Markers of Increased Risk of Intracerebral Hemorrhage After Intravenous Recombinant Tissue Plasminogen Activator Therapy for Acute Ischemic Stroke in Clinical Practice
The Multicenter rt-PA Acute Stroke Survey
Background— Intravenous recombinant tissue plasminogen activator (rtPA) is an effective therapy for acute ischemic stroke, but it is associated with risk of intracerebral hemorrhage (ICH). Our aim was to identify, in a large cohort of patients, readily available baseline factors that are associated with thrombolysis-related ICH.
Methods and Results— In a multicenter retrospective and prospective investigation of individual data from 1205 patients treated in routine clinical practice with intravenous rtPA within 3 hours of stroke symptom onset, 72 patients (6%) developed symptomatic ICH and 86 additional patients (7%) had asymptomatic ICH identified on a routine follow-up CT. In analyses based on clinical variables alone, the main attributes associated with ICH were a history of diabetes mellitus and cardiac disease, increasing stroke severity, advancing age, use of antiplatelet agents other than aspirin before stroke onset, and elevated pretreatment mean blood pressure. In additional analyses that incorporated baseline CT and laboratory findings (in a subset of patients), the main associations were early ischemic CT changes, in particular if exceeding one third of middle cerebral artery territory; increasing stroke severity; diabetes mellitus or elevated serum glucose; and lower platelet counts. Final independent attributes associated with parenchymatous hematoma, defined by purely radiologically based criteria, were similar to those of symptomatic ICH.
Conclusions— Readily available factors can identify acute ischemic stroke patients at high and low risk for rtPA-related ICH. These factors require confirmation in a prospective cohort before clinical implementation.
Received November 9, 2001; revision received January 29, 2002; accepted January 29, 2002.
Intravenous recombinant tissue plasminogen activator (rtPA) administered within 3 hours of symptom onset is an effective therapy for acute ischemic stroke.1,2⇓ Although data regarding the use of rtPA in clinical practice are somewhat limited, several postmarketing studies3–9⇓⇓⇓⇓⇓⇓ have demonstrated outcomes comparable to those of the pivotal National Institute of Neurological Disorders and Stroke (NINDS) rtPA Stroke Trial.1
At present, only a small fraction of potentially eligible stroke patients in the United States are receiving tPA therapy, and it is estimated that the rate of tPA use averages not more than 2%.10 The most critical risk of rtPA therapy, hampering widespread endorsement by clinicians, is symptomatic intracerebral hemorrhage (ICH). In the NINDS trial, 6.4% (20/312) patients developed symptomatic ICH within 36 hours of treatment, compared with 0.6% in the placebo group. In the European Cooperative Acute Stroke Study (ECASS-II), treatment within 6 hours of onset was associated with an 8.8% risk of symptomatic ICH, compared with 3.4% in the placebo arm.11
Identification of pretreatment predictors of ICH may improve patient selection and may be useful for individualized counseling of patients and families. Initial stroke severity, older age, heart disease, high blood pressure, and early CT abnormalities were reported to predict thrombolysis-related ICH after stroke, but findings have been inconsistent, 12–15⇓⇓⇓ and analyses were limited by insufficient statistical power.16
In contrast to randomized clinical trials, the present data are derived from patients treated in broad-based real-life clinical practice and thus may be more generalizable. Our aim was to identify readily available factors associated with rtPA-related ICH in a large cohort of acute ischemic stroke patients treated in clinical practice.
Individual patient data were collected from multiple data sets comprising 1205 patients with acute ischemic stroke treated in clinical practice with intravenous rtPA within 3 hours of stroke onset, following the NINDS rt-PA Stroke Trial protocol. Fifty-six centers/groups following the NINDS rt-PA Stroke Trial protocol were invited to participate; 2 declined. Centers had experience with stroke thrombolysis. Some centers have previously reported their clinical experience but could not study associations of ICH because of lack of statistical power. Data sources included the Standard Treatment with Alteplase to Reverse Stroke (STARS) study (296 patients)6; the Minnesota Stroke Treatment in the Community (STIC) study (151)7; Cologne, Germany (132)3; 2 centers in Calgary, Canada (60)5; and 49 centers in the United States (566) (including data from Houston, Tex ,8 a US multicenter survey ,4 and the OSF stroke network 9).
