuroradiology. A focal neurologic deficit associated with a focal collection of blood in the brain seen on CT without evidence of preceding ischemic infarction was classified as an ICH. The primary site was defined as the lesion most likely to be responsible for each patient's presenting symptoms and signs. A primary site was considered multilobar if multiple cerebral lobes were involved through contiguous spread. The presence of hemorrhage extending to other parenchymal sites, ventricles, subarachnoid space, or combination of sites was noted, as were multiple, distinct lesions.

The size of the ICH at the primary site was estimated by a modification of the methods of Hier et al ^[25] and Kase et al ^[26]. The volume of the hematoma was estimated by multiplying the maximum length and width (cross-sectional area) by the height of the high absorption lesion on the CT. This number was then divided by two in an attempt to correct for the deviation of the lesion from cuboid shape ^[25,26].

Neuropathology. Postmortem records were available for five of the 11 fatal cases (45%). A limited number of hematoxylin-eosin-stained glass slides and paraffin blocks of the brain from four patients and slides stained with Luxol fast blue-periodic acid Schiff (LFB-PAS) from one were available for review by one of the authors (C.K.P.). Immunohistochemistry to detect amyloid beta protein (ABP) was performed on all five cases using unstained glass slides of cerebral cortex in four and decolorized LFB-PAS-stained slides in the fifth. The slides were exposed to 88% formic acid for 5 minutes, ^[27] exposed to hydrogen peroxide (H₂ O₂) for 20 minutes, and incubated with normal goat serum for 30 minutes. They were then incubated with rabbit polyclonal anti-ABP (gift of Dr. Samuel Gandy, Rockefeller University) diluted 1:500 in phosphate-buffered saline with 1% bovine serum albumin at 4 degrees C overnight. Following this, the slides were incubated with biotinylated secondary antibody at 27 degrees C for 30 minutes and the avidin-biotin complex ^[28] (Vector Labs, Carpenteria, CA) for 50 minutes. They were incubated with 3,3-diaminobenzidine and H₂ O₂ for 5 minutes and lightly counterstained with hematoxylin. Phosphate-buffered saline with bovine serum albumin was substituted for the primary antibody as a negative control and a previously diagnosed case of cerebral amyloid angiopathy (CAA) was used as a positive control. Congo red stain and the modified Bielschowsky silver stain were performed on two to three sections of cerebral cortex from the four patients whose paraffin blocks were available for resectioning. The frequency of diffuse and neuritic plaques was evaluated according to the criteria of Khachaturian ^[29] and Mirra et al ^[30].

Results. Data from this study are presented in tabular form as follows. Table 1contains each patient's age, weight, time interval between treatment initiation and neurologic symptom onset, time interval between neurologic symptom onset and performance of CT, site and size of primary ICH, earliest clinical abnormalities, time interval from neurologic symptom onset to maximal deficit, mortality, neurologic follow-up time, and residual deficit. Table 2contains information on medical history (hypertension, atrial fibrillation, and prior cerebrovascular disease), ventricular arrhythmias, blood pressures, fibrinogen and aPTT (when available), presence of multiple hemorrhages, and underlying pathologic lesion (if known). Table 3summarizes the neuropathologic data from the five autopsied patients.

Baseline characteristics. Of the 23 patients in this series, 14 (61%) were men and nine (39%) women. Twenty patients (87%) were white. The age range was 42 to 75 years, with a mean age of 63.3 years; 57% of the patients were in the seventh decade and 22% were in the eighth decade. Weight at study entry was available in 21 patients (91%); low body weight (<70 kg) was known to be present in three of 10 patients (30%) in the 150-mg rt-PA group and six of 11 patients (54%) in the 100-mg rt-PA group (table 1). Fourteen patients (61%) had histories of hypertension. Five patients had evidence of prior cerebrovascular disease: three with stroke, one with transient ischemic attacks, and one with transient ischemic attacks and stroke. No information is available about the type or location of stroke. The ECG location of myocardial infarction was anterior in 15 patients (65%).

