Rudiger von Kummer, MDHolger Bourquain, MDStefano Bastianello, MDLuigi Bozzao, MDClaude Manelfe, MDDieter Meier, MDWerner Hacke, MDFor the European

Cooperative Acute StrokeStudy II Group

Index terms:Brain, CT, 10.12111Brain, infarction, 10.781Brain, ischemia, 10.781

Radiology 2001; 219:95–100

Abbreviations:ECASS II 5 European Cooperative

Acute Stroke Study IIMCA 5 middle cerebral arteryNIHSS 5 National Institutes of Health

Stroke Scale

1 From the Departments of Neurora-diology, University of Technology,Fetscherstrasse 74, D-01307 Dresden,Germany (R.v.K., H.B.); Universita LaSapienza, Rome, Italy (S.B., L.B.); andUniversity of Toulouse, France (C.M.);the Department of Neurology, Univer-sity of Heidelberg, Germany (W.H.);and Boehringer-Ingelheim, Germany(D.M.). Received April 24, 2000; revi-sion requested June 12; revision re-ceived July 17; accepted August 11.Address correspondence to R.v.K. (e-mail: kummer-r@rcs.urz.tu-dresden.de).© RSNA, 2001

Author contributions:Guarantor of integrity of entire study,R.v.K.; study concepts and design, allauthors; definition of intellectual con-tent, all authors; literature research,R.v.K., H.B.; clinical studies, R.v.K.,D.M., W.H.; data acquisition, R.v.K.,S.B., L.B., C.M., D.M.; data analysis/interpretation, all authors; statisticalanalysis, R.v.K.; manuscript prepara-tion, all authors; manuscript editing,D.M., W.H., R.v.K.; manuscript revi-sion/review and final version approval,all authors.

Early Prediction ofIrreversible Brain Damageafter Ischemic Stroke at CT1

PURPOSE: To assess the capability of computed tomography (CT) in the predictionof irreversible ischemic brain damage and its association with the clinical coursewithin 6 hours of stroke onset.

MATERIALS AND METHODS: Serial CT scans obtained within 6 hours of strokeonset, at 22–96 hours (median, 1 day), and at 2–36 days (median, 7 days) aftersymptom onset in 786 patients with ischemic stroke were prospectively studied, andfollow-up CT scans were used as the reference. Clinical variables were assessedprospectively and independently of CT evaluation.

RESULTS: The specificity and positive predictive value of ischemic edema at baselineCT for brain infarcts were 85% (95% CI: 77%, 91%) and 96% (95% CI: 94%, 98%),respectively. Sensitivity and negative predictive values were 64% (95% CI: 60%, 67%)and 27% (95% CI: 23%, 32%), respectively. Patients without early CT findings were lessseverely affected (P , .001), developed smaller infarcts (P , .001), had fewer intracra-nial bleeding events (P , .001), and had a better clinical outcome at 90 days (P , .001)compared with patients with hypoattenuating brain tissue at early CT.

CONCLUSION: After ischemic stroke, x-ray hypoattenuation at CT is highly specificfor irreversible ischemic brain damage if detection occurs within the first 6 hours.Patients without hypoattenuating brain tissue have a more favorable clinical course.

Neurologic symptoms caused by focal brain ischemia do not necessarily represent irrevers-ible brain damage. The critical perfusion level for functional disturbance is at 15–25mL/100 g per minute and above the critical perfusion level for tissue death (10–15 mL/100g per minute) (1–3). Neuronal tissue can survive with cerebral blood flow values higherthan 12 mL/100 g per minute in a state of dysfunction for an undefined period (2). Thisdifference in critical blood flow levels for reversible dysfunction and irreversible tissuedamage allows spontaneous or treatment-induced recovery. The focal degree of ischemiadetermines the delay between the onset of symptoms and induction of irreversible braindamage (4). The heterogeneous pattern of cerebral ischemia and its variation within shortperiods may provide different time windows to initiate treatment.

Information on whether the ischemic brain has a chance to survive or is already dead isimportant to assess the effect of treatment. Because of time constraints, a reliable methodof cerebral blood flow measurement that could clearly facilitate the prediction, fromcerebral blood flow values, of which brain regions will die and which will survive is notroutinely applicable. Even with histologic staining, identification of irreversible damage isdifficult in experimental animals within the first hours after arterial occlusion (5). It wasshown (6–8), however, that brain tissue affected by severe ischemia with a perfusion levelof less than 15 mL/100 g per minute takes up water. This process becomes irreversiblewithin a short time (9).

