ARTICLES

Validity and reliability of a quantitative computed tomography

score in predicting outcome of hyperacute stroke before
thrombolytic therapy

Philip A Barber, Andrew M Demchuk, Jinjin Zhang, Alastair M Buchan, for the ASPECTS Study Group

Summary Introduction
There is evidence that intravenous recombinant tissue
Background Computed tomography (CT) must be done
plasminogen activator (alteplase) is an important
before thrombolytic treatment of hyperacute ischaemic
treatment for acute ischaemic stroke. A systematic
stroke, but the significance of early ischaemic change on review of 17 clinical trials suggested that thrombolysis,
CT is unclear. We tested a quantitative CT score, the though associated with an increased risk of symptomatic
Alberta Stroke Programme Early CT Score (ASPECTS). intracerebral haemorrhage, may increase the proportion
Methods 203 consecutive patients with ischaemic stroke of patients with stroke surviving and able to live
were treated with intravenous alteplase within 3 h of independently.1 The most convincing evidence for the
symptom onset in two North American teaching hospitals. efficacy of alteplase comes from the National Institute of
All pretreatment CT scans were prospectively scored. The Neurological Disorders and Stroke (NINDS) study,2
score divides the middle-cerebral-artery territory into ten
which randomised patients within 3 h of stroke.
However, there is uncertainty about who to treat—for
regions of interest. Primary outcomes were symptomatic
instance, the elderly, patients with severe strokes, and
intracerebral haemorrhage and 3-month functional
those with early ischaemic change on computed
outcome. The sensitivity and specificity of ASPECTS for the
tomography (CT).
primary outcomes were calculated. Logistic regression was
CT in acute stroke is highly sensitive for the detection
used to test the association between the score on of intracerebral haemorrhage. Concern has arisen about
ASPECTS and the primary outcomes. the reliable detection of early ischaemic change on CT
Findings Ischaemic changes on the baseline CT were seen and of its significance in relation to functional outcome
in 117 (75%) of 156 treated patients with anterior- and the risk of symptomatic haemorrhage before the
circulation ischaemia included in the analysis (23 had administration of thrombolytic therapy. The European
ischaemia in the posterior circulation and 24 were treated Cooperative Acute Stroke Study (ECASS) trials
outside the protocol). Baseline ASPECTS value correlated identified the importance of early CT ischaemic changes
inversely with the severity of stroke on the National
in predicting benefit with intravenous thrombolysis.3,4
Patients were eligible only if there was CT ischaemia
Institutes of Health Stroke Scale (r=
0·56, p<0·001).
involving less than a third of the distribution territory of
Baseline ASPECTS value predicted functional outcome and
the middle cerebral artery. This method is not reliable,
symptomatic intracerebral haemorrhage (p<0·001,
however, and even experienced stroke clinicians have
p=0·012, respectively). The sensitivity of ASPECTS for
difficulty in recognising and quantifying such changes by
functional outcome was 0·78 and specificity 0·96; the
currently available methods.5–9 A system is needed to
values for symptomatic intracerebral haemorrhage were improve the general reading of CT scans.
0·90 and 0·62. Agreement between observers for The main aim of our study was to assess the validity,
ASPECTS, with knowledge of the affected hemisphere, was reliability, and usefulness of a standardised quantitative
good ( statistic 0·71–0·89). CT grading system, the Alberta Stroke Programme Early
Interpretation This CT score is simple and reliable and CT Score (ASPECTS), in acute anterior-circulation
identifies stroke patients unlikely to make an independent ischaemic stroke. We hypothesised that by quantification
recovery despite thrombolytic treatment. of early ischaemic change detected on CT scan before
the administration of alteplase, outcome in terms of
Lancet 2000; 355: 1670–74 independence, dependence, and symptomatic
intracerebral haemorrhage could be predicted. We
assumed that other factors may modify the effects of
ASPECTS (ie, serum glucose and age).2,10,11 Such a
score, if reliable and practical, could be applied to future
clinical trials to identify the most appropriate patients for
interventional stroke therapy with thrombolytic or
Departments of Clinical Neurosciences, Alberta Stroke Programme,
University of Calgary, Calgary, Alberta, Canada (P A Barber MRCP,
potential neuroprotective drugs.
A M Demchuk FRCPC, J Zhang MSc, Prof A M Buchan FRCPE) and Stroke
Program, University of Texas, Houston, TX, USA (A M Demchuk) Methods
Correspondence to: Prof Alastair M Buchan, Department of Clinical Patients
Neurosciences, Room 1162, Foothills Hospital, 1403 29th Street Consecutive stroke patients at two North American teaching
NW, Calgary, Alberta, Canada T2N 2T9 hospitals who met established NINDS criteria were treated
(e-mail: abuchan@ucalgary.ca) within 3 h with intravenous alteplase. Only patients thought to