At the time of study entry, 10 patients (43%) had systolic blood pressures greater than 140 mm Hg, and 10 patients (43%) had diastolic blood pressures greater than 90 mm Hg (table 2). Thirteen patients (56.5%) developed or maintained a blood pressure >=160 mm Hg systolic or >=90 mm Hg diastolic during the rt-PA infusion. Ten patients (43%) had systolic blood pressures >=160 mm Hg during the rt-PA infusion; two of these were normotensive at study entry. Thirteen patients had diastolic blood pressures >=90 mm Hg (five were >=100 mm Hg) during the rt-PA infusion; one had been normotensive at study entry. One of these patients had an ongoing hypertensive crisis that was difficult to control (patient 9). Of the eight patients with blood pressures recorded at or near the time of symptom onset, five had systolic blood pressures >=160 mm Hg and/or diastolic blood pressures >=100 mm Hg. In comparison, of the 3,901 patients who did not have ICH, 1,024 (26.2%) had a blood pressure >=160 mm Hg systolic and/or >=90 mm Hg diastolic during the first 3 hours of the rt-PA infusion.

Six of 23 patients (26%) had life-threatening cardiac arrhythmias (table 2). Ventricular fibrillation occurred in five and ventricular tachycardia occurred in one. The time of occurrence of ventricular fibrillation was before the rt-PA infusion in three, during the rt-PA infusion in one, and after termination of the rt-PA infusion but before onset of neurologic symptoms in one. The one instance of ventricular tachycardia occurred after termination of the rt-PA infusion and after onset of neurologic symptoms. Ventricular fibrillation occurred before or during the rt-PA infusion in four of 23 patients (17.4%) with ICH and 181 of 3,901 patients (4.6%) without ICH. Atrial fibrillation occurred in only two of the 23 patients (9%). In one of these (patient 2), symptoms occurred after cardioversion of rapid atrial fibrillation in the setting of acute anterior-wall myocardial infarction and a history of prior stroke.

Timing, clinical features, and temporal course. The earliest possible times of onset for 19 of the 23 hemorrhages (83%) were within the first 24 hours after initiation of treatment. Nine (39%) occurred during the 6-hour study drug infusion; six (26%) between 6 and 12 hours after initiation of the infusion; and four (17%) between 12 and 24 hours after initiation of the infusion. The times of occurrence of symptoms among the six patients with multiple hemorrhages were 4.0 hours, 5.7 hours, 7.5 hours, 9.5 hours, 11.8 hours, and 13.9 hours after initiation of treatment.

The most frequent earliest documented neurologic symptom and sign in the 23 patients studied was a decreased level of consciousness, observed in 15 patients (65%); two patients (9%) were in coma. Eight patients (35%) had focal neurologic deficits. Other presenting complaints included vomiting in six patients (26%), seizures in five (22%), and headache in three (13%). Headache was noted in two patients with lobar hematomas and the one patient with multiple cerebellar hematomas. Eleven patients (48%) had more than one neurologic abnormality at symptom onset (table 1). In addition, documented symptoms and signs of neurologic deterioration at the time of examination in the 23 patients with ICH included decreased level of consciousness (lethargy or stupor) in 82%, coma in 30%, hemiparesis in 62%, disordered oculomotor function in 48%, nausea/vomiting in 39%, and disorientation in 30%. Of the five patients with seizures, four had lobar hemorrhages (25% of the 16 lobar cases), and one had a midbrain hemorrhage with extension to the subarachnoid space.

The onset of symptoms was rapid (less than 10 minutes) in seven (30%) and gradual in 16 (70%) patients. The time to maximal deficit was less than 6 hours in 14 (61%), 6 to 12 hours in four (17%), 12 to 24 hours in three (13%), and 24 to 48 hours in two (9%) patients. There was possible clinical improvement within 24 hours of onset in six patients (nos. 5, 10, 15, 17, 19, and 22). In patients 10 and 17, this may have represented recovery from a postictal state. In patient 15, the sedative effect of diazepam may have been resolving.

