Cerebrovascular Disease Center, Centre de Recherche et d'Applications en Traitement de l'Image et du Signal (CREATIS) UMR Centre National de la recherche scientifique (CNRS) 5515, Institut National de la Sante et de la Recherche Médicale U 630, University of Lyon, Lyon, France
Members of the MR Stroke Study Group are listed in the Appendix on page xx.
The risk for symptomatic intracerebral hemorrhage (sICH) associated with thrombolytic treatment has not been evaluated in large studies using diffusion-weighted imaging (DWI). Here, we investigated the relation between pretreatment DWI lesion size and the risk for sICH after thrombolysis.
In this retrospective multicenter study, prospectively collected data from 645 patients with anterior circulation stroke treated with intravenous or intraarterial thrombolysis within 6 hours (<3 hours: n = 320) after symptom onset were pooled. Patients were categorized according to the pretreatment DWI lesion size into three prespecified groups: small (≤10ml; n = 218), moderate (10–100ml; n = 371), and large (>100ml; n = 56) DWI lesions.
In total, 44 (6.8%) patients experienced development of sICH. The sICH rate was significantly different between subgroups: 2.8, 7.8, and 16.1% in patients with small, moderate, and large DWI lesions, respectively (p < 0.05). This translates to a 5.8 (2.8)-fold greater sICH risk for patients with large DWI lesions as compared with patients with small (or moderate) DWI lesions. The results were similar in the large subgroup (n = 536) of patients treated with intravenous tissue plasminogen activator. DWI lesion size remained an independent risk factor when including National Institutes of Health Stroke Scale, age, time to thrombolysis, and leukoariosis in a logistic regression analysis.
This multicenter study provides estimates of sICH risk in potential candidates for thrombolysis. The sICH risk increases gradually with increasing DWI lesion size, indicating that the potential benefit of therapy needs to be balanced carefully against the risk for sICH, especially in patients with large DWI lesions. Ann Neurol 2007
Symptomatic intracerebral hemorrhage (sICH) is the main complication of recombinant tissue plasminogen activator (tPA) therapy in acute stroke.1 The identification and possible exclusion of patients at high risk for sICH would be of major importance to reduce the overall complication rate and to increase efficacy of this treatment.
Several clinical and laboratory parameters have been suggested as markers of an increased risk for sICH.2–4 In addition, early ischemic changes (EICs) on computed tomography (CT) are associated with an increased risk for sICH, especially if exceeding one third of the middle cerebral artery (MCA) territory.2, 5, 6 However, it is still a matter of debate whether the presence of extensive EICs justifies exclusion from thrombolysis.7 Moreover, EICs are often subtle, particularly in the first hours, resulting in poor interrater reliability.8, 9
Recently, the use of diffusion- (DWI) and perfusion-weighted magnetic resonance imaging has been shown to allow the identification of patients with lesion patterns associated with an improved response to recanalization.10 In addition, small magnetic resonance imaging (MRI) and CT studies suggest that parameters indicating severe ischemia (low apparent diffusion coefficient [ADC] or severe perfusion deficit) could be useful markers of sICH risk, but the role of MRI in defining patients at high risk for sICH has not been established in larger studies.11–15 In particular, DWI lesion size has not been evaluated in sufficiently powered studies as a potential risk factor for sICH.
The aim of this large multicenter analysis was, therefore, to evaluate the relation between pretreatment DWI lesion size and sICH risk after thrombolysis in patients with acute stroke.
Patients and Methods
This retrospective multicenter analysis was performed based on prospectively collected data provided by 10 well-established academic stroke centers (Barcelona, Frankfurt, Hamburg, Jena, University of California Los Angeles, Lyon, Mannheim, Paris, Stanford [for the DEFUSE investigators10], Seoul). The study was conducted as an initiative of the MR Stroke Study group, which is an international group of stroke physicians who meet regularly with the aim to coordinate and standardize MRI research in acute stroke. Each of the centers participating in this study uses standardized MRI protocols for acute stroke patients, approved by all local ethics committees.
