MRI Detection of Early Blood-Brain Barrier DisruptionParenchymal Enhancement Predicts Focal Hemorrhagic

Transformation After Thrombolysis

Niels Hjort, MD, PhD; Ona Wu, PhD; Mahmoud Ashkanian, MD; Christine Sølling, MD;Kim Mouridsen, PhD; Søren Christensen, MSc; Carsten Gyldensted, MD, PhD;

Grethe Andersen, MD, PhD; Leif Østergaard, MD, PhD

Background and Purpose—Blood-brain barrier disruption may be a predictor of hemorrhagic transformation (HT) inischemic stroke. We hypothesize that parenchymal enhancement (PE) on postcontrast T1-weighted MRI predicts andlocalizes subsequent HT.

Methods—In a prospective study, 33 tPA-treated stroke patients were imaged by perfusion-weighted imaging, T1 andFLAIR before thrombolytic therapy and after 2 and 24 hours.

Results—Postcontrast T1 PE was found in 5 of 32 patients (16%) 2 hours post-thrombolysis. All 5 patients subsequentlyshowed HT compared to 11 of 26 patients without PE (P�0.043, specificity 100%, sensitivity 31%), with exactanatomic colocation of PE and HT. Enhancement of cerebrospinal fluid on FLAIR was found in 4 other patients, 1 ofwhich developed HT. Local reperfusion was found in 4 of 5 patients with PE, whereas reperfusion was found in all casesof cerebrospinal fluid hyperintensity.

Conclusions—PE detected 2 hours after thrombolytic therapy predicts HT with high specificity. Contrast-enhanced MRImay provide a tool for studying HT and targeting future therapies to reduce risk of hemorrhagic complications. (Stroke.2008;39:1025-1028.)

Key Words: blood brain barrier � brain infarction � imaging � intracerebral hemorrhage � MRI � thrombolysis

Thrombolytic therapy of acute ischemic stroke increasesthe risk of hemorrhagic transformation (HT).1 A central

mechanism is thought to be reperfusion of brain tissues,where prolonged, severe hypoperfusion damages the blood-brain barrier (BBB).2 In stroke, increased BBB permeabilityis observed as hyperintensities by contrast-enhanced T1-weighted MRI in the parenchyma3 and fluid-attenuatedinversion-recovery (FLAIR) MRI in the cerebrospinal fluid(CSF), termed “hyperintense acute reperfusion marker”(HARM).4 By the dynamics of CSF, HARM may not besensitive to subcortical BBB damage. On the other hand,parenchymal enhancement (PE) seen as postcontrast hyper-intensity on T1-weighted imaging may provide a more directlocalization of BBB disruption relative to the diffuse hyper-intensity in HARM.

The purpose of this study was to compare the ability ofcontrast-enhanced T1 and FLAIR to predict HT in the firstfew hours after thrombolytic therapy. Secondly, we aimed tocharacterize the relation between early reperfusion, BBBdisruption, and HT.

Methods

PatientsAll patients undergoing MRI-based intravenous treatment withrecombinant tissue plasminogen activator (tPA) from April 2004 toApril 2006 at our unit were prospectively studied. The study wasapproved by the local Ethics Committee. Eligibility criteria includedthe National Institute of Neurological Disorders and Stroke criteria1

with the following modifications: (1) diffusion-weighted imaging(DWI) and PWI consistent with acute ischemic stroke, (2) DWIlesion volume comprising �50% of the territory supplied by themiddle cerebral artery (MCA). Perfusion-diffusion mismatch (ratio�1.2) was required if DWI lesion volume comprised 33% to 50% ofthe MCA territory.

ImagingWe used a 3.0T (26 of 33 patients) or 1.5T MRI scanner (SignaExcite/Horizon, GE, USA). The protocol consisted of DWI, T2*gradient recalled echo (GRE), FLAIR (TR/TE�8650/120 ms), T1(TR/TE�700/18 ms), and PWI. Follow-up MRI was performed 2and 24 hours post-treatment. Mean transit time (MTT) maps werecalculated5 and lesions on DWI and MTT maps were outlinedsemiautomatically. A neuroradiologist assessed all images blindedto clinical and other MRI data. Reperfusion was defined as a

Received July 4, 2007; accepted August 1, 2007.From the Center of Functionally Integrative Neuroscience (N.H., O.W., M.A., C.S., K.M., S.C., C.G., L.O.), Department of Neuroradiology, Århus

University Hospital, Århus C, Denmark; the Department of Neurology (N.H., G.A.), Århus University Hospital, Århus C, Denmark; the Athinoula AMartinos Center for Biomedical Imaging (O.W.), Department of Radiology, Massachusetts General Hospital, Charlestown, Mass.