For 19 of 23 patients (83%), the time interval between the earliest possible time of neurologic symptom onset and performance of the first CT was known. The time interval was less than 2 hours in eight, 2 to 6 hours in two, 6 to 12 hours in five, 12 to 24 hours in three, and more than 24 hours in one. For four patients (nos. 5, 9, 11, and 15), the exact time interval could not be determined (table 1). For the patients with known time intervals between neurologic symptom onset and performance of CT, the time interval ranges were 0.50 to 7.50 hours (mean, 2.63 hours) for the four patients with clinical improvement within 24 hours of ICH onset, and 0.25 to 42.25 hours (mean, 8.52 hours) for the 16 patients without clinical improvement within 24 hours of ICH onset.

Hemostatic measurements. In the 17 patients with available platelet counts, thrombocytopenia (platelet count <150,000/mm³) was not present at the onset of symptoms. The aPTT was greater than 65 seconds in close temporal relation to symptom onset in 12 of the 23 patients (52%). Three patients in the 150-mg rt-PA group had profound hypofibrinogenemia (15 mg/dl, 15 mg/dl, and 32 mg/dl) at or near the time of symptom onset; only one patient in the 100-mg rt-PA group had a fibrinogen level as low as 85 mg/dl (table 2). Three of the six patients with multiple hemorrhages had fibrinogen levels measured at or near the time of symptom onset; 15 mg/dl each in two patients (nos. 5 and 6) and 85 mg/dl in one (no. 15). There was no obvious relation between degree of hypofibrinogenemia and size of ICH.

Site and size of intracerebral hemorrhages. The sites of the primary hemorrhage for the 23 cases were lobar in 16 (70%), thalamic in four (17%), and brainstem-cerebellum in three (13%) (table 1). Specific primary lobar sites included frontal (5), parietal (1), temporal (1), occipital (1), and multiple contiguous sites (8). Multilobar sites were frontoparietal (3), parieto-occipital (2), temporoparietal (2), and frontotemporal (1). Extension of the ICH to other sites occurred in five patients (22%). Specific primary sites in the posterior fossa were midbrain (2) and cerebellum (1). Extension of the hemorrhage to the ventricular system occurred in 13 cases (57%) and to the subarachnoid space in nine cases (39%). In one patient (no. 18), a left temporoparietal ICH was associated with a serpentine structure coursing medially and inferiorly, suggestive of an arteriovenous malformation. Multiple distinct hemorrhages occurred in six of 23 patients (26%; nos. 5, 6, 8, 15, 20, and 21), three in the 150- mg rt-PA group and three in the 100-mg rt-PA group. Five of the six patients had multiple cerebral lobar hemorrhages and one had multiple cerebellar hemorrhages. Only one (no. 21) had more than five sites involved.

The size of the primary hematomas ranged from 1 cc to 166 cc, with a mean size of 35.8 cc (table 1). There were three distinct size groups: seven (30%) small-volume hemorrhages (1 to 9 cc), 12 (52%) moderate-volume hemorrhages (10 to 49 cc), and four (17%) large-volume hemorrhages (>=50 cc). Within 12 hours of initiation of rt-PA therapy, four of four large, eight of 12 medium, and three of seven small hemorrhages occurred. There were no important differences in the sizes of the hemorrhages associated with the 150-mg and 100-mg rt-PA regimens.

Neuropathologic findings. Table 3summarizes the clinical and neuropathologic findings of the five patients with fatal ICH who had autopsy examinations. Four of these patients developed neurologic symptoms and signs within 12 hours of study entry, two of five patients (nos. 6 and 21) had ventricular arrhythmias, and none had uncontrolled hypertension.

Three patients had CAA associated with multiple subcortical hemorrhages. The arteries and arterioles of the cerebral leptomeninges and cortex were dilated, and the walls were thickened and replaced by amorphous pink material that was strongly immunoreactive to ABP. The vascular changes were extensive in two of the three patients but were only mild in the third. Congo red stains performed in two of the three patients (nos. 6 and 21) were positive and showed characteristic yellow-green dichroism when viewed under polarized light. In addition, frequent cortical neuritic plaques (>15 per 200 x field) were present in two of these patients, and frequent diffuse plaques and a few neuritic plaques were found in the third. The etiology of the multiple subcortical hemorrhages in patient 4 was not identified. The cortical blood vessels were only slightly thickened, amyloid was not identified with either Congo red stain or immunohistochemistry, and neither thromboemboli nor infarcts were observed.

The massive intraventricular hemorrhage in one patient (no. 2) was anatomically contiguous with a hemorrhagic infarction in the periventricular white matter. The age of the infarct was consistent with the 3-month interval between the hemorrhagic complication and death. In retrospect, the diagnosis of hemorrhagic infarction may have been suspected in view of four distinctive clinical features: history of prior stroke, acute anterior-wall myocardial infarction, cardioversion for rapid atrial fibrillation, and late occurrence of neurologic symptoms (74 hours after initiation of thrombolytic therapy). In this patient, the time interval between neurologic symptom onset and performance of the CT was 18 hours.

Management and outcome. Eighteen of the 23 patients with ICH were treated medically with discontinuation of heparin (17), administration of protamine (5) and epsilon-aminocaproic acid (1), and infusion of fresh frozen plasma (4) and cryoprecipitate (4). One patient (no. 22) was not receiving anticoagulants at the time of onset of neurologic symptoms. Steroids or mannitol were given to six patients. Five had discontinuation of heparin and surgical evacuation of the hematoma, and two received ventriculostomies for hydrocephalus. Three of the five patients who underwent craniotomy died.

The overall case fatality rate within 1 month was 11 of 23 patients (48%): seven of 12 patients (58%) in the 150-mg rt-PA group and four of 11 patients (36%) in the 100-mg rt-PA group (table 2). Fatal hemorrhages occurred in five of nine patients (56%) with onset of symptoms within less than 6 hours after initiation of study drug infusion and five of six (83%) within 6 to 12 hours. In eight of these 10 fatal cases (80%), patients were comatose when initially examined. One patient succumbed due to underlying cardiac disease, possible mesenteric ischemia, and sepsis (no. 10), and another had a massive pulmonary embolus (no. 15).

Discussion. In TIMI II, 23 patients (0.58%) met previously defined criteria ^[16] for primary ICH associated with rt-PA and heparin therapy for acute myocardial infarction. Most of the hemorrhages (83%) occurred within 24 hours (with 15, or 65%, within 12 hours) of initiation of rt-PA infusion. This is similar to findings in other studies ^{[1-5,13,15,31-41]} and suggests a temporal relationship between rt-PA administration and ICH, with occurrence and enlargement of the hemorrhage during the period of active fibrinolysis ^[42,43]. This bleeding may be potentiated by concomitant administration of antithrombotic therapy ^{[1,2,16,19,31,33,37,38,42,44]} and other factors. An arteriovenous malformation was observed on CT in one patient. Pathologic examination in five ICH patients revealed important clues to pathogenesis in four patients: CAA in three and hemorrhagic transformation of cerebral infarction in one.

Clinical factors contributing to ICH occurrence. Fourteen patients (61%) had histories of hypertension, and 13 (57%) developed or maintained elevated blood pressure (>=160/90 mm Hg) during rt-PA infusion. In reported series of ICH occurring secondary to chronic hypertension, hemorrhages tended to be found most often in the basal ganglia. The predominance of lobar hemorrhages (70%) in TIMI II is unusual for hypertensive ICH ^[45-47]. The distribution of the primary hemorrhage sites in this series is similar to the distribution of ICH associated with long-term oral anticoagulant therapy, ^[48-50] although the cerebellum ^[48] was not a frequent site of bleeding.

These findings suggest that in TIMI II, chronic hypertension may not be the preeminent cause of ICHs associated with thrombolytic therapy ^[26,45-47]. Uncontrolled hypertension was a reason for exclusion of patients from enrollment in TIMI II. However, after enrollment, acute hypertension, ie, >=160 mm Hg systolic or >=100 mm Hg diastolic, arising during infusion of fibrinolytic agents may have contributed to the occurrence of ICH in 10 patients (43%) ^[14,51]. Although elevated blood pressure was noted more frequently in the patients with ICH, this may reflect either different intensity of patient observation between the groups or a cardiovascular response to the ICH.

Life-threatening ventricular arrhythmias occurred in six patients (26%), with five of six (83%) cases occurring before onset of neurologic symptoms. Prolonged cardiopulmonary resuscitation (CPR) may lead to global hypoxic-ischemic encephalopathy. Several authorities recommend that CPR for less than 10 minutes should not be a contraindication to thrombolytic therapy ^[52,53]. Whether CPR is brief or prolonged, marked fluctuations in blood pressure due to pharmacologic therapy and electrical cardioversion are associated with hemodynamic instability that may lead to dramatic changes in cerebral perfusion. If cerebral vessels are structurally weakened, then the combined effect of thrombolytic therapy and blood pressure fluctuations may promote the occurrence of intracranial bleeding. While ventricular fibrillation before or during the rt-PA infusion was noted more frequently in the patients with ICH, this may reflect either a different level of patient observation between the groups or an association not previously reported that requires further study.

Underlying cerebrovascular pathology. The reported frequency of multiple sites of ICH in patients treated with thrombolytic agents is 15 to 38% ^[13,15,42]. The 26% frequency of multiple ICHs in TIMI II confirms these findings. Limited data in the literature ^[54-56] suggest that CT-detectable multiple hemorrhages may occur in 2% ^[55] to 11% ^[54] of spontaneous ICH cases. They occur most frequently in patients with leukemia and other blood dyscrasias, coagulopathies, neoplasms (primary and metastatic), vasculitis, venous sinus thrombosis, and CAA ^[57-60] but rarely with chronic hypertension ^[54,55].

Structural lesions in the brain may predispose to ICH. Hemorrhage might occur by induction of de novo bleeding from a previously unruptured lesion or rebleeding from a vessel that had bled previously. In the setting of thrombolytic therapy for acute myocardial infarction, arteriovenous malformations ^[16,39] and CAA ^[40-42,61] are the only structural lesions that have been reported in patients with ICH.

CAA is due to the infiltration of cerebrum-specific amyloid into the media and adventitia of small to medium cerebral or cerebellar arteries, arterioles, and veins ^[57-60]. The affected vessels are structurally brittle and unable to withstand trauma or blood pressure changes ^[57,59]. The suspicion that CAA contributes to ICH in patients treated with thrombolytic agents is based on the cerebral or cerebellar lobar location, multiplicity, increasing frequency in older patients, ^{[6-8,13-16,42,62,63]} and the growing number of reports demonstrating the association in surgical ^[42,61] and autopsy ^[40,41] specimens. In the present study, four of five patients whose brains were examined at autopsy had subcortical lobar hemorrhages characteristic of CAA. In three of these four patients, we found multiple hemorrhages and immunohistochemical evidence of CAA. Four of the six multiple ICHs were fatal; three of these patients had CAA. In addition, two of these three patients had numerous neuritic plaques in the cerebral cortex; these plaque frequencies would have been consistent with the diagnosis of Alzheimer's disease had the patients been demented ^[29,30]. However, not all patients with CAA-related ICH after thrombolysis have microscopic Alzheimer's disease ^[41,61].

Recent data support the possibility of a connection between increasing age, ICH, and CAA. The GUSTO Trial reported a three- or fourfold increase in the risk of hemorrhagic stroke in patients >75 years old treated with streptokinase or accelerated rt-PA compared with patients <=75 years old ^[12]. In future studies, it would be interesting to briefly assess cognitive function in patients, particularly those over age 75, before thrombolytic treatment is given.

One of the five patients (no. 2) had pathologic evidence for a confluent hemorrhagic infarction that extended to the ventricular system. No other patient in this series had the combination of clinical features suggestive of preceding ischemic cerebral infarction that were present in this patient (no. 2). In the other patient with atrial fibrillation (no. 14), the timing of the ICH after thrombolytic treatment and the typical pattern of extension of the massive thalamic ICH make hemorrhagic transformation of a cerebral infarction less likely.

Even with stringent clinical and radiologic criteria, it may be extremely difficult to distinguish a primary ICH from a confluent hemorrhagic infarction ^[62-66]. Mutlu et al ^[65] observed that four cases of fatal lobar or white matter hemorrhages were actually "catabolic rupture hemorrhages" associated with thrombotic occlusion. Bogousslavsky et al ^[66] demonstrated that confluent hemorrhagic transformation of a cerebral infarction may occur within 16 + 3 hours of stroke onset, particularly in patients with potential cardiac sources of embolism, prior TIAs, and previously undiagnosed stroke on CT. In one case in that study, ^[66] autopsy failed to demonstrate histologic evidence of an underlying ischemic cerebral infarction.

Some studies ^[67,68] suggest that hemorrhagic infarction following cardioembolic stroke generally occurs within 4 days of onset, rarely on the first day or as late as 11 days after onset. In TIMI II, there was a shorter mean time interval between ICH onset and performance of CT in patients with clinical improvement within 24 hours of ICH onset than in those who did not improve within 24 hours. The longer time to CT performance in patients who did not improve makes it less likely that confluent hemorrhagic infarctions occurred in patients who did improve and were not detected due to a delay in performance of CT.

Of the 29 ischemic cerebral infarctions in TIMI II, ^[64] eight (28%) had hemorrhagic infarctions. It is not known how often this event occurred in other studies ^{[1-10,13,14,31-38,62,69]}. It is presently unknown what proportion of "ICH" diagnosed on clinical grounds or with neuroimaging techniques would be classified pathologically as confluent hemorrhagic infarctions. In the GUSTO trial, ^[12] hemorrhagic transformation of ischemic cerebral infarction occurred in a small proportion of patients who had ischemic strokes. We therefore believe that the one misclassified patient (no. 2) in our series represents a small minority of patients who have hemorrhagic infarctions misclassified as ICHs.

Hemostatic measurements. In nine cases in the literature ^{[15,36,44,70-72]} and in three of our patients, ICH occurred in association with documented severe hypofibrinogenemia. In TIMI II, ^[73] patients with or without ICH who received 150 mg rt-PA had higher plasma rt-PA levels, higher fibrinogen degradation product levels, and lower plasminogen levels, and were more likely to have lower fibrinogen levels, than patients receiving 100 mg rt-PA. However, as previously reported, ^[16] there were no large differences in fibrinogen, fibrinogen degradation products, rt-PA antigen level, plasminogen levels, and aPTT among the 23 patients with ICH, the 29 patients with cerebral infarction, and all other patients who did not have cerebrovascular events. The small number of observations precludes firm conclusions regarding the relation between the existence of a systemic lytic state and CNS bleeding.

Role of multiple factors. A number of investigators have identified factors that may increase the risk of intracranial hemorrhage following thrombolytic therapy for acute myocardial infarction ^[10,74]. In a multiple logistic regression analysis of data on intracranial hemorrhage patients from a number of clinical trials, including the present study, Simoons et al ^[74] suggested that four variables known at hospital admission appeared to be related to intracranial hemorrhage: elderly age (>65 years), low body weight (<70 kg), hypertension upon admission (systolic blood pressure >=170 mm Hg and/or diastolic blood pressure >=95 mm Hg), and alteplase (rt-PA) regimen. In patients receiving alteplase, the odds ratios were 3.2 (95% CI = 1.8-5.6) for elderly age, 2.5 (95% CI = 1.4-4.4) for low body weight, and 2.0 (95% CI = 1.0-3.9) for hypertension on admission. Assuming an overall intracranial hemorrhage risk of 0.75%, the risk estimate varied from 0.26% for a patient without intracranial hemorrhage risk factors who received streptokinase to 5.0% for an elderly hypertensive patient with low body weight treated with alteplase ^[74].

In TIMI II, prior cerebrovascular disease, wide fluctuations in blood pressure due to the occurrence and treatment of ventricular arrhythmias, marked hypofibrinogenemia, and prolonged aPTT after rt-PA and heparin therapy may have enhanced bleeding from weakened amyloid-infiltrated vessels and led to the multiple ICHs in one patient (no. 6). Ventricular arrhythmias in the setting of a prolonged aPTT may lead to ICH (patient 12). Persistent or acute hypertension (either systolic or diastolic) ^[14,44,51] in the presence of a prolonged aPTT may lead to single (patients 4, 8, 10, 16, and 23) or multiple (patient 15) ICH. In other ICH cases, presently unknown factors, other than the effects of combined thrombolytic/anticoagulant therapy, may be responsible for ICH.

Clinical features. The finding of a decreased level of consciousness at presentation in 15 of 23 (65%) and on initial examination in 20 of 23 patients (87%) is striking. Our patients frequently had a rapid onset and progression of symptoms and signs; 61% reached their maximal deficit within 6 hours (78% within 12 hours) of symptom onset. This temporal course is more characteristic of "hypertensive" hemorrhage ^[75] than of ICH related to long-term anticoagulant therapy ^[48]. Our findings differ from those of Wijdicks and Jack, ^[42] who found focal neurologic deficits at onset followed by a rapid decrease in level of consciousness in all eight of their patients. The variations in clinical presentation may be due to the site or size of ICH, rapidity of onset and progression before clinical recognition, and other factors. The occurrence of decreasing alertness, decreased level of consciousness, or focal neurologic deficits, especially if rapid in onset and progression, within 24 hours of rt-PA infusion for acute myocardial infarction should be interpreted as suggesting ICH, and this diagnosis must be excluded as soon as possible.

The reported observation of possible improvement within 24 hours of onset in six of the 23 cases (26%) is unusual for parenchymatous ICH ^[75]. In our study, two patients had motor seizures at presentation, and improvement in these cases may have reflected resolution of the postictal state. Improvement in the other four cases may have reflected the resolution of more clinically subtle complex partial seizures, toxic-metabolic disturbances, sedation, or improved cardiac function. There is no direct evidence to support or refute the hypothesis that these four patients had confluent hemorrhagic transformation of a preceding ischemic cerebral infarction.

Prognosis of intracerebral hemorrhage. Previous studies ^[76-78] have shown that decreased level of consciousness is associated with a poor prognosis in patients with ICH. In this study, 13 of the 20 patients (65%) with decreased level of consciousness at initial examination died (table 1). There was a high case fatality rate (10/15, or 67%) for patients with ICH within 12 hours of initiation of treatment. These findings are similar to those of Wijdicks and Jack ^[42].

Therapeutic implications. These results have several implications for treatment of acute myocardial infarction with fibrinolytic agents. First, the occurrence of ICH following thrombolytic therapy for acute myocardial infarction may be difficult to predict. Second, the occurrence of a decreased level of alertness or appearance of a neurologic deficit, particularly in the 24 hours after initiation of thrombolytic therapy, must be suspected to reflect ICH until proven otherwise. Third, it may be informative to evaluate cognitive/neurologic function in older patients before giving thrombolytic therapy. Fourth, blood pressure should be carefully controlled before, during, and after infusion of the fibrinolytic agent. Finally, more data are needed to delineate the microvascular/microscopic pathology associated with ICH after thrombolytic therapy for acute myocardial infarction.

Acknowledgments

The authors express gratitude to the clinicians and pathologists who made data available for this study, and thank Myra Franklin and Patricia Haworth for preparation of the manuscript.