We studied the hypothesis that an increase in brain tissue water content, indicated bya decrease in x-ray attenuation, might herald irreversible ischemic tissue damage. We usedthe prospective computed tomography (CT) protocol of the European Cooperative AcuteStroke Study II (ECASS II) (10), a large multicenter trial on thrombolysis, to study thecapability of CT in the early prediction of irreversible ischemic brain damage. We further

Neuroradiology

95

studied whether the extent of ischemicbrain edema at initial CT is associatedwith the clinical course.

MATERIALS AND METHODS

ECASS II was a controlled, randomized,double-blind, multicenter trial (10) tostudy the effect of recombinant tissueplasminogen activator on clinical out-come and infarct volume at CT afteracute ischemic stroke. The study protocolincluded three CT examinations for eachpatient: before randomization within 6hours of symptom onset, at 22–36 hoursafter symptom onset, and at 6–8 daysafter symptom onset. Local ethics com-mittees approved the study. All CT scanswere consecutively and prospectivelyevaluated by the local investigators andadditionally and independently by a CTreading panel that consisted of three ex-perienced neuroradiologists (includingS.B., L.B., R.v.K., C.M.) from three Euro-pean countries.

Patients

Patients (325 women, 461 men; me-dian age, 68 years; age range, 18–84years) were enrolled in Europe, Australia,and New Zealand from October 1996 toJanuary 1998 after informed consent wasobtained and if they were 18–80 years ofage and had a moderate to severe hemi-spheric stroke within the last 6 hours.Intracranial hemorrhage, a hypoattenu-ating area exceeding one-third of themiddle cerebral artery (MCA) territory atCT, severe stupor, coma, and recoverybefore treatment initiation were amongthe exclusion criteria (10). Four hundrednine patients were assigned to receive re-combinant tissue plasminogen activator(0.9 mg per kilogram of body weight,with an upper limit of 90 mg, and 391were assigned to receive placebo.

CT Evaluation

All local investigators of ECASS II (532physicians; 372 (70%) neurologists, 69(13%) radiologists, 64 (12%) neuroradiolo-gists, and 27 (5%) internists or others) par-ticipated in training courses to improve thequality of both the CT scanning procedureand CT scan evaluation (11). Baseline andfollow-up CT scans were obtained withoutcontrast enhancement and with the samescanner if possible. Windows and centerlevels were set to optimally distinguishgray and white matter.

Following the prospective ECASS pro-tocol, all CT scans were evaluated twice,

first by the local investigators and thenindependently by three members of theCT reading panel, who were blinded totreatment allocation, follow-up CT find-ings, and the reading results of the localinvestigators. The chairman of the CTreading panel gathered the categorizedfindings from the panel members, checkedthem for disagreements, disclosed dis-agreements to the other panel members,and discussed each discrepant CT findingto achieve a consensus. The final judg-ments were sent to the data managementcenter, which then was allowed to sendthe follow-up scans to the members ofthe CT reading panel.

The members of the CT reading panelwere informed which hemisphere was af-fected and reviewed the CT scans for hy-poattenuating arterial territories respon-sible for the symptoms of acute stroke.We defined hypoattenuation as a visibledecrease in x-ray attenuation of brain tis-sue compared with the attenuation inother portions of the same anatomicstructure or its contralateral counterpart.We categorized the extent of hypoat-tenuation of the MCA territory as no hy-poattenuation, 33% or less (small), orgreater than 33% (large). On follow-upscans, we measured the volume of acuteischemic lesions by multiplying the max-imum diameter of the hypoattenuatingarea, the maximum diameter of the areaperpendicular to it in the same section,the number of sections affected, the sec-tion distance, and a conversion factor of0.5 by using the formula for irregular vol-umes.

In patients with hemorrhagic transfor-mation, the entire lesion was measured.We classified hemorrhagic events ashemorrhagic infarction or parenchymalhematoma according to the definitionsused in the ECASS II.

Clinical Assessments

Local investigators used the NationalInstitutes of Health Stroke Scale (NIHSS)

to assess the neurologic deficits of thepatients at baseline; at 24 hours aftersymptom onset; and at 7, 30, and 90 days(SD, 14). The NIHSS is a 42-point scalethat is used to quantify neurologic defi-cits in 11 categories. For example, a mildfacial paresis is assigned a score of 1, anda complete hemiplegia with aphasia, gazedeviation, hemianopia, dysarthria, andsensory loss is assigned a score of 25.