Inclusion criteria for patients in this analysis were as follows: (1) patients presented with anterior circulation stroke, (2) thrombolytic therapy was administered within 6 hours of symptom onset, and (3) MRI including DWI was performed before the initiation of treatment. Symptomatic ICH was defined, as in the National Institute of Neurological Disorders and Stroke-tPA trial, as CT (or MRI)-documented hemorrhage that occurred within 36 hours from treatment onset and was temporally related to deterioration in the patient's clinical condition in the judgment of the clinician, associated with a worsening of at least one point on the National Institutes of Health Stroke Scale (NIHSS).1
Central reading was performed of follow-up CT or MRI scans of sICH patients for classification of hemorrhage according to European-Australasian Acute Stroke Study criteria.16 In addition, ICH volume was estimated using the ABC/2 formula.17 The following clinical variables that have been associated with sICH were obtained for each patient: age, NIHSS score on admission, and time between symptom onset and thrombolytic treatment (categorized as <3 hours and 3–6 hours). Furthermore, the NIHSS at the time of clinical deterioration or on routine follow-up (days 2–7) was obtained. An NIHSS score of 42 points was assigned to patients who died within the follow-up period. Most patients were treated with standard intravenous (IV) tPA therapy (n = 536). Two smaller subgroups were treated with either intraarterial thrombolysis (IA; n = 74) or combined intravenous/intraarterial (IV+IA) treatment (n = 35).
Magnetic Resonance Imaging Data Acquisition and Analysis
MRI scans were performed on 1.5-Tesla scanners, all equipped with echo planar imaging data acquisition capabilities. Stroke protocols were not entirely uniform in the 10 participating centers, but all included an echo planar DWI sequence. There were no predefined MRI patterns (such as DWI/perfusion-weighted imaging mismatch), which were required for inclusion in the study.
Pretreatment DWI lesion volumes were measured by the participating stroke centers using locally available software. The areas of hyperintensity on diffusion-weighted images were manually traced on each slice, summed, and multiplied with the slice thickness and interslice gap to obtain the DWI lesion volume.18 To obtain an estimate of the interrater reliability, two blinded investigators (O.C.S., M.C.H.) reanalyzed a subset of pretreatment diffusion-weighted images (n = 78 patients from 7 centers; between 7 and 17 data sets per center, according to the proportion of data sets contributed to the study). Lesion volumes were compared with the lesion volumes measured by the local investigators. The overall intraclass correlation coefficient was 0.968 (95% confidence interval, 0.951–0.980; p < 0.001).
Patients were categorized into 3 prespecified groups according to the pretreatment DWI lesion size: a small DWI lesion group (≤10ml, n = 218), a moderate DWI lesion group (10–100ml, n = 371), and a large DWI lesion group (>100ml, n = 56). The cutoff of 100ml was chosen because this translates roughly into an involvement of one-third of the MCA territory, a commonly used criterion in CT-based studies for assessment of extensive EICs.2, 19, 20 The lower cutoff (≤10ml) was arbitrarily chosen to distinguish between patients with small and moderate DWI lesions. Because DWI lesions were not normally distributed (predominance of small and moderate DWI lesions) and sample size was not balanced within the 3 prespecified groups, we additionally performed a separate analysis using quintiles (n = 128–130 patients in each group; median DWI lesion [minimum to maximum range]): group 1, 3ml (0–6); group 2, 9ml (6–13); group 3, 18ml (13–26); group 4, 37ml (26–55); and group 5, 95ml (55–382).
Data on the extent of leukoariosis, using the scale that Fazekas and coworkers21 developed, was available in 346 patients as part of a previously published study.22
As the variables age, initial NIHSS and initial DWI volume were not normally distributed (Kolmogorov–Smirnov test); the Mann–Whitney U test was used to test for statistically significant differences between groups. If more than two groups were compared, the Kruskal–Wallis test was used before the Mann–Whitney U test. Using binary logistic regression analysis, we analyzed whether DWI lesion volume is an independent risk factor for sICH. Results were considered statistically significant at the 5% level. All values are given as median (25th, 75th percentile), unless otherwise stated.
Data from all 645 patients meeting the inclusion criteria were analyzed. Forty-four (6.8%) patients developed sICH; median ICH volume was 36ml (12, 67). Initial DWI lesion volume was 18ml (7, 45; see Fig 1 for examples of different lesion sizes). DWI lesion volume was significantly larger in patients experiencing subsequent sICH than in those without sICH (44 [15, 93] vs 18ml [7, 42]; p < 0.001). The risk for sICH was 2.8, 7.8, and 16.1% in the small, moderate, and large DWI lesion group, respectively (Fig 2). This translates to a 2.8-fold increase of sICH risk in the moderate DWI lesion group as compared with the small DWI lesion group and to a 5.8-fold increase of sICH risk in the large DWI lesion group as compared with the small DWI lesion group. Results were comparable when patients were grouped into quintiles: The sICH risk was 1.5% for the 20% of patients with the smallest DWI lesions (median DWI lesion volume, 3ml) and increased to 14% in the patients with the largest lesions (median DWI lesion volume, 95ml) (Fig 3).
On univariate analysis, initial NIHSS score and the presence of extensive leukoariosis were also significantly associated with a greater risk for sICH, whereas age and time to thrombolysis were not (Table 1). Similar sICH rates were seen in the subgroup of patients treated with IV tPA (n = 536; Table 2 and Figs 4 and 5). In this population, the overall sICH rate was 5.2% (n = 28), and it was 2.1, 6, and 12.8% in patients with small, moderate, and large DWI lesions, respectively.
In a multivariate model (n = 634), DWI lesion volume was an independent risk factor for sICH when controlling for age, initial NIHSS score, and time to thrombolytic treatment (odds ratio, 1.069 per 10ml increase in DWI lesion volume [95% confidence interval, 1.017–1.123]; p < 0.01). Results were similar when leukoariosis (n = 346) was added to the binary logistic regression model (Table 3). When the analysis was limited to patients treated with IV tPA, DWI lesion volume was again an independent risk factor for sICH when controlling for age, initial NIHSS, time to thrombolytic treatment (<3 vs 3–6 hours), and leukoariosis (odds ratio 1.098 per 10ml increase in DWI lesion volume [95% confidence interval, 1.010–1.194]; p < 0.05).
Table 3. Multivariate Analysis of Prediction of Symptomatic Intracerebral Hemorrhage (n = 346)
Of the 44 patients with sICH, 30 had radiological evidence of grade 2 parenchymal hematoma, 11 had grade 1 parenchymal hematoma, 2 had grade 2 hemorrhagic infarction, and 1 had grade 1 hemorrhagic infarction. Table 4 gives an overview of the influence of different definitions of sICH on the prediction of sICH by the initial DWI lesion volume. DWI lesion volume remains an independent risk factor for sICH even when the most strict definition of sICH (evidence of grade 2 parenchymal hematoma and NIHSS worsening of four or more points) is applied.23
Table 4. Different Definitions of Symptomatic Intracerebral Hemorrhage and Prediction of Symptomatic Intracerebral Hemorrhage by Initial Diffusion-Weighted Imaging Lesion Volume
Short-term outcome data are presented in Table 5. Patients with small (≤10ml) and moderate (10–100ml) DWI lesions improved significantly between symptom onset and short-term follow-up (days 2–7), whereas patients with large DWI lesions (>100ml) did not improve significantly on follow-up examination, even if patients with sICH were excluded from analysis. All patients with sICH had deteriorated at follow-up, regardless of pretreatment DWI lesion size.
Table 5. Short-Term Outcome Depending on Initial Diffusion-Weighted Imaging Lesion Size and Occurrence of Symptomatic Intracerebral Hemorrhage
Diffusion-weighted MRI lesions reflect areas of severe ischemic damage with disturbed energy metabolism.24 Our retrospective multicenter analysis of prospectively collected data suggests that the size of the DWI lesion is an independent factor determining the risk for sICH associated with thrombolytic treatment. There is a significant increase of sICH risk with increasing pretreatment DWI lesion size, from about 2% in patients with very small lesions to about 12 to 16% in patients with large DWI lesions involving more than one-third of the MCA territory (>100ml).