Investigators at each center reviewed records of consecutive patients treated with rtPA from the time of the publication of the NINDS rt-PA Stroke Trial results until the end of 1998. A standard data collection instrument with written instructions was used. We collected data on variables that, on the basis of previous thrombolysis data (in ischemic stroke or myocardial infarction) and/or biological plausibility, were potentially related to ICH or outcome. Head CTs were read by a qualified investigator at each center (except in the STIC data set, in which all CTs were read centrally) who was aware of the clinical presentation but not of outcome and presence/absence of ICH. Early CT findings were assessed and categorized as ischemic changes involving more or less than one third of the middle cerebral artery (MCA) territory. Time from stroke onset to treatment was categorized as <1 hour, 1 to 2 hours, 2 to 3 hours, or >3 hours. Centers/groups were categorized into 5 groups (STARS, STIC, Texas, Cologne, and other centers) to assess the effect of site. Pretreatment National Institute of Health stroke scale (NIHSS) scores,17 measuring neurological impairment, were categorized in 5-point intervals (≤5, 6 to 10, 11 to 15, 16 to 20, or >20) and were extracted from the medical records unless specifically documented.18
ICH was categorized primarily as symptomatic versus asymptomatic, in accord with the NINDS trial and the postmarketing prospective STARS trial. Symptomatic ICH was defined as a CT-documented hemorrhage temporally related to clinical deterioration as judged by the treating physician. Asymptomatic ICH was defined as any CT-documented hemorrhage identified on routine follow-up without clinical deterioration. In addition, ICH was categorized by a radiologically based approach as either parenchymatous hematoma (PH) or hemorrhagic infarction. PH was defined, according to the NINDS trial criteria, as findings of a typical homogeneous, hyperdense lesion with a sharp border with or without edema or mass effect. This hyperdense lesion could arise at a site remote from the vascular territory of the ischemic stroke or within but not necessarily limited to the territory of the presenting cerebral infarction. Hemorrhage with an intraventricular extension was considered a PH. Hemorrhagic infarction was defined as findings of acute infarction with punctuate or variable hypodensity/hyperdensity with an indistinct border within the vascular territory suggested by the acute neurological signs and symptoms. Follow-up CTs were performed at the discretion of the treating physician and recorded when performed within 36 hours.
Initial bivariable comparisons were performed among 3 groups: symptomatic ICH, asymptomatic ICH, and no ICH. Multivarible analyses were performed using logistic regression models. Pretreatment variables that were significant with P≤0.2 in bivariable analyses were analyzed in multivariable models and adjusted for age and sex. Two models were performed for each end point: the first incorporated baseline demographic and clinical variables; others also incorporated results of relevant laboratory and CT attributes (collected in a subset of the cohort). Goodness of fit was evaluated by the Hosmer-Lemeshow statistic, and discriminatory ability was evaluated using the C statistic, or area under a receiver-operator characteristic curve.
The study was powered to identify as many as 10 important variables that were predictive of symptomatic ICH. Therefore, the sample required ≈50 to 100 patients with this outcome in the study.16 Because symptomatic ICH was expected to occur in ≈6% of patients, the requisite sample size was 800 to 1600 patients and was achieved via recruitment of active centers and existing databases.
Of 1205 patients with acute ischemic stroke treated with intravenous rtPA, 72 patients (6%) had symptomatic ICH within 36 hours and 86 additional patients (7%) had asymptomatic ICH identified on a routine follow-up CT scan. Follow-up CT was not performed in 184 patients (15%) without evidence of neurological deterioration who were assumed not to have ICH. Mortality rates among these patients without follow-up CT (7%) did not differ from those without hemorrhage identified on follow-up CT (10%).