1670 THE LANCET • Vol 355 • May 13, 2000

For personal use only. Not to be reproduced without permission of The Lancet.
 ARTICLES

Figure 1: ASPECTS study form

A=anterior circulation; P=posterior circulation; C=caudate; L=lentiform; IC=internal capsule; I=insular ribbon; MCA=middle cerebral artery; M1=anterior MCA
cortex; M2=MCA cortex lateral to insular ribbon; M3=posterior MCA cortex; M4, M5, and M6 are anterior, lateral, and posterior MCA territories immediately
superior to M1, M2, and M3, rostral to basal ganglia.
Subcortical structures are allotted 3 points (C, L, and IC). MCA cortex is allotted 7 points (insular cortex, M1, M2, M3, M4, M5, and M6).

have anterior-circulation ischaemia (including some severe conditions. Parenchymal hypoattenuation was defined as a
lacunar strokes), at presentation, were included in the analysis. region of abnormally low attenuation of brain structures relative
to attenuation of other parts of the same structures or of the
Procedures contralateral hemisphere. Focal brain swelling was defined as
Before treatment, all patients had a CT brain scan and the score any focal narrowing of the cerebrospinal-fluid space due to
on the National Institutes of Health Stroke Scale (NIHSS) was compression by adjacent brain structures such as effacement of
recorded by a stroke neurologist; in five cases the NIHSS score the cortical sulci or ventricular compression.
had to be extrapolated from the medical records.12 All the CT The affected territory of the middle cerebral artery was
scans were done on fourth-generation scanners with 10 mm graded by a systematic quantitative scoring system, ASPECTS,
slice thickness without contrast enhancement. The NIHSS and according to the rule of one-third or less or more than a
score was recalculated just before the follow-up 24 h CT scan. third of the territory affected by ischaemia,5 done at the same
Both the baseline and 24 h NIHSS scores were done without time. The ASPECTS value was calculated from two standard
knowledge of the results of baseline or 24 h CT scan or the axial CT cuts, one at the level of the thalamus and basal
clinical progress of the patient. ganglia, and one just rostral to the ganglionic structures.
Stroke severity, measured by the NIHSS (an incremental For ASPECTS, the territory of the middle cerebral artery is
scale, with higher scores indicating a more severe neurological allotted 10 points (figure 1). 1 point is subtracted for an area of
deficit), was categorised into five groups: 0–5, 6–10, 11–15, early ischaemic change, such as focal swelling, or parenchymal
16–20, and more than 20. Primary outcome measures were hypoattenuation, for each of the defined regions. A normal CT
score on the modified Rankin scale at 3 months (independence scan has an ASPECTS value of 10 points. A score of 0 indicates
0–2 vs dependence 3–5 and death), and symptomatic diffuse ischaemia throughout the territory of the middle
intracerebral haemorrhage. The Rankin scale score at 3 months cerebral artery.
was assessed by a stroke neurologist or nurse stroke practitioner As a separate part of the study, the reliability within and
who was not aware of the results of the baseline CT, baseline between observers for detection and quantification of early CT
NIHSS score, or the acute clinical events. ischaemia was assessed between three pairs of clinicians: stroke
To detect intracerebral haemorrhage, CT scans were done neurologists (PAB, AMD), radiology trainees (JNS, DK), and
24 h after the onset of stroke and when the patient’s experienced neuroradiologists (WYH, MEH) on a sample of 68
neurological state had deteriorated. A haemorrhage was scans. Each clinician assessed the baseline CT scan in isolation
classified as symptomatic if it had not been seen on a previous at a viewing box initially without access to any clinical
CT scan and there had subsequently been a decline in the information and then at least 3 weeks later with the benefit of
neurological status. only the knowledge of the affected hemisphere.
The baseline CT scan was subsequently assessed with
knowledge of the side affected but without knowledge of the Statistics
baseline stroke severity, 24 h CT scan, or the clinical outcome Spearman’s rank correlation coefficient was used to test the
by a panel of CT reviewers consisting of stroke neurologists association between baseline ASPECTS value and baseline
(PAB, AMD, AMB) and experienced neuroradiologists (RJS, NIHSS score, ASPECTS value at 24 h, and functional
WYH, MEH) for the presence of haemorrhage, parenchymal outcome. A review of the data suggested that baseline
hypoattenuation in the territories of the anterior, middle, and ASPECTS value in two categories (7 and >7) discriminated
posterior cerebral arteries and the vertebrobasilar artery independence from dependence and death. Similarly, previous
distribution, focal swelling, and a hyperdense middle-cerebral- studies1 have suggested that more severe stroke is associated
artery sign. The 24 h scan was assessed under the same with greater risk of poor outcome and symptomatic