Correspondence to Niels Hjort, MD, PhD, Center for Functionally Integrative Neuroscience, Århus University Hospital, Nørrebrogade 44, Building 30,8000 Århus C, Denmark. E-mail niels@pet.auh.dk

© 2008 American Heart Association, Inc.

Stroke is available at http://stroke.ahajournals.org DOI: 10.1161/STROKEAHA.107.497719

1025

decrease of the MTT lesion volume of �30% from baseline to thefollow-up study. Tissue with PE was characterized as showinglocal reperfusion if MTT values in the tissue volume hadnormalized after 2 and 24 hours (visual inspection). PostcontrastT1 PE had to be clearly distinct from vascular enhancement andnot present on baseline images. HARM was defined as FLAIRhyperintensity in the CSF.4 Hemorrhagic transformation wasclassified as either hemorrhagic infarct (HI) or parenchymalhematoma (PH).

ResultsOf 141 MRI-screened patients, 33 received tPA within 3hours of symptom onset (age 68�9 years, NIHSS 11�6,time-to-MRI 104�33 min). Mean DWI lesion volume was16 mL. Five patients (16%) displayed PE 2 hours post-tPA(Figure 1); hemorrhagic transformation occurred in those 5patients, compared with 11 of 26 without PE (P�0.043,specificity�100%, sensitivity�31%). Three of 16 cases ofHT were classified as PH. None of the hemorrhages weresymptomatic (NIHSS increase of �4 during 24 hours).

Local reperfusion was found in 4 of 5 cases of PE 2 hourspost-tPA (Figure 2). Reperfusion (�30% MTT lesion shrink-age) was found in 1 of 4 patients after 2 hours compared with10 of 19 without PE (P�0.59). There was no significantrelationship between reperfusion and subsequent HT (Table).Reperfusion 2 hours post-tPA was found in 7 of 14 patientswith subsequent HT. At 24 hours, local reperfusion hadoccurred in 12 of 13 patients with HT.

Four patients (12%) displayed HARM 2 hours post-tPA.After 24 hours, HARM was only found in those 4 patients;

early reperfusion had occurred in all, but subsequent HT wasonly found in one of them. None of the patients displayedboth PE and HARM at any time.

Predictors of Hemorrhagic TransformationLogistic regression showed correlation between baselineDWI lesion volume and subsequent HT (P�0.04). Mean

Figure 1. Magnetic resonance imagesobtained before, 2 hours, and 24 hours afterthrombolytic therapy. Evidence of BBB dis-ruption is found 2 hours after treatment initia-tion in the basal ganglia region on the left side.The hyperintensity on the T1-weighted imagesis resulting from leakage of the gadolinium-based contrast agent injected at the baselineperfusion-weighted examination 3 hours earlier.On the following day, a hemorrhagic transfor-mation is found in the same location on the T2*GRE images.

Table. Univariate Analysis of Association of Imaging VariablesWith Hemorrhagic Transformation

HemorrhagicTransformation

Yes No P

Parenchymal enhancement, 2 hours

Yes 5 (100) 0 0.04

No 11 15

HARM, 2 hours

Yes 1 (25) 3 0.60

No 15 13

Reperfusion, 2 hours

Yes 7 (64) 4 0.70

No 7 6

Reperfusion, 24 hours

Yes 12 (63) 7 0.54

No 1 2

Percentages are shown in parentheses. Univariate tests of independence in2�2 contingency tables by Fisher exact test.

1026 Stroke March 2008

DWI lesion volume at baseline was significantly higher inthe HT-group than in the non-HT group (26.0 versus 6.6mL, P�0.02, Student t test). Similarly, admission NIHSSwas related to HT (P�0.04) and mean NIHSS was signif-icantly higher in the HT group than in the non-HT group(13.3 versus 8.5, P�0.03). An exact anatomic relationshipbetween PE and subsequent HT was found in all cases.

DiscussionThis study of contrast agent (CA) leakage in tPA-treatedstroke patients showed PE as early as 2 hours post-treatment.Furthermore, we found that local tissue reperfusion precededBBB disruption in the majority of patients.

All patients with PE 2 hours after thrombolysis developedHT, confirming the high specificity and low sensitivityrecently noted in a retrospective study.6 Parenchymal en-hancement and subsequent HT colocalized in all of cases. Incontrast, only 1 of 4 patients showing HARM developed HT.Latour et al4 retrospectively identified HARM in 53% oftPA-treated patients, of whom 46% developed HT, andfound a correlation between HARM and reperfusion within1 week. We extend that finding by demonstrating earlyreperfusion in all cases of HARM. We speculate that thenumber of cortical and periventricular infarctions accountsfor differences in the occurrence of HARM between the 2studies.