Only few large, sufficiently powered MRI studies have investigated markers of an increased risk for sICH after thrombolysis. The presence of severe leukoariosis was recently established as an independent risk factor for sICH, whereas the role of old microbleeds remains elusive.22, 25–27 Our data indicate that both leukoariosis and DWI lesion volume are independent predictors of sICH. In addition, our data demonstrate a trend toward increased sICH risk associated with greater baseline NIHSS scores. Taken together, the extent of both acute (DWI lesion) and chronic (leukoariosis) ischemic injury are likely to influence the rate of sICH after thrombolysis.
Our data are in line with the findings of several smaller sized studies examining predictors of ICH in patients treated less than 3,13 less than 7,28 or 3 to 6 hours29 from symptom onset with tPA: Average pretreatment DWI lesion volumes were greater in patients with subsequent (symptomatic and asymptomatic) ICH, but DWI lesion volume remained an independent predictor of ICH risk after correction for several clinical and imaging parameters in only one of these studies.29 Others reported that DWI lesion size is predictive for hemorrhagic transformation, but not for parenchymal hemorrhage or sICH.30 However, this finding has to be interpreted with caution because this study30 included only four patients with sICH. Further evidence comes from a large CT-based study indicating that patients with minor EICs of less than a third of the MCA territory have a 2.9-fold increased sICH risk as compared with patients without EICs, which increases up to 4-fold in patients with extensive EICs.2 Although the low interrater reliability of EIC detection on CT limits its use in clinical practice, improvements have been made by standardized CT scoring.7, 8, 31 In contrast, DWI is more sensitive to EICs,9, 32, 33 and interrater agreement of DWI lesion volumetry is excellent (intraclass correlation coefficient = 0.968 in this study).34
Smaller studies have investigated the prognostic value of the severity of the diffusion restriction within the DWI lesion using ADC measurements. Either the absolute number of voxels with severely reduced ADC13 or the percentage of ADC values below a given threshold in a certain region11, 35 has been reported to predict subsequent hemorrhagic transformation. The ADC has been shown to be significantly lower in areas of subsequent hemorrhagic transformation than in the remainder of the DWI lesion.15 Thus, in addition to performing simple measurements of DWI lesion size, focusing on areas with severe ADC reductions could further increase the prognostic power of ICH risk assessment; this, however, was not systematically analyzed in our study.
A recent multicenter open-label IV tPA stroke study (DEFUSE) identified a small subgroup of patients (n = 6) with a “malignant pattern” with either large DWI lesions or severe perfusion abnormalities exceeding 100ml.10 This pattern was associated with a high rate (50%) of reperfusion-associated brain hemorrhage after thrombolytic treatment. In our study, the rate of sICH in the subgroup with large lesions was substantially lower (12–16%) than in the DEFUSE study. In addition, we found a gradual increase in sICH risk with increasing DWI lesion size less than 100ml, indicating that there is no clear threshold differentiating between high and low risk. However, we did not find a further increase in the sICH rate in the group with DWI lesions exceeding 100ml, possibly because of the relatively small sample size in the large DWI lesion group (n = 56; sICH = 9). Risk assessment will thus need to take into account DWI lesion size not only in patients with very large lesions, but also in patients with moderate lesions.
The main limitation of our analysis is the retrospective nature of the study. However, data were prospectively collected in the single centers and the cutoffs for patient grouping (≤10ml, >100ml) were prespecified. Importantly, follow-up MRI was not required to prevent dropout because of bad clinical status. Secondly, no long-term follow-up was performed, prohibiting inferences on long-term harm of thrombolytic therapy in our sICH patients with large pretreatment lesions. Furthermore, measurements of DWI lesions were performed by the local investigators and not in a central imaging laboratory, possibly introducing heterogeneity. However, DWI lesion volumetry can be performed with high interrater reliability.34 There is an ongoing debate, which is the most appropriate outcome parameter for bleeding risk analysis,36, 37 and the sICH definition used in this study has been criticized for a lack of specificity, probably overestimating the rate of hemorrhage causing clinical deterioration (in contrast with deterioration caused by reocclusion or brain edema). However, the fact that the use of more strict sICH criteria23 (see Table 4) did not influence the main findings of this study argues for the robustness of our data and sICH risk assessment by the initial DWI lesion size. Several variables potentially influencing the ICH risk such as glucose level, platelet count, albuminuria, or pre-existent antiplatelet therapy were not included in our multivariate analysis because these parameters were not assessed systematically. Finally, CT before thrombolytic therapy was not required for this study. Therefore, no direct comparison between DWI lesion volumetry and detection of EICs on CT for sICH risk assessment can be derived from the present data.