THE LANCET • Vol 355 • May 13, 2000 1671

For personal use only. Not to be reproduced without permission of The Lancet.
 ARTICLES

100 treated outside the established NINDS protocol,14 and

were prospectively excluded from the analysis. Four
patients who presented within the 3 h treatment window
were excluded from treatment on the basis that the
80 baseline CT scan revealed ischaemic changes too
extensive to treat with alteplase. At 3 months, all were
dependent or had died. The main reasons for treatment
Proportion (%)

60 to be withheld from patients presenting with cerebral

ischaemia within 3 h were non-disabling or mild
symptoms and rapidly improving deficit.
The final analysis included 156 patients (mean age 68
40 years [SD 14]; 72 women and 84 men) treated within
3 h of stroke onset with intravenous alteplase between
March, 1996, and May, 1999. 117 (75%) of the 156
20 patients had evidence of early ischaemic change on
baseline CT and 141 (90%) had such features on the
24 h scan.
Among the 154 patients with NIHSS scores at
0 presentation, five (3·2%) had scores of 0–5, 41 (26·6%)
10 9 8 7 6 5, 4, 3 2, 1, 0
scores of 6–10, 35 (22·7%) scores of 11–15, 41 (26·6%)
ASPECTS scores of 16–20, and 32 (20·8%) scores above 20. In
Independence Dependence Death total, 31 (19·9%) of these 154 patients had intracerebral
(Rankin 0–2) (Rankin 3–5) haemorrhage confirmed by the 24 h CT; in ten (6·4%)
Figure 2: ASPECTS and functional outcome this haemorrhage was associated with neurological
Spearman correlation coefficient –0·69, p<0·001. deterioration. For two patients, NIHSS and Rankin
scores at 3 months were not available but neither had a
symptomatic intracerebral haemorrhage. At 3 months, of
intracerebral haemorrhage, and therefore, NIHSS scores of 15 the remaining patients, 74 (48·1%) were independent,
or less (less severe) and more than 15 (more severe) were used. 54 (35·1%) were dependent, and 26 (16·9%) had died.
The validity of ASPECTS (7 vs >7), NIHSS score (15 vs
The median baseline ASPECTS value was 8. The
>15), and a third or less versus more than a third of middle-
baseline ASPECTS correlated with the baseline NIHSS
cerebral-artery territory were assessed by calculation of the
sensitivity and specificity for functional outcome and score (r=
0·56 [95% CI
0·46 to
0·78], p<0·001),
symptomatic haemorrhage. Logistic regression was used to the ASPECTS at 24 h (r=0·88 [0·84 to 0·91], p<0·001);
examine the predictive ability of ASPECTS (7 vs >7) in terms functional outcome at 3 months (r=
0·69 [
0·57 to
of functional outcome (independence vs dependence and death)
0·75], p<0·001). A baseline ASPECTS of 7 or less
and symptomatic haemorrhage, in the presence of other sharply discriminated independence, dependence, and
potential risk factors that may modify the effect of ASPECTS death at 3 months (65 patients had scores of 7 or less,
(ie, age and serum glucose). and 89 had scores of more than 7; figure 2).
 statistics13 were used to assess the clinicians’ agreement The sensitivity and specificity of the baseline
with each other for the presence of early CT ischaemic change ASPECTS value, NIHSS score, and the one-third
and for the quantification of early CT ischaemia according to middle-cerebral-artery rule in terms of functional
the one-third middle-cerebral-artery territory rule and outcome (independence vs dependence and death) and
ASPECTS. A  value of 0 indicates agreement no better than symptomatic intracerebral haemorrhage are shown in
chance, and a value of 1 indicates perfect agreement. The
table 1. Because there were very few symptomatic
within-rater agreement was measured for each individual
haemorrhages, the data for ASPECTS and for the one-
observer.
third middle-cerebral-artery rule were identical.
In logistic regression analysis, baseline ASPECTS
Results value was a significant predictor of both functional
Of 203 consecutive patients treated with intravenous outcome (odds ratio 82 [95% CI 23–290], p<0·001) and
alteplase, 23 were identified clinically to have sustained symptomatic haemorrhage (14 [2–117], p=0·012). Age
ischaemia in the posterior circulation, and 24 were (78 vs >78 years) was a significant predictor of