Patients with ICH more often had diabetes mellitus, atrial fibrillation, or other cardiac disease (Table 1). They more often used antiplatelet agents before stroke and had higher pretreatment blood pressure. Increasing age and stroke severity were associated with higher risk of ICH (Figure 1). A particularly low rate of symptomatic ICH was noted in patients <60 years old (2%) or with NIHSS ≤10 (3%). Early ischemic changes on baseline CT were more often observed in patients with ICH. Hyperdense artery sign was noted more often in patients with asymptomatic but not symptomatic ICH. The rate of symptomatic ICH was more than doubled in patients with minor (<33% of MCA territory) early ischemic changes and more than 4-fold higher in patients with major (>33%) changes compared with those without early ischemic changes. Higher levels of serum glucose and lower platelet counts were associated with increasing rate of ICH (Figure 2).
The associations of ICH derived from logistic regression models are presented in Tables 2 and 3⇓. In analyses based solely on clinical variables, significant associations of symptomatic ICH were older age, history of other cardiac disease, diabetes mellitus, and use of antiplatelets other than aspirin before stroke onset. In an analysis that also adjusted for site, categorized into 5 groups, similar associations emerged, including site as an additional independent attribute (P=0.02). Associations of all ICH were increasing stroke severity, elevated pretreatment blood pressure, history of diabetes mellitus, and atrial fibrillation. The goodness-of-fit statistic (P=0.20) indicated a good fit of this model, and the C statistic was 0.64.
In separate analyses incorporating CT and laboratory data into the models, these variables tended to take precedence over the clinical data. Thus, in a model evaluating associations of symptomatic ICH, only diabetes mellitus remained significant, whereas lower platelet counts and early CT abnormalities became important indicators of risk. Similarly for all ICH, independent associations were baseline stroke severity, higher glucose levels, lower platelet counts, and presence of early ischemic CT changes. Early ischemic CT changes >33% of the MCA territory were the most robust attribute. The goodness-of-fit statistic (P=0.35) indicated a good fit of the model, and the C statistic was 0.71.
Because some variables were not collected by all groups contributing to this study, the analyses of CT and laboratory data were limited to a subset of patients. To determine whether uncollected data were likely to affect the results, an additional logistic regression model for the end point of all ICH was performed that included only clinical variables among the subset of patients for whom all CT and laboratory variables were present. This analysis demonstrated results similar to those obtained from the complete cohort, suggesting that patients for whom no laboratory and CT data were collected were comparable to the overall population.
Additional analyses were performed to study PH (69 cases of 900 for whom data on this outcome were available). In analyses based on clinical variables, PH was significantly associated with atrial fibrillation (OR 2.28, 95% CI 1.21 to 4.32), diabetes mellitus (OR 2.20, 95% CI 1.15 to 4.18), history of hypertension (OR 2.28, 95% CI 1.21 to 4.32), male sex (OR 2.29, 95% CI 1.20 to 4.35), use of antiplatelets other than aspirin before stroke onset (OR 5.01, 95% CI 1.49 to 17.21), and NIHSS score (OR 1.35 per 1 category increase, 95% CI 1.05 to 1.73). Site had borderline statistical significance (P=0.048). Separate analyses incorporated CT and laboratory data into the model. In a final model that used the significant attributes, the independent variables and odds ratios were diabetes mellitus (OR 2.69, 95% CI 1.28 to 5.68), platelet counts (OR 0.67 per 50 000 increase, 95% CI 0.50 to 0.90), early CT changes <33% of MCA (OR 3.17, 95% CI 1.42 to 7.04), and >33% of MCA (OR 9.38, 95% CI 3.68 to 23.90).