In conclusion, this multicenter study provides estimates of sICH risk in potential candidates for thrombolysis. The sICH risk increases gradually with increasing DWI lesion size from 2% (small lesions) to 12 to 16% (lesions exceeding 100ml). This study was not designed to determine whether patients with large DWI lesions shall receive thrombolytic therapy. A controlled trial is needed to answer this issue properly. As this requires a large number of patients and is unlikely to be performed in the near future, this analysis, based on one of the largest data sets of patients imaged with MRI before thrombolysis published so far, may serve as a guide for clinicians in their assessment of sICH risk associated with thrombolytic treatment.
The study was initiated by the MR Stroke Study group without dedicated funding. Local data sampling was supported by different funding sources: German centers—Bundesministerium für Bildung und Wissenschaft within the Kompetenznetzwerk Schlaganfall (Stroke Imaging Net) (O.C.S., M.C.H., J.F., A.K., A.G., T.N.-H.); Stanford—NIH (National Institute for Neurological Disorders and Stroke, RO1 NS39325, K23 NS051372) (G.W.A., M.G.L.): Los Angeles—NIH (National Institutes for Neurological Disorders and Stroke, 5K23NS054084 5P50NS044378-03) (D.S.L.); Paris—Programme Hospital de Recherche Clinique (PHRC) (AOM 03 008 EVAL-USINV) (C.R.); Seoul—Brain Research Center of the 21st Century Frontier Research Program funded by the Ministry of Science and Technology of Korea (M103KV010010 06K2201 01010) (J.S.K.).
Members of the MR-Stroke Study Group are as follows: co-chairs: Geoffrey A. Donnan, Stephen M. Davis; contributing members: Germany—University Hospital Frankfurt (Tobias Neumann-Haefelin, Oliver C. Singer, Silke Hoelig, Matthias W. Lorenz, Marek C. Humpich, Matthias Sitzer, Helmuth Steinmetz, Bernard Yan, Joachim Berkefeld); University Hospital, Hamburg Eppendorf (Jens Fiehler, Joachim Roether, Thomas Kucinski, Hermann Zeumer, Goetz Thomalla); University Hospital, Jena (Andreas Kastrup); University Hospital, Mannheim (Achim Gass, Olivera Lecei); France—Salpêtrière, Paris (Yves Samson, Charlotte Rosso, Sandrine Deltour, Sophie Crozier, Anne Leger, Pr Jacques Chiras and his Neuroradiology team); Hopital Neurologique, Lyon (Laurent Derex, Norbert Nighoghoshian, Marc Hermier); Spain—Hospital Universitari Vall d'Hebron, Barcelona (Alex Rovira, Raquel Delgado, Carlos Molina, Jose Alvarez-Sabin); United States and Canada—University of California Los Angeles Stroke Center, University of California, Los Angeles (David S. Liebeskind, Jeffrey L. Saver, Jeffry R. Alger, Latisha K. Ali, Brian H. Buck, Gary R. Duckwiler, Reza Jahan, Doojin Kim, Bruce Ovbiagele, Noriko Salamon, Nerses Sanossian, Sidney Starkman, Paul M. Vespa, J. Pablo Villablanca, Fernando Viñuela); Stanford Stroke Center, Stanford University (Gregory W. Albers, Roland Bammer, Anna Caulfield Finley, Scott Hamilton, Wataru Kakuda, Stephanie Kemp, Maarten G. Lansberg, Michael Marks, Michael Moseley, Neil E. Schwartz, Vincent N. Thijs, Christine A.C. Wijman); UPMC Stroke Institute and Department of Neurology, University of Pittsburgh (Lawrence Wechsler); Beth Israel Deaconess Medical Center and Harvard Medical School, Boston (Gottfried Schlaug); Department of Neurology, University of Utah, Salt Lake City (Elaine Skalabrin); Division of Neurology, University of Alberta, Edmonton (William Coplin); Korea—Asian Medical Center, Seoul (Jong S. Kim, Dong-Wha Kang, Sun U. Kwon).