The investigators used the Barthel in-dex and the modified Rankin score at 90days 6 14 (SD) as clinical endpoints. TheBarthel index is a reliable and valid mea-sure of the ability of the patient to per-form activities of daily living. A score of100 indicates complete independence.The modified Rankin scale is a simplifiedoverall assessment of function in which ascore of 0 indicates absence of symptomsand a score of 6, death. If values weremissing, data from the last observationwere used. For the modified Rankin scoreand the Barthel index, a worst-case impu-tation was made for missing values at day90. The primary endpoint of ECASS IIwas the proportion of patients who had afavorable outcome (score , 2) on themodified Rankin scale.

Data Analysis

We used an intention-to-treat analysisfor the radiographic endpoints. Ischemiclesions at follow-up CT as described bythe CT reading panel served as the refer-ence for irreversible tissue damage irre-spective of whether they were detected22–96 hours (first follow-up CT) or 2–36days (second follow-up CT) after symp-tom onset. In four patients with missingCT scans from day 1, we used the infor-mation provided at CT on day 7. Thefindings of the first follow-up CT exami-nation were used in 38 patients withmissing second follow-up CT findings.

Baseline CT findings were consideredtrue-positive if follow-up CT scans con-firmed the ischemic lesion in the samelocation, and they were considered true-

TABLE 1Predictive Values with Baseline CT

MeasureLocal ECASS IIInvestigators

ECASS II CT ReadingPanel P Value*

Sensitivity (%) 40 (36, 44; 271 of 679) 64 (60, 67; 433 of 679) ,.001Specificity (%) 79 (70, 85; 84 of 107) 85 (77, 91; 91 of 107) ..05Positive predictive value (%) 92 (89, 95; 271 of 294) 96 (94, 98; 433 of 449) ,.01Negative predictive value (%) 17 (14, 21; 84 of 492) 27 (23, 32; 91 of 337) ,.001Accuracy (%) 45 (42, 49; 355 of 786) 67 (63, 70; 524 of 786) ,.001

Note.—Data in parentheses are the 95% CIs and the number of patients.* P values were determined with use of the McNemar test.

96 z Radiology z April 2001 von Kummer et al

negative if baseline and follow-up CTscans did not reveal an ischemic infarct.If follow-up CT scans did not show thetypical development (increased hypoat-tenuation, sharper demarcation) of anischemic lesion seen on the baseline scanor if they were normal in the locationindicated at baseline CT, we classified thefinding at baseline as false-positive. Wethen reviewed the false-positive findingsfor an explanation, such as obliquity ofthe scan or partial volume artifacts. Nor-mal baseline scans with ischemic lesionsat follow-up were called false-negative.

On the basis of these definitions, wecalculated sensitivity, specificity, positiveand negative predictive values, and accu-

racy with their 95% CIs. The positive pre-dictive value is the proportion of findingson the baseline scans that are confirmedon follow-up scans. The negative predic-tive value describes the proportion ofnegative findings on baseline scans thatare still negative on the follow-up scans.

Statistical Analysis

Numeric variables are presented asmeans with SDs or 95% CIs. All propor-tions are presented with 95% CIs. To cal-culate the 95% CI, we used the formulasuggested by Berry (12). Differencesbetween the distributions were testedwith the Mann-Whitney U test and theKruskal-Wallis test. Differences betweenproportions were tested with the x2 testand the McNemar test. We accepted .05as a level of significance.

RESULTS

The study population consisted of 786patients after exclusion of 14 patients inwhom we did not obtain baseline CTscans (n 5 9) or any follow-up CT scans(n 5 5). The median times for both fol-low-up CT examinations were 1 and 7days after stroke onset; 46 (6%) patientswere examined outside the study win-dows (22–36 hours and 6–8 days) ofECASS II.

Findings at Baseline CT

The local investigators of ECASS IIidentified 294 (37%; 95% CI: 34%, 41%)patients with ischemic hypoattenuationat baseline CT; among them were 271(34%) patients with true-positive find-ings and 23 (3%) patients with false-pos-itive findings compared with the findingsat follow-up CT. Their findings were true-negative in 84 patients. Follow-up CTshowed infarcts in 408 (52%) patientsthat were not seen at baseline (Table 1).

The members of the ECASS II CT read-

ing panel identified 449 (57%; 95% CI:54%, 61%) patients with ischemic hy-poattenuation at baseline CT; amongthem were 37 patients with hypoattenu-ation of greater than 33% of the MCAterritory (ECASS II exclusion criterion).Compared with the follow-up CT find-ings, the findings of the CT reading panelwere true-positive in 433 (55%) patients,false-positive in 16 (2%), and true-nega-tive in 91 (12%). The retrospective reviewof false-positive findings did not revealregression of a brain infarct. Three patientsdeveloped infarcts in another location (Fig1), and 13 patients had hypoattenuatingareas without further demarcation at fol-low-up CT that were retrospectively ex-plained as partial volume artifacts or oldsubcortical ischemic lesions. In some pa-tients, motion artifacts and low scanquality explained the false-positive CTdiagnosis.