Outcome of Patients
Mortality among patients with symptomatic ICH, asymptomatic ICH, and no ICH were 60%, 17%, and 10%, respectively; severe disability rates (modified Rankin scale 4 to 5) were 32%, 29%, and 26%, respectively; and excellent outcome rates (modified Rankin scale 1 to 2) were 4%, 21%, and 34%, respectively (P=0.001).
The present study assists in risk prognostication for rtPA-related ICH based on clinical information, laboratory data, and head CT. Although these factors appear to influence the probability of ICH, their presence or absence does not necessarily reflect which patients are more or less likely to benefit from rtPA. In this regard, secondary analysis of the NINDS rt-PA Stroke Trial failed to identify independent factors that altered response to rtPA therapy, other than time to treatment.19,20⇓
This study corroborates previous findings8 suggesting an association between baseline serum glucose and/or diabetes mellitus with ICH after rtPA. Marked hyperglycemia and diabetes mellitus produce damaging effects on the microvasculature that may result in increased edema and hemorrhagic transformation after reperfusion. Early ischemic changes on CT, especially if extensive, were strongly associated with ICH, and in particular PH. These ischemic changes represent early cytotoxic edema, extended hypoperfusion, and possibly the development of irreversible injury.21 The frequency of early ischemic changes (31%) found in the present study is comparable to that in clinical trials. The interrater reliability of identifying early CT changes and their extent is only fair.22 Therefore, although our results demonstrate the important predictive role of early CT changes as read by physicians in clinical practice, our findings are limited by a lack of centralized standard CT reading.
Borderline-low platelet counts were associated with a higher rate of ICH. In addition, patients taking antiplatelet agents other than aspirin, mainly ticlopidine, were at high risk of symptomatic ICH or PH. Relatively few patients were taking these antiplatelet agents, and thus, these finding are preliminary and should be interpreted with caution. Borderline-low platelet counts and/or decreased platelet function may be associated with greater risk of ICH in the setting of a compromised hemostatic system after rtPA therapy. Platelets play a pivotal role in mediating the activity of the plasma fibrinolytic system, whereas thrombolytic agents have reciprocal effects on platelet function. The role of platelets and antiplatelet agents deserves close scrutiny in future studies of thrombolysis for acute stroke.
Initial stroke severity assessed by the NIHSS score was found to be an independent marker of subsequent ICH, as previously shown.12 Finally, site-related variability was observed, and this finding suggests the importance of other factors possibly related to ICH risk, such as expertise, experience, protocol adherence, or other unknown elements.
Data were derived in part from pooling postmarketing studies and are thus limited by methodological differences among the studies. Protocol violations may have occurred, and our data reflect these potential errors as part of the reality of clinical practice.
Most centers had previous experience with acute stroke trials and thrombolysis and therefore may not represent all hospital types. Data included all consecutive patients treated with rtPA, but they were obtained by voluntary reporting without systematic auditing. Because study procedures were not prospectively mandated (eg, follow-up CT), the incidence of some events such as asymptomatic ICH may be underestimated. The data were therefore not used to assess absolute risk, but rather to identify associations. We categorized ICH primarily as symptomatic or not as defined in the NINDS trial. Clinical deterioration as judged by the treating physician, however, could be part of the natural course of a stroke with evolving edema, independent of the coinciding hemorrhage, especially if hemorrhagic transformation is only a minor part of a large infarct. Indeed, recent post hoc analysis from ECASS II revealed that only large PH independently impaired prognosis.23 Markers of subsequent PH in our analysis, however, were comparable to those of symptomatic ICH. Laboratory and CT data were collected only for a subset of patients. Analyses showed little difference between patients with and without these data, however, and also little effect on multivariable models. Analyses in this study were exploratory and data-driven and should therefore be considered to be primarily hypothesis-generating, requiring validation in future studies.
In conclusion, key clinical variables, laboratory results, and early ischemic CT changes are associated with development of rtPA-related ICH in routine clinical practice. At present, the factors identified should not dissuade physicians from treating patients according to accepted treatment guidelines. These associations may be indicators of higher risk situations rather than features causally related to ICH. They may provide clues about the pathophysiology of rtPA-related ICH that should be evaluated prospectively in future studies of the merits of thrombolytic therapy.