Test Functional outcome Symptomatic haemorrhage

Independent Dependent Sensitivity Specificity Odds ratio Yes No Sensitivity Specificity Odds ratio
or dead (95% CI) (95% CI)
NIHSS
15 (n=81) 56 25 0·69 0·76 6·8 (3·4–14) 6 75 0·40 0·52 0·7 (0·2–2·7)
>15 (n=73) 18 55 4 69
ASPECTS
7 (n=89)* 71 18 0·78 0·96 82 (23–290) 1 90 0·90 0·62 14 (1·8–117)
>7 (n=65) 3 62 9 56
MCA rule
1/3 (n=89)* 67 22 0·73 0·91 25 (10–63) 1 90 0·90 0·62 14 (1·8–117)
>1/3 (n=65) 7 58 9 56
*n=91 for symptomatic haemorrhage analysis.
Table 1: Sensitivity, specificity, and odds ratio for NIHSS score, baseline ASPECTS value, and one-third middle-cerebral-artery (MCA)
rule for functional outcome and symptomatic haemorrhage

1672 THE LANCET • Vol 355 • May 13, 2000

For personal use only. Not to be reproduced without permission of The Lancet.
 ARTICLES

 statistic for agreement between two observers and symptomatic haemorrhage. For patients with
Stroke Radiology Experienced
identical ASPECTS values, inclusion of age and serum
neurologists trainees neuroradiologists glucose in the model further improved the predictive
Without knowledge of clinical information ability. ASPECTS has high sensitivity and specificity for
Evidence of early ischaemia 0·25 0·52 0·20 both functional outcome and symptomatic intracerebral
Side of lesion 0·34 0·58 0·35 haemorrhage. For example, in an individual with an
Evidence of focal brain swelling 0·25 0·41 0·36
1/3 vs >1/3 MCA 0·49 0·14 0·37 ASPECTS value of 7 or less, the risk of symptomatic
ASPECTS 7 vs >7 0·69 0·39 0·47 intracerebral haemorrhage with alteplase is 14 times that
With knowledge of affected hemisphere of patients with a score greater than 7. In patients with
Evidence of early ischaemia 0·45 0·34 0·33 scores above 7 the rate of symptomatic intracerebral
Evidence of focal brain swelling 0·24 0·23 0·42
1/3 vs >1/3 MCA 0·61 0·64 0·52
haemorrhage is 1%, slightly higher than the frequency of
ASPECTS 7 vs >7 0·85 0·71 0·89 symptomatic ischaemic haemorrhagic transformation in
the placebo group in the NINDS trial (0·6%).2
Table 2: Assessment of agreement between observers
ASPECTS is simple and quick and has good between-
observer reliability. However, the results of our study
functional outcome (6·4 [2–20], p=0·002) but not of should not be misinterpreted as showing that patients
symptomatic haemorrhage, after inclusion of ASPECTS who have an ASPECTS of 7 or less should be excluded
value and serum glucose in the model (p=0·68). Serum from thrombolysis, since we cannot know how these
glucose ( 10 vs >10 mmol/L) was a significant patients would have done if they had not received
predictor of symptomatic haemorrhage (4·9 [1–21], treatment. Validation of this score in a randomised
p=0·032) but not functional outcome, after inclusion of controlled study is needed.
ASPECTS value and age in the model (p=0·36). No This scoring method has implications for the design of
interactions were observed between ASPECTS value future stroke trials. It also underlines the arbitrary nature
and age (p=0·86) or serum glucose (p=0·83). Therefore, of use of rigid time windows as the basis for treatment
two multiple logistic regression models were developed. decisions. This concept is supported by the observation
The model that provided the best prediction of from ECASS II of no significant difference in outcome
functional outcome included ASPECTS value and age; between patients treated at 0–3 h and at 3–6 h.4
the model that provided the best prediction of Definition of an arbitrary treatment window ignores
symptomatic intracerebral haemorrhage included evidence of wide variation among individuals in
ASPECTS value and serum glucose. After control for potentially salvageable brain tissue.17 Future “tissue-
ASPECTS value in the two models, the probability of window” studies could assess how a systematic approach
dependence and death was higher for patients older than to early CT ischaemic change, such as ASPECTS, can
78 years than for those aged 78 years and younger, and be used to redefine time windows. For instance, a
the probability of symptomatic haemorrhage was higher patient with a persistent deficit presenting after the
for patients with serum glucose higher than 10 mmol/L defined time window but with a CT scan that shows only
than for those with concentrations of 10 mmol/L or
a small area of early ischaemic change (and therefore a
lower.
high ASPECTS value) might be considered for
Table 2 shows the agreement between observers. The
thrombolytic therapy. In addition to determining trial
within-rater reliability ranged from 0·26 to 0·76 for the
eligibility, the ASPECTS system has the potential to be
one-third middle-cerebral-artery rule and from 0·67 to
used as a surrogate endpoint to provide objective
0·82 for ASPECTS.
evidence in placebo-controlled neuroprotective trials.
Discussion Contributors
Early ischaemic changes identified on CT during the Philip Barber helped to design the study, collected and analysed the
data, and cowrote the paper. Andrew Demchuk helped with the study,
first few hours after stroke onset represent early collected data, and cowrote the paper. Jinjin Zhang helped with the
cytotoxic oedema and possibly the development of design and carried out the analyses. Alastair Buchan invented the
irreversible injury.15 Attempts to assess the prognostic ASPECTS scoring system, designed the study, helped to collect data,
and cowrote the paper. These individuals represent the ASPECTS
value of these early ischaemic changes on CT in terms of Study Group. The other members of the group are: J C Grotta (Stroke
functional outcome and the risk of intracerebral Program, University of Texas-Houston), M Hudon, R Sevick, W Hu,
haemorrhage before administration of thrombolytic J N Scott, D Kaura (Department of Radiology, University of Calgary),
therapy have had misleading results.5 Many have cited S Rose (Community Health Sciences, University of Calgary),
H Karbalai (Medical Student, University of Calgary).
the potential superiority of diffusion-weighted magnetic
resonance imaging over CT, but discrimination of
salvageable from irretreviably injured brain tissue with References
MRI is not yet possible.16 Although diffusion-weighted 1 Wardlaw JM, Yamaguchi T, del Zoppo G. Thrombolytic therapy
magnetic resonance imaging may become the imaging versus control in acute ischaemic stroke. In: Cochrane Library,
issue 4. Oxford: Update Software, 1999.
method of choice, most physicians treating stroke will 2 National Institute of Neurological Disorders and Stroke rt-PA Stroke
depend on CT because it is more accessible. Study Group. Tissue plasminogen activator for acute ischemic
We have shown that a systematic approach to stroke. N Engl J Med 1995; 333: 1581–87.
quantification of early CT ischaemic change can 3 Hacke W, Kaste M, Fieschi C, et al., for the ECASS Study Group.
Intravenous thrombolysis with recombinant tissue plasminogen
improve the identification of cerebral ischemia and is of activator for acute hemispheric stroke: the European Cooperative
prognostic value even before treatment is administered. Acute Stroke Study (ECASS). JAMA 1995; 274:
For patients treated with alteplase, as the ASPECTS 1017–25.
value decreases, the probabilities of dependence, death, 4 Hacke W, Kaste M, Fieschi C, et al. Randomised double-blind
placebo-controlled trial of thrombolytic therapy with intravenous
and symptomatic intracranial haemorrhage increase. alteplase in acute ischaemic stroke (ECASS II). Lancet 1998; 352:
ASPECTS is a predictor for both functional outcome 1245–51.