Follow-up CT depicted ischemic lesionsin 246 (31%) patients that were not seen atbaseline. In Table 1, the predictive valuesof the baseline CT findings provided bythe local investigators are compared withthose of the CT reading panel. Treatmentdid not affect the predictive value of thebaseline CT findings (Table 2).

Figure 2 shows the hypoattenuation ofthe lentiform nucleus on a CT scan ob-tained 22 minutes after stroke onset andthe follow-up CT findings. The propor-tion of patients with true-positive find-ings did not vary considerably during thefirst 5 hours after stroke onset and de-creased in the 6th hour (Table 3). Thelater the patients were examined withCT, the less severe were the symptomswith which they presented (Kruskal-Wal-lis test, P , .001).

Figure 3 shows the NIHSS scores at dif-ferent time points after stroke onset inpatients with no, small, or large volumesof hypoattenuating brain tissue at base-line CT. Findings at CT were used to sep-arate patient groups with different neu-

TABLE 2Predictive Values with Baseline CT Findings of Ischemic Lesions at Follow-upCT on Days 1 and 7 for the Two Treatment Groups as Assessed by theCT Reading Panel

Measure Placebo Recombinant Tissue Plasminogen

Sensitivity (%) 64 (59, 70; 213 of 233) 64 (58, 68; 220 of 346)Specificity (%) 77 (63, 87; 37 of 48) 92 (82, 96; 54 of 59)Positive predictive value (%) 95 (91, 97; 213 of 224) 98 (95, 99; 220 of 225)Negative predictive value (%) 24 (18, 31; 37 of 157) 30 (24, 37; 54 of 180)Accuracy (%) 66 (61, 70; 250 of 381) 68 (63, 72; 274 of 405)

Note.—Data in parentheses are the 95% CIs and the number of patients.

Figure 1. False-positive CT reading by thepanel. At baseline (top row), the right insularcortex (arrow) was interpreted as hypoattenu-ating, as compared with the contralateralinsular cortex. Because of beam-hardening ar-tifacts, a safe judgment in the brainstem (ar-rowhead) was impossible. On day 1 after strokeonset (middle row), a right paramedian pon-tine infarct (arrowhead) became visible. Atten-uation of the right insular cortex (arrow) isunchanged. On day 7 after stroke onset (bot-tom row), the pontine infarct (arrowhead) isclearly visible. The right insular cortex (arrow)appears normal.

Volume 219 z Number 1 Ischemic Brain Damage: Early CT Prediction z 97

rologic impairments at baseline and atfollow-up. Patients with no or small hy-poattenuating tissue volume recovereduntil day 30, although to a differentlevel. The condition of patients with alarge hypoattenuating tissue volume de-teriorated in the first week, and the pa-tients did not again achieve the level ofneurologic deficit they had immediatelyafter stroke onset.

We observed significant differencesamong these three groups with regard tothe severity of neurologic symptoms atbaseline and outcome, the infarct volumeat day 7, the incidence of bleeding events,and the fatality rate (Table 4). Of the 337patients with normal CT findings at base-line, only 14 (4%) died of a cerebral cause(ischemic edema, hemorrhage). The fre-quency of cerebral deaths was 12 (32%) of37 patients in those with large hypoattenu-ating tissue volume at baseline (P , .001).

Infarcts Confirmed at Follow-up CT

Forty-two patients underwent one fol-low-up CT examination, the findings ofwhich confirmed the findings (32 in-farcts, three normal scans) at baseline in35 patients. In seven patients, an infarctappeared on the day 7 CT scan that wasnot seen on the baseline scan. Findings ofthe second follow-up CT examinationconfirmed the findings of the first fol-low-up CT examination (620 infarcts,104 normal CT scans) in 724 (97%) of744 patients with two follow-up scans. In16 (2%) patients, infarcts appeared at CTbetween days 1 and 7 after stroke onset.The condition of one of these patientsdeteriorated by 14 points on the NIHSSbetween days 1 and 7; the others re-mained clinically stable or slightly im-proved. In four (0.5%) patients, the in-farct was no longer visible at the secondfollow-up CT examination.

Remarkably, the mean infarct volumeincreased between days 1 and 7 from 48mL 6 77 to 61 mL 6 93 (Mann-WhitneyU test, P , .001). Despite the increase inmean infarct volume, the median NIHSSscore improved from 8 to 6 (Mann-Whit-ney U test, P , .001) between days 1 and7. The infarct volume at the second fol-low-up CT correlated directly with theNIHSS at day 7 (r2 5 0.47, P , .001), atday 90 (r2 5 0.32, P , .001), and indi-rectly with the Barthel index at day 90(r2 5 20.29, P , .001).