The following data sets, institutions, and investigators contributed to this survey (number of patients treated in parentheses, in decreasing order): STARS data set: S.A. Hamilton (296); STIC data set (23 hospitals): S.K. Hanson (151); Klinik für Neurologie der Universität zu Köln, Cologne, Germany: M. Grond (132); Hermann, St Luke’s Episcopal, Southwest Memorial, and Northwest Memorial Hospitals, Houston, Tex: J.C. Grotta, L. Morgenstern, D. Krieger, S.E. Kasner, C. Villar, D. Chiu, T. Wein, S. Hickenbottom, R. Felberg, A.M. Demchuk (131); Foothills Medical Center, Calgary, Canada: A.M. Buchan, A.M. Demchuk, P. Barber, N. Newcommon (60); Dent Neurological Institute, Buffalo, NY: V.E. Bates, K. Vereczkey-Porter (50); University of Pennsylvania Medical Center, Philadelphia, Pa: S.E. Kasner, J.C. Chalela, M.L. McGarvey (35); OSF Stroke Network, St Francis Medical Center, Peoria, Ill: D.Z. Wang (35); UPMC Stroke Institute, Pittsburgh, Pa: J. DeCampo, L.R. Wechsler (30); Medical College of Wisconsin, Milwaukee, Wis: J.R. Binder, D. Book (27); Henry Ford Hospital, Detroit, Mich: D. Tanne, S. Daley, M.J. Gorman, P. Mitsias, C.A. Lewandowski, S.R. Levine (26); Barrow Neurological Institute, Phoenix, Ariz: J.L. Frey, H. Jahnke (26); University of South Alabama, Mobile: R. Zweifler, N. Koscicki (26); Souers Stroke Institute, St Louis, Mo: S. Cruz-Flores (23); Seton-Hall University, East Orange, NJ: P. Verro (20); Stroke Treatment and Research Center, Medical College of Ohio, Toledo: G. Tiejen (18); Washington University School of Medicine, St Louis, Mo: J.M. Lee, T. Lowenkopf (16); University of Virginia Health Sciences Center, Charlottseville: B.B. Worrall, K.C. Johnston (14); Yale University School of Medicine, New Haven, Conn: P. Fayad, J. Boiser (14); St John Hospital, Detroit, Mich: T. Giancarlo, A. Schuster (11); Harborview Medical Center, Seattle, Wash: K.J. Becker (10); Marshfield Clinic, Marshfield, Wis: P.N. Karanjia (10); University of Alberta Hospital, Edmonton, Canada: A. Shuaib, P. Kochanski (8); University of Michigan Medical Center, Ann Arbor, Mich: P. Scott (8); Sparrow Hospital, Lansing, Mich: S. Wehner (6); Thomas Jefferson University Hospital, Philadelphia, Pa: J.M. Dayno, R. Bell (5); University of Rochester, Rochester, NY: C. Benesch (5); Wayne State University, Detroit, Mich: W.M. Coplin, S. Chaturvedi (5); University of Wisconsin, Madison, Wis: D. Dulli (4); UCLA School of Medicine, Los Angeles, Calif: M.A. Kalafut, C.S. Kidwell, J.L. Saver (3). Coordinated at Henry Ford Hospital and Health Science Center: D. Tanne, S. Daley, L. Salowich-Palm, L.R. Schultz, L. D’Olhaberriague, S.R. Levine.
This study was supported by National Institutes of Health (NIH) grant K23-NS02147 (Dr Kasner), NIH training grant T32-NS07412 to the University of Texas–Houston Medical School, the Canadian Institutes of Health, the Alberta Heritage Foundation for Medical Research (Dr Demchuk), an American Heart Association Grant-in-Aid (Dr Hanson), and NIH grant R01-NS30896 (Dr Levine).