The sensitivity of PE was low; it failed to precede all casesof HT at 2 hours, perhaps due to the short half-life ofgadobutrol (1.5 hour). Our preliminary data suggest that

repeated T1-imaging immediately following 2-hour PWIincreases sensitivity (results not shown). Furthermore, CAmay not enter tissue with severe ischemia and hence fail todelineate BBB abnormality. The majority of patients wereexamined with a 3T scanner. No PE was recorded at 1.5T at2-hour follow-up, and we speculate that 1.5T images may beless sensitive to CA leakage. All hemorrhages were asymp-tomatic. In contrast to thrombolysis-related “severe HT”(PH), “mild HT” (HI) may be a marker of successfulrecanalization without adverse impact on neurological out-come.7 Thomalla et al8 identified predictors of the 2 types ofHT and suggested that they have different pathogenesis. Thesmall number of PH in our study precludes us for statingwhether specific BBB leakage patterns are associated withPH rather than HI. We speculate that the extent of PE and theseverity of the subsequent HT are correlated. Indeed, studiesusing serial postcontrast T1-weighted imaging, PWI, andDWI may give insight into the pathology of BBB disruptionin ischemic tissue and help target and monitor future adjunc-tive therapies for reducing hemorrhagic complications inthrombolytic treatment.

Sources of FundingThis study was supported by The Danish National Research Foun-dation (NH, CS, MA, SC, KM, OW, LØ), The Danish MedicalResearch Council (NH, LØ), The Velux Foundation (NH, CS), andThe Toyota Foundation (NH, CS). Contrast agent was kindlysponsored by Schering Diagnostics AG, Berlin, Germany.

DisclosuresNone.

Figure 2. Magnetic resonance images from apatient who suffered hemorrhagic transforma-tion after thrombolytic therapy. Imaging wasperformed before thrombolysis, 2 hours, and24 hours after treatment. Baseline DWI showsacute infarction of the right insular cortex. Theacute MTT map depicts hypoperfusion in themajority of the tissue supplied by the right mid-dle cerebral artery. T2* GRE shows hemor-rhagic infarction 2 hours after treatment andprogression to parenchymal hematoma after 24hours. Local reperfusion of the putamen isfound at the 2-hour MTT map and correspondsanatomically to the parenchymal enhancementon the concomitant postcontrast T1-weightedimages. Gd indicates gadolinium-based con-trast agent.

Hjort et al Blood-Brain Barrier Disruption After Thrombolysis 1027

References1. The National Institute of Neurological Disorders and Stroke rt-PA Stroke

Study group. Tissue plasminogen activator for acute ischemic stroke.N Engl J Med. 1995;333:1581–1587.

2. del Zoppo GJ, von Kummer R, Hamann GF. Ischaemic damage of brainmicrovessels: Inherent risks for thrombolytic treatment in stroke. J NeurolNeurosurg Psychiatry. 1998;65:1–9.

3. Knight RA, Barker PB, Fagan SC, Li Y, Jacobs MA, Welch KM. Pre-diction of impending hemorrhagic transformation in ischemic stroke usingmagnetic resonance imaging in rats. Stroke. 1998;29:144–151.

4. Latour LL, Kang DW, Ezzeddine MA, Chalela JA, Warach S. Earlyblood-brain barrier disruption in human focal brain ischemia. Ann Neurol.2004;56:468–477.

5. Wu O, Ostergaard L, Weisskoff RM, Benner T, Rosen BR, Sorensen AG.Tracer arrival timing-insensitive technique for estimating flow in MRperfusion-weighted imaging using singular value decomposition with a

block-circulant deconvolution matrix. Magn Reson Med. 2003;50:164–174.

6. Kim EY, Na DG, Kim SS, Lee KH, Ryoo JW, Kim HK. Prediction ofhemorrhagic transformation in acute ischemic stroke: role of diffusion-weighted imaging and early parenchymal enhancement. AJNR Am JNeuroradiol. 2005;26:1050 –1055.

7. Molina CA, Alvarez-Sabin J, Montaner J, Abilleira S, Arenillas JF,Coscojuela P, Romero F, Codina A. Thrombolysis-related hemorrhagicinfarction: a marker of early reperfusion, reduced infarct size, andimproved outcome in patients with proximal middle cerebral arteryocclusion. Stroke. 2002;33:1551–1556.

8. Thomalla G, Sobesky J, Kohrmann M, Fiebach JB, Fiehler J, ZaroWeber O, Kruetzelmann A, Kucinski T, Rosenkranz M, Rother J,Schellinger PD. Two tales: Hemorrhagic transformation but not paren-chymal hemorrhage after thrombolysis is related to severity andduration of ischemia: MRI study of acute stroke patients treated withintravenous tissue plasminogen activator within 6 hours. Stroke. 2007;38:313–318.

1028 Stroke March 2008