Patients with infarcts at follow-up CT(n 5 679) had a poorer clinical outcomeat 90 days than patients without a detect-able infarct (n 5 107). The mean Barthelindex was 68 6 37 in patients with in-

farct and 89 6 26 in patients withoutinfarct at follow-up CT (Mann-WhitneyU test, P , .001). Only 227 (33%) of 679patients with infarct at follow-up CT hada modified Rankin score of 0 or 1 com-pared with 75 (70%) of 107 patients with-out infarct (odds ratio, 0.21; 95% CI:0.14, 0.33).

Treatment did not affect the number ofinfarcts detected at follow-up CT. Themean infarct volume increased similarlyduring CT follow-up in both treatmentgroups.

DISCUSSION

This was a relatively large cohort of acutestroke patients who were prospectivelyevaluated with repeated brain imaging.The patients were primarily selected tostudy the effect of recombinant tissueplasminogen activator and had a moder-ate to severe hemispheric stroke. Becauseof the exclusion criteria of ECASS II, pa-tients with brainstem and cerebellarstrokes, patients with reversible symp-toms, and patients with large areas ofhypoattenuation at baseline CT are un-

derrepresented. The study drug did notaffect the prospective values at CT andthe proportion of patients with brain in-farcts at follow-up. We discuss, therefore,both treatment groups together.

In addition to the local investigators ofECASS II, three experienced neuroradi-ologists individually evaluated the CTscans and had the opportunity to consulteach other. The finding of hypoattenuat-ing brain tissue at baseline CT had a highpositive predictive value for an ischemicinfarct at follow-up CT, irrespective ofwhether the local investigators or the CTreading panel assessed the CT scans. Thelocal investigators were less sensitive andless specific (difference not significant) indetecting hypoattenuating brain tissuethan were the neuroradiologists. The dif-ference can be explained by the differentsettings in which the CT scans were read:The local investigators had to evaluatethe CT findings as quickly as possible toallow early randomization and treat-ment. In contrast, the three neuroradi-ologists could evaluate the scans withoutany time constraints and could finallyconsult each other, but they had less clin-

Figure 2. CT scan obtained 22 minutes after onset of left-sided hemiparesis. Hypoattenuation ofthe right lentiform nucleus (arrows) causes obscuration of this structure (left). Ischemic lesion iswell delineated at days 1 (middle) and 7 (right) after stroke onset.

TABLE 3Sensitivity and Positive Predictive Values with Baseline CT for Ischemic Infarctsat Intervals after Stroke Onset

Interval (h)* Sensitivity (%)†Positive Predictive

Value (%)†Median

NIHSS Score

1 (n 5 36) 58 (41, 74; 18 of 31) 100 (82, 100; 18 of 18) 12.52 (n 5 210) 66 (59, 73; 117 of 176) 95 (90, 98; 117 of 123) 12.03 (n 5 242) 65 (58, 71; 142 of 219) 98 (94, 99; 142 of 145) 11.04 (n 5 172) 64 (56, 71; 94 of 147) 96 (90, 98; 94 of 98) 10.05 (n 5 97) 66 (55, 76; 51 of 77) 93 (83, 97; 51 of 55) 9.06 (n 5 29) 38 (21, 59; 8 of 21) 80 (49, 94; 8 of 10) 9.0

* Numbers in parentheses are the numbers of patients.† Data in parentheses are the 95% CI and the number of patients.

98 z Radiology z April 2001 von Kummer et al

ical information. This setting was chosento determine the capability of CT underoptimal conditions. Training and experi-ence may also have contributed to theneuroradiologists’ higher sensitivity andspecificity.

In agreement with others (13), we didnot observe a single patient with resolu-tion of tissue hypoattenuation within thefirst 24 hours after stroke onset in bothtreatment groups. Infarcts seen at thefirst follow-up CT at 24 hours were con-firmed by the second follow-up CT at 7days, with only four (0.5%) exceptionsthat are best explained by a fogging ef-fect, a transient increase of hypoattenua-tion within ischemic brain lesions at CTin the 2nd and 3rd week (14,15). Onlypitfalls such as partial volume artifacts,motion artifacts, scan obliquity, or isch-emic subcortical demyelination impairedthe specificity of the baseline CT find-ings.

These observations are in line withthose of others (13,16–18) and suggestthat hypoattenuating brain tissue cover-ing an arterial territory that appeared ona CT scan after a stroke is a distinctpathophysiologic finding that representsirreversible damage. During the initial 3

hours of ischemia, an increase in waterand sodium contents is almost exclu-sively confined to gray matter (6,19).Brain cortex water content increases im-mediately after arterial occlusion (7,8) ifperfusion decreases below 12 mL/100 gper minute, the critical flow level for struc-tural integrity (3). If brain tissue water con-tent increases by 1%, x-ray attenuationdecreases by 2–3 HU (20). In animal ex-periments, x-ray attenuation declined by7.5 HU 6 1.6 within 4 hours of MCAocclusion (21). Brain tissue hypoattenua-tion after arterial occlusion may, there-fore, be interpreted as ischemic edema,and it may be used as an indicator forsevere focal ischemia with subsequentnecrosis. Ischemic edema appearing ingray matter first causes a diminished con-trast to adjacent white matter and, thus,a loss of anatomic margins. Such graymatter hypoattenuation explains nega-tive phenomena such as obscuration oflentiform nucleus and loss of the insularribbon (17,22).

The subtlety of early ischemic graymatter edema at CT may be one reason toexplain why ischemic infarcts are oftenmissed at CT during the first 6 hours aftersymptom onset, as the ECASS II local in-vestigators did, and the interobserveragreement is only moderate (23,24). Itdoes not explain, however, why experi-enced reviewers do not detect any is-chemic edema at CT in about one-thirdof infarcts that appear later at CT. In thisstudy, the proportion of patients withx-ray hypoattenuation within the first 6hours of stroke was in the range (56%–92%) of that of other studies (13,17,18,

22,25–27), which is higher than that as-sumed in recent reviews (28–30).

A normal CT finding in a patient withacute stroke can be explained by focalischemia above the critical flow level ofstructural integrity, by an early stage ofischemic edema causing hypoattenua-tion below contrast resolution, by ische-mia mainly confined to white matter, orby small volumes of ischemic brain tissueor lesions near the skull base where beamhardening artifacts impair recognition.We presume that the development ofischemic edema is delayed beyond thefirst 6 hours in a considerable proportionof patients. This delay occurred in 246(73%) of 337 patients with normal initialCT findings in ECASS II. A standard forirreversible ischemic brain damage is notavailable for the first few hours afterstroke onset. Areas of disturbed protondiffusion at diffusion–weighted magneticresonance imaging may be unspecific forpermanent damage (31,32). Therefore,problems may arise in the assessment ofthe sensitivity and accuracy of early CTfor irreversible tissue damage. Low valuesmay be due to a delay in the develop-ment of such tissue changes and notcaused by the incapability of CT.

Like others (18), we found that hypoat-tenuating brain tissue at CT within thefirst 6 hours of stroke is associated with alarger infarct volume, more severe symp-toms, a less favorable clinical course, anda high risk for secondary cerebral hemor-rhage and death compared with normalfindings at CT. The association betweenthe extent of hypoattenuation at baselineCT and the frequency of cerebral deaths

TABLE 4Extent of Hypoattenuation at Baseline CT, Neurologic Deficit, Bleeding Events,and Clinical Outcome at 90 Days

Clinical Data

Hypoattenuation of Brain Parenchyma at Baseline

P ValueNone

(n 5 337)

#33% ofMCA Territory

(n 5 412)

.33% ofMCA Territory

(n 5 37)

NIHSS score*At baseline 9 6 4 13 6 6 17 6 7 ,.001†

At 90 d 6 6 12 10 6 13 19 6 18 ,.001†

Infarct volume at 7 d (mL)* 20 6 51 82 6 98 194 6 120 ,.001†

Barthel index at 90 d* 77 6 35 67 6 37 46 6 43 ,.001†

Modified Rankin score of 0or 1 at 90 d‡

159 (47) 136 (33) 7 (19) ,.001§

Fatalities‡ 28 (8) 42 (10) 13 (35) ,.001§

Hemorrhages‡ ,.001§

Hemorrhagic infarct 55 (16) 208 (50) 22 (59) No dataParenchymal hematoma 15 (4) 41 (10) 4 (11) No data

* Data are the mean plus or minus the SD.† P value was determined with use of the Kruskal-Wallis test.‡ Data are the number of patients. Data in parentheses are percentages.§ P value was determined with use of the x2 test.

Figure 3. Mean NIHSS scores (vertical axis)with 95% CIs in patients with no hypoattenu-ation at early CT (n 5 337, Œ), in patients withsmall hypoattenuating tissue volumes of 33%or less of the MCA territory (n 5 412, F), andin patients with greater than 33% MCA terri-tory hypoattenuation (n 5 37, ■) at baseline(NIHSS 0), at day 1 (NIHSS 1), at day 7 (NIHSS7), at day 30 (NIHSS 30), and at day 90 (NIHSS90) after stroke onset. The NIHSS is used toscore the neurologic deficit with a 42-pointscale; 0 means no deficit, and 42, a severe def-icit. Patients with normal CT findings within 6hours of stroke onset had a better score atbaseline and a better clinical course than pa-tients with positive CT findings. Patients withsmall hypoattenuation at early CT improvedclinically, as did the patients with normal CTfindings. Patients with hypoattenuation ex-ceeding 33% of the MCA territory did not im-prove during the first 3 months after stroke.

Volume 219 z Number 1 Ischemic Brain Damage: Early CT Prediction z 99

was similar to that in ECASS I (33). Thisobservation supports the assumptionthat patients with less severe strokes anda good prognosis develop no, delayed, oronly small ischemic brain lesions (18,34).The association between CT findings andclinical performance explains the lowerproportion of positive CT findings in lessseverely affected patients examined inthe 6th hour after stroke onset. As inECASS I, the feature of ischemic infarc-tions at CT was not stable even after 24hours after stroke in our study (35).

New infarcts appeared on the CTscans in 16 patients, and the mean vol-ume of lesions increased between days 1and 7, although the patients clinicallyimproved, on average. These findings re-flect that the symptoms of stroke are notdirectly linked to the underlying ische-mic tissue damage. The eloquence ofbrain regions is different, and enlarge-ment of ischemic edema may occur evenwith clinical recovery. We have no infor-mation on whether other lesions may ap-pear even after 7 days after stroke onsetand whether these lesions further in-crease in volume beyond our observationperiod. We found, however, a significantcorrelation between the presence of le-sions and their volume at the follow-upCT, with clinical outcome at 90 days.This correlation suggests that ischemiclesions, as assessed at day 7 after strokeonset, are clinically relevant and that thepossibility of their early detection isworth investigating.

In summary, we found that CT ishighly specific in the detection of irre-versible ischemic brain tissue damagewithin the first hours of stroke onset.Ischemic edema is present within the first6 hours of symptom onset in about two-thirds of patients. A normal early CT scanin a patient with stroke indicates a lesssevere disease. Ischemic tissue damage isdelayed in these patients.

Acknowledgments: We thank the local in-vestigators, all study committee members, andthe colleagues of the study monitoring centerfor their great support. Boehringer-Ingelheim,Germany, sponsored ECASS II and CT readingtraining and continues to support the use ofthe study data for scientific analyses.

References1. Astrup J, Siesjo B, Symon L. Thresholds in

cerebral ischemia: the ischemic penum-bra. Stroke 1981; 12:723–725.

2. Heiss W, Rosner G. Functional recovery ofcortical neurons as related to degree and

duration of ischemia. Ann Neurol 1983;14:294–301.

3. Hossmann KA. Viability thresholds andthe penumbra of focal ischemia. AnnNeurol 1994; 36:557–565.

4. Jones T, Morawetz R, Crowell R, et al.Thresholds of focal cerebral ischemia inawake monkeys. J Neurosurg 1981; 54:773–782.

5. Garcia J, Liu KF, Ho KL. Neuronal necrosisafter middle cerebral artery occlusion inWistar rats progresses at different timeintervals in the caudoputamen and thecortex. Stroke 1995; 26:636–643.

6. Watanabe O, West CR, Bremer A. Experi-mental regional cerebral ischemia in themiddle cerebral artery territory in pri-mates. II. Effects on brain water and elec-trolytes in the early phase of MCA stroke.Stroke 1977; 8:71–76.

7. Schuir FJ, Hossmann KA. Experimentalbrain infarcts in cats. II. Ischemic brainedema. Stroke 1980; 11:593–601.

8. Mintorovitch J, Yang G, Shimizu H, Ku-charczyk J, Chan P, Weinstein P. Diffu-sion-weighted magnetic resonance imag-ing of acute focal cerebral ischemia:comparison of signal intensity withchanges in brain water and Na1,K1

-ATPase activity. J Cereb Blood FlowMetab 1994; 14:332–336.

9. Ianotti F, Hoff J. Ischemic brain edemawith and without reperfusion: an experi-mental study in gerbils. Stroke 1983; 14:562–567.

10. Hacke W, Kaste M, Fieschi C, et al. Ran-domised double-blind placebo-controlledtrial of thrombolytic therapy with intra-venous alteplase in acute ischaemicstroke (ECASS II). Lancet 1998; 352:1245–1251.

11. von Kummer R. Effect of training in read-ing CT scans on patient selection forECASS II. Neurology 1998; 51(suppl 3):S50–S52.