THE LANCET • Vol 355 • May 13, 2000 1673

For personal use only. Not to be reproduced without permission of The Lancet.
 ARTICLES

5 von Kummer R, Allen KL, Holle R, et al. Acute stroke: usefulness of 11 Demchuk AM, Morgenstern LB, Krieger D, et al. Is baseline serum
early CT findings before thrombolytic therapy. Radiology 1997; 205: glucose a predictor of hemorrhage after rtPA therapy in acute stroke?
327–33. Stroke 1999; 30: 34–39.
6 Dippel DWJ, Du RY, van Beest Holle M, et al. The validity and 12 Kasner S, Chalela JA, Luciano JM, et al. Reliability and validity of
reliability of early infarct signs on CT in acute ischaemic stroke. estimating the NIH stroke scale score from medical records. Stroke
Cerebrovasc Dis 1997; 7 (suppl 4 ): 15 (abstr). 1999; 30: 1534–37.
7 Schriger D, Kalafut M, Starkman S, et al. Cranial computed 13 Armitage P, Berry G. Statistical methods in medical research.
tomography interpretation in acute stroke. Physicians’ accuracy in Oxford: Blackwell Scientific Publications, 1987.
determining eligibility for thrombolytic therapy. JAMA 1998; 279: 14 Buchan AM, Barber PA, Newcommon NJ, et al. Effectiveness of t-
1293–97. PA in acute ischemic stroke: outcome relates to appropriateness.
8 Grotta J, Chiu D, Lu M, et al. Agreement and variability in the Neurology 2000; 54: 679–84.
interpretation of early CT changes in stroke patients qualifying for 15 del Zoppo GJ, von Kