Cerebral hemorrhage after intra-arterial thrombolysis for ischemic stroke

Subjects and methods.

The study design of the PROACT II trial has been described in detail elsewhere.7 In brief, subjects with acute ischemic stroke less than 6 hours from onset were eligible if they had angiographic documentation of an occlusion of the M1 or M2 segment of the MCA, CT scan with no evidence of hemorrhage or acute infarction encompassing more than one-third of the MCA territory, and an NIHSS score between four and 30 at the time of randomization. After informed consent, eligible subjects were randomized to either IA r-proUK at a total dose of 9 mg delivered directly into the clot via a microcatheter over a 120-minute period, along with fixed-dose IV heparin (2000 U initial bolus followed by 500 U/h for 4 hours, starting together with the r-proUK infusion), or fixed-dose IV heparin alone (the control group). After completion of the IA infusion of r-proUK, the subjects were transferred to the intensive care unit for close monitoring of blood pressure and neurologic function. Regardless of clinical condition, all patients had a head CT scan 24 hours after r-proUK infusion, or earlier if clinical deterioration occurred.

The main safety parameter was the presence of ICH with neurologic deterioration, defined as an increase of four or more points in the NIHSS score in comparison with the preangiography score, within 36 hours from treatment initiation. The adjudication of all cases was performed by an independent committee composed of three neurologists (C.S.K., R.G.H., G.F.M.) and a member of the Neuroradiology Core facility (H.A.R.), who evaluated all cases reported by the sites as serious adverse events, as well as all procedural complications and deaths, and all instances of any clinical or CT evidence of intracranial bleeding, either symptomatic or asymptomatic. The detailed analysis of each case was done during a total of 14 periodic meetings in which the clinical and imaging data were reviewed by the four members of the committee who were blinded to treatment assignment. To ensure that the reviewers were fully blinded to treatment assignment, angiographic films that contained evidence of randomization group, such as the presence of a microcatheter, were omitted from review. This study incorporated a formal safety “stopping rule” that was to be applied if IA r-proUK was associated with an unacceptably high rate of ICH with neurologic deterioration after any of the periodic reviews. This particular safety review was performed after each successive group of 15 patients was enrolled. Based on reported data from the literature on the frequency of symptomatic ICH in trials or clinical series of thrombolysis in patients with acute ischemic stroke, the stopping rule was set at a rate of symptomatic ICH that exceeded 15% as the lower limit of the 95% CI of the current point estimate. In the event of symptomatic ICH surpassing this limit, the trial would temporarily (pending review of the data) stop further patient accrual. To maintain blinding within the committee, the calculations were performed by one of the statistical consultants (M.G.), who was required to simply report whether the “stopping rule” was met (i.e., there was no disclosure of interim data if the stopping guideline was not met).

During the review sessions, the committee members classified all cases as to type of event, relationship to procedure (angiography), and potential relationship to study drug. In addition, all cases with hemorrhagic lesions on CT scan were labeled as having either hemorrhagic infarction (areas of irregular, mottled hyperintensity, at times confluent, on a background of hypodensity of infarction) or parenchymal hemorrhage (dense and homogeneous areas of hyperintensity, with associated mass effect depending on size, with potential for intraventricular extension), with or without clinical deterioration. A listing of these cases with hemorrhagic lesions and all those resulting in death regardless of mechanism was forwarded to the medical monitor (L.S.F.), who in turn provided an unblinded listing to one of the statistical consultants (M.G.) to monitor differential ICH rates on an ongoing basis. After this analysis, the committee, the sponsor, and the sites were informed as to whether case enrollment into the study should continue. None of the periodic reviews of the rate of symptomatic ICH required the study to discontinue recruitment.

A total of 24 baseline patient demographic and clinical features was considered in association with symptomatic ICH. One postrandomization variable, recanalization of the target vessel (based on the 120-minute angiogram), was also included. These factors included both quantitative (e.g., age, NIHSS score) and qualitative (e.g., sex, history of diabetes) variables. Each variable was examined separately for association with symptomatic ICH; for qualitative variables, this was achieved via Fisher’s exact test8 and for quantitative variables via linear logistic regression.9 For this exploratory analysis, statistical significance was accepted at the 0.05 level, without adjustment for multiple comparisons. The rate of ICH was compared between patients receiving IA r-proUK and controls with a Fisher’s exact test. All p values were two sided.

Results.

Of a total of 180 subjects enrolled into the trial, 121 were randomized to receive r-proUK, and 59 were randomized to the control group. For a variety of reasons, 13 patients randomized to the r-proUK group did not receive the drug (not meeting angiographic criteria [four]; exceeding the 6-hour window [three]; because of technical difficulties [two]; and extreme agitation, post-angiogram neurologic deterioration, pharmacy error, and improper informed consent [one each]), whereas two patients randomized to the control group received r-proUK (in both instances owing to pharmacy error). Three of the 13 subjects randomized to the r-proUK group who did not receive the drug, along with three control subjects, received out-of-protocol IA urokinase and were excluded from analysis for this study. This resulted in 110 patients being treated with r-proUK and 64 not receiving any thrombolytic agent (figure 1). The focus of this study is on the 110 patients treated with r-proUK.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Study group, formed by the final numbers of patients in the r-proUK and control groups, after changes from the original randomization groups occurred.

Symptomatic ICH occurred in 12 of the 110 patients (10.9%) treated with r-proUK and in two of the 64 (3.1%) who received IV heparin alone. This difference was not significant (p = 0.086; relative risk [RR] 3.5; 95% CI, 0.8 to 31.5). The CT scans of these 14 instances of ICH are shown in figure 2. The hemorrhages in the r-proUK–treated patients corresponded to either large confluent hemorrhagic infarcts (Cases 1, 9, and 11) or large deep hemispheric hematomas (Cases 3 through 8 and 10), whereas others were relatively small hemorrhagic areas within large hypodense, edematous MCA distribution infarcts (Cases 2 and 12). All these hemorrhages occurred in the area of the preceding infarct and had the first onset of ICH symptoms at a mean of 10.2 ± 7.4 hours after the onset of treatment. The interval between stroke onset and time of treatment with r-proUK in the 12 patients with symptomatic ICH (5.17 ± 0.9 hours) was not different from that of the remaining 98 patients who did not develop symptomatic ICH (5.14 ± 0.8 hours).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. CT scans of the symptomatic ICH in the patients treated with r-proUK (Patients 1 through 12) and in the two controls (bottom row).

To relate the severity of the initial clinical deficit to various outcomes, the baseline NIHSS score was stratified in three groups: 4 to 10, 11 to 20, and 21 to 30. All 12 of the r-proUK–related symptomatic ICH occurred in patients with NIHSS scores higher than 10 (nine ICH among 75 patients [12%] in the 11 to 20 NIHSS score stratum and three ICH in the 23 patients [12%] in the 21 to 30 stratum).

The main symptoms at onset of ICH in the 12 patients with r-proUK–related symptomatic ICH included decline in the level of consciousness (75%), sudden rise in blood pressure (25%), headache and vomiting (17%), increase in the severity of the presenting focal neurologic deficit (17%), and seizures (7%). The mortality for patients experiencing symptomatic ICH was 83% (10 of 12 patients).

In an attempt to assess predictors of symptomatic ICH in the 12 subjects treated with r-proUK, we evaluated several baseline and one posttreatment variable (angiographic recanalization). The univariate association between various demographic and clinical factors and symptomatic ICH is summarized in table 1 (qualitative factors) and table 2 (quantitative factors). Only three of the variables considered, history of hypercholesterolemia, CT infarct volume, and serum glucose, showed relatively strong statistical evidence of association with symptomatic ICH. The strongest evidence was found for the relationship with serum glucose, at a nominal p value of 0.014. The postrandomization variable angiographic recanalization, evaluated at 120 minutes after treatment onset, was not associated with an increased risk of symptomatic ICH.

Because only one relatively strong prognostic variable was identified via the univariate analyses, the multivariable logistic model was unnecessary. A logistic model predicting symptomatic ICH from serum glucose measurements (included in the model as a quantitative variable) indicated a 1.3% increase in the odds of symptomatic ICH for each 1 mg/dL increase in serum glucose. Adjustment for baseline NIHSS score, a predictor of risk of thrombolysis-related ICH,1 did not alter the association: the unadjusted regression coefficient of 0.0128 and the p value of 0.014 in the logistic model became 0.0124 and 0.018, respectively, after adjusting for baseline NIHSS score. When serum glucose was stratified into categories of increasing levels of hyperglycemia, those subjects with values of >200 mg/dL had a risk of symptomatic ICH of 36% (figure 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Rate of symptomatic ICH after IA r-proUK therapy and baseline serum glucose.

None of the variables recorded at the time of onset of symptoms of ICH compared with values obtained at similar time points in those without symptomatic ICH was found to be significantly different between the r-proUK–treated subjects with and without symptomatic ICH. These included systolic and diastolic blood pressure, mean arterial pressure, and maximal activated partial thromboplastin time (aPTT) (data not shown).

Discussion.

The risk of symptomatic intracranial hemorrhage after thrombolysis for acute ischemic stroke is substantial, regardless of whether the agent is infused by the IV or IA routes. There are already considerable data on the ICH risk after IV treatment with t-PA in clinical trials1,10-13⇓⇓⇓⇓ and with its use in the community.2,3,14,15⇓⇓⇓ The use of IV t-PA in various studies has several differences that may account for the varying rates of ICH, including t-PA dosage, length of the therapeutic window, and, possibly, deviations from the guidelines for t-PA administration,14,15⇓ although this latter factor had no impact on the rate of ICH in two recently reported studies.2,3⇓ Potential local vascular factors related to ICH risk after IV t-PA treatment are unknown because this agent is routinely administered without preceding angiography. Studies such as PROACT II, which included angiography, provide a unique opportunity to study the potential role of local vascular abnormalities in the pathogenesis of the hemorrhagic complications.

The 10.9% rate of symptomatic ICH for the 110 patients who received IA r-proUK in the PROACT II trial is higher than most of the rates reported from the trials of IV t-PA: 6.4% in the National Institute of Neurological Disorders and Stroke study,1 8.8% in ECASS II,12 and 7.2% in ATLANTIS.13 The higher ICH rates in PROACT II and PROACT I (15%)6 reflect the occurrence of this complication in a well-defined, homogeneous group of patients with angiographically documented occlusion of the M1 or M2 segments of the MCA. Conversely, the IV t-PA studies included a more heterogeneous group of patients with ischemic stroke of any mechanism (large vessel, cardioembolic, or lacunar), some of which have a substantially lower risk of bleeding, in particular, those with a negligible propensity to develop secondary hemorrhagic transformation as part of their natural history such as infarcts of lacunar mechanism.16 Conversely, a high frequency of hemorrhagic transformation characterizes cardioembolic strokes,17 which were the predominant subtype in PROACT II and were recently reported to be a risk factor for post-IV t-PA symptomatic ICH.18 Furthermore, the higher rate of symptomatic ICH in the PROACT II trial might be expected from the greater baseline stroke severity as measured by the NIHSS score (medians of 17 in PROACT II7 compared with 11 in ECASS II12 and ATLANTIS,13 and 14 in the National Institute of Neurological Disorders and Stroke trial1). The baseline NIHSS score was found to relate significantly to ICH risk in the National Institute of Neurological Disorders and Stroke trial.4 An additional factor related to increased ICH risk in the two PROACT trials may be their longer treatment window (6 hours) compared with the 3 hours in the National Institute of Neurological Disorders and Stroke study.1 The early experience with IV t-PA suggested that longer treatment windows were significantly associated with higher rates of symptomatic ICH: in an angiographic study that had a therapeutic window of 8 hours,10 those subjects who developed hemorrhagic lesions were treated significantly later than those who did not develop intracranial hemorrhagic complications (6.1 ± 1.5 versus 5.3 ± 1.7 hours; p = 0.006).

The 12 symptomatic ICH in the PROACT II trial occurred in the area of the qualifying ischemic infarct, suggesting that local vascular factors, probably related to an abnormal permeability of the blood-brain barrier or to other mechanisms of ischemic vascular injury, played a role in their pathogenesis. This situation is clearly different from that of ICH that occurs in the setting of IV t-PA use in patients with acute myocardial infarction in whom the ICH occur randomly in the brain parenchyma and in the absence of acute abnormalities in the cerebral circulation.19-21⇓⇓ The use of thrombolysis after acute cerebral infarction, either IV or IA, is likely to lead to bleeding as a result of acute changes in the local vasculature. The phenomenon of reperfusion hemorrhage as a result of clot lysis and recirculation into an ischemic and abnormally permeable vascular bed, thought to be one of the mechanisms of hemorrhagic infarction after cerebral embolism,22 needs to be considered as a possible explanation for the observed ICH in the PROACT II trial. Although there was no association shown between degree of post-r-proUK angiographic recanalization (at 120 minutes from onset of treatment) and risk of ICH (see table 1), the possibility of delayed (after 120 minutes) recanalization with ICH cannot be dismissed because no recanalization data were obtained past that time point. It is also possible that the ICH observed after thrombolysis do not correspond to the mechanism of clot lysis–related reperfusion,23 but rather that they develop as a result of a different mechanism of reperfusion—that which occurs via retrograde collaterals into the area of cerebral ischemia, a mechanism of hemorrhagic transformation that does not require the proximal clot to be lysed to occur.24 The issue of collaterals to the MCA territory in the baseline angiogram in the PROACT II trial is currently under investigation. A role of the use of IV heparin in the pathogenesis of these ICH is possible because the immediate use of full IV heparinization was associated with a significant increase in the rate of ICH in PROACT I,6 as well as in a study of IV t-PA plus heparin in acute carotid territory stroke.25 However, a heparin effect is unlikely because the level of anticoagulation used in PROACT II for the purpose of preventing arterial reocclusion was subtherapeutic, without prolongation of the aPTT beyond the normal range and with the same low and fixed heparin dose for a period of 4 hours in both r-proUK–treated patients and controls.

The risk factors for symptomatic ICH after thrombolytic treatment of acute ischemic stroke have been studied after use of IV t-PA,4,25,26⇓⇓ IV streptokinase,27 and IA r-proUK.6 In the IV studies,4,25-27⇓⇓⇓ a consistent27 risk factor for symptomatic ICH was the presence of signs of a large infarct on baseline CT, whereas the severity of the neurologic deficit at entry was also correlated with ICH risk in three of these four studies.4,25,26⇓⇓ Other factors encountered in some studies, but not in others, included advanced age,26 diabetes,27 and use of high-dose heparin.6,25⇓ Most of these were not found to relate to increased risk of ICH in PROACT II. Although some, like large infarct volume on CT and higher NIHSS scores, showed a trend indicative of increased associated ICH risk, their values did not reach significance, almost certainly reflecting the small numbers of cases included in the analyses. It is possible that a larger data set may have established a significant correlation between these factors and increased risk of symptomatic ICH.

The only factor that reached nominal significance in its association with ICH risk in our exploratory analysis was baseline blood glucose level. However, because we examined 24 potential prognostic variables, even a nominal p value of 0.014 should be judged as weak evidence of association. The presence of diabetes or hyperglycemia, however, has been correlated with ICH risk in several other studies. In the National Institute of Neurological Disorders and Stroke trial, admission glucose >300 mg/dL was associated with increased ICH risk in univariate analysis, but this variable was not significantly associated with symptomatic ICH in the multivariate analysis.4 Another study27 documented a role for history of diabetes (not further defined) in a multivariate analysis of baseline variables, with an OR of 3.0 (95% CI, 1.1 to 7.9) and p = 0.03. Recently reported experience with the use of IV t-PA in the community3 showed that subjects with symptomatic ICH had higher baseline blood glucose levels than patients without ICH (198.6 ± 85.1 versus 150.5 ± 61.1 mg/dL; p = 0.03). Finally, another study5 found baseline serum glucose as an independent predictor of symptomatic ICH in a group of 13 patients who developed symptomatic ICH after treatment with IV t-PA. Both diabetes and elevated baseline serum glucose were independent predictors of increased ICH risk, with an observed 25% rate of ICH with serum glucose >11.1 mmol/L (equivalent to approximately 200 mg/dL).5 These authors found an OR of 1.008 (95% CI, 1.000 to 1.015) for symptomatic ICH by serum glucose increases of 1.0 mg/dL, which is similar to the one calculated for our PROACT II data (OR, 1.013; 95% CI, 1.003 to 1.023). Elevated baseline serum glucose has been established as a factor predictive of poor prognosis independently from stroke severity.28 We confirmed this by finding that the strength of the glucose association in our data was unaffected by adjusting for baseline NIHSS scores, thus suggesting that serum glucose appears to act as an independent influence on the risk of symptomatic ICH and not merely as a correlate of stroke severity.

The apparent association between baseline hyperglycemia and increased ICH risk has biologic plausibility that is supported by experimental data. Hyperglycemia in experimental stroke models has been shown to increase the damage to the blood-brain barrier29 and to increase the hemorrhagic conversion rate and hemorrhage size after reperfusion.30,31⇓ The mechanisms by which hyperglycemia exerts its detrimental effects in acute stroke are probably multiple but are likely centered on its effects at the level of the microcirculation, causing injury that in turn promotes ischemia32 and enhances the effects of the local ischemic injury,31 possibly promoting vascular rupture in the event of reperfusion.

The failure of this exploratory analysis to clearly identify other factors that influence the risk of symptomatic ICH does not necessarily mean that none truly exists. PROACT II was a relatively small study and thus could have failed to detect one or more other important variables. The importance of baseline hyperglycemia as a predictor of increased risk of symptomatic ICH and its potential impact in the management of patients with acute ischemic stroke needs to be further evaluated in larger databases studied prospectively in groups of patients treated with IV or IA thrombolytic agents. The confirmation of this observation could then lead to a more active approach to the management of hyperglycemia before the administration of thrombolytic agents.