12. Berry C. A tutorial on confidence inter-vals for proportions in diagnostic radiol-ogy. AJR Am J Roentgenol 1990; 154:477–480.

13. Bozzao L, Bastianello S, Fantozzi LM, An-geloni U, Argentino C, Fieschi C. Corre-lation of angiographic and sequential CTfindings in patients with evolving cere-bral infarction. AJNR Am J Neuroradiol1989; 10:1215–1222.

14. Becker H, Desche H, Hacker H, Pencz A.CT fogging effect with ischemic cerebralinfarcts. Neuroradiology 1979; 18:185–192.

15. Skriver E, Olsen T. Transient disappear-ance of cerebral infarcts on CT scan, theso-called fogging effect. Neuroradiology1981; 22:61–65.

16. Skriver EB, Olsen TS. Contrast enhance-ment of cerebral infarcts: incidence andclinical value in different states of cere-bral infarction. Neuroradiology 1982; 23:259–265.

17. Tomura N, Uemura K, Inugami A, FujitaH, Higano S, Shishido F. Early CT findingin cerebral infarction. Radiology 1988;168:463–467.

18. Horowitz SH, Zito JL, Donnarumma R,Patel M, Alvir J. Computed tomographic-angiographic findings within the first five

hours of cerebral infarction. Stroke 1991;22:1245–1253.

19. Garcia J. Morphology of global cerebralischemia: a review. Crit Care Med 1988;16:979–987.

20. Unger E, Littlefield J, Gado M. Water con-tent and water structure in CT and MRsignal changes: possible influence in de-tection of early stroke. AJNR Am J Neuro-radiol 1988; 9:687–691.

21. von Kummer R, Weber J. Brain and vas-cular imaging in acute ischemic stroke:the potential of computed tomography.Neurology 1997; 49:S52–S55.

22. Truwit C, Barkovich A, Gean-Marton A,Hibri N, Norman D. Loss of the insularribbon: another early CT sign of acutemiddle cerebral artery infarction. Radiol-ogy 1990; 176:801–806.

23. von Kummer R, Holle R, Grzyska U, et al.Interobserver agreement in assessingearly CT signs of middle cerebral arteryinfarction. AJNR Am J Neuroradiol 1996;17:1743–1748.

24. Marks M, Holmgren E, Fox A, Patel S, vonKummer R, Froehlich J. Evaluation ofearly computed tomographic findings inacute ischemic stroke. Stroke 1999; 30:389–392.

25. von Kummer R, Meyding-Lamade U,Forsting M, et al. Sensitivity and prognos-tic value of early computed tomographyin middle cerebral artery trunk occlusion.AJNR Am J Neuroradiol 1994; 15:9–15.

26. Hacke W, Kaste M, Fieschi C, et al. Intra-venous thrombolysis with recombinanttissue plasminogen activator for acutehemispheric stroke: the European Coop-erative Acute Stroke Study (ECASS). JAMA1995; 274:1017–1025.

27. Moulin T, Cattin F, Crepin-Leblond T, etal. Early CT signs in acute middle cerebralartery infarction: predictive value for sub-sequent infarct locations and outcome.Neurology 1996; 47:355–375.

28. Castillo M. Prethrombolysis brain imag-ing: trends and controversies. AJNR Am JNeuroradiol 1997; 18:1830–1834.

29. Gilman S. Imaging of the brain. N EnglJ Med 1998; 338:812–820.

30. Beauchamp N, Barker P, Wang P, van ZijoP. Imaging of acute cerebral ischemia. Ra-diology 1999; 212:307–324.

31. Kidwell C, Alger J, Di Salle F, et al. Diffu-sion MRI in patients with transient isch-emic attacks. Stroke 1999; 30:1174–1180.

32. Lecouvet F, Duprez T, Raymackers J,Peeters A, Cosnard G. Resolution of earlydiffusion-weighted and FLAIR MRI ab-normalities in a patient with TIA. Neurol-ogy 1999; 52:1085–1087.

33. von Kummer R, Allen K, Holle R, et al.Acute stroke: usefulness of early CT find-ings before thrombolytic therapy. Radiol-ogy 1997; 205:327–333.

34. Skriver E, Olsen T. Repeated computedtomography in lacunar infarcts of thebrain. Acta Radiol 1989; 30:1–6.

35. Pantano P, Caramia F, Bozzao L, Dieler C,von Kummer R. Delayed increase in in-farct volume after cerebral ischemia: cor-relations with thrombolytic treatmentand clinical outcome. Stroke 1999; 30:502–507.

100 z Radiology z April 2001 von Kummer et al