↵*Participants in the Multicenter rt-PA Stroke Survey Group are given in the Appendix.
- ↵Adams HP, Brott TG, Furlan AJ, et al. Guidelines for thrombolytic therapy for acute stroke: a supplement to the guidelines for the management of patients with acute ischemic stroke: a statement for healthcare professionals from a special writing group of the Stroke Council, American Heart Association. Stroke. 1996; 27: 1711–1718.
- ↵Schmülling S, Grond M, Rudolf J, et al. One-year follow-up in acute stroke patients treated with rt-PA in clinical practice. Stroke. 2000; 31: 1552–1554.
- ↵Tanne D, Bates VE, Verro P, et al. Initial clinical experience with intravenous t-PA for acute ischemic stroke: a multicenter survey. Neurology. 1999; 53: 424–427.
- ↵Buchan AM, Barber PA, Newcommon NJ, et al. Effectiveness of t-PA in acute ischemic stroke: outcome relates to appropriateness. Neurology. 1999; 54: 679–674.
- ↵Davenport J, Hanson SK, Altafullah IM, et al. tPA: a rural network experience (letter). Stroke. 2000; 31: 1457.
- ↵Demchuk AM, Morgenstern LB, Krieger DW, et al. Serum glucose level and diabetes predict tissue plasminogen activator–related intracerebral hemorrhage in acute ischemic stroke. Stroke. 1999; 30: 34–39.
- ↵Wang DZ, Rose JA, Honings DS, et al. Treating acute stroke patients with intravenous tPA: the OSF Stroke Network experience. Stroke. 31: 77–81.
- ↵The NINDS rt-PA Stroke Study Group. Intracerebral hemorrhage after intravenous t-PA therapy for ischemic stroke. Stroke. 1997; 28: 2109–2118.
- ↵Larrue V, von Kummer R, del Zoppo G, et al. Hemorrhagic transformation in acute ischemic stroke: potential contributing factors in the European Cooperative Acute Stroke Study. Stroke. 1997; 28: 957–960.
- ↵Jaillard A, Cornu C, Durieux A, et al. Hemorrhagic transformation in acute ischemic stroke: the MAST-E Study. Stroke . 1999; 30: 1326–1332.
- ↵Larrue V, von Kummer R, Muller A, et al. Risk factors for severe hemorrhagic transformation in ischemic stroke patients treated with recombinant tissue plasminogen activator: a secondary analysis of the European-Australasian Acute Stroke Study (ECASS II). Stroke. 2001; 32: 438–441.
- ↵Brott T, Adams HP, Olinger CP, et al. Measurements of acute cerebral infarction: a clinical examination scale. Stroke. 1989; 20: 864–870.
- ↵Kasner SE, Chalela JA, Luciano JM, et al. Reliability and validity of estimating the NIH Stroke Scale from medical records. Stroke. 1999; 30: 1534–1537.
- ↵The NINDS tPA Stroke Study Group. Generalized efficacy of tPA for acute stroke: subgroup analysis of the NINDS tPA Stroke Trial. Stroke. 1997; 28: 2119–2125.
- ↵Marler JR, Tilley BC, Lu M, et al. Early stroke treatment associated with better outcome. The NINDS rt-PA Stroke Study. Neurology. 2000; 55: 1649–1655.
- ↵Grond M, von Kummer R, Sobesky J, et al. Early x-ray hypoattenuation of brain parenchyma indicates extended critical hypoperfusion in acute stroke. Stroke. 2000; 31: 133–139.
- ↵Grotta JC, Chiu D, Lu M, et al. Agreement and variability in the interpretation of early CT changes in stroke patients qualifying for intravenous rtPA therapy. Stroke. 1999; 30: 1528–1533.
- ↵Berger C, Fiorelli M, Steiner T, et al. Hemorrhagic transformation of ischemic brain tissue: asymptomatic or symptomatic? Stroke. 2001; 32: 1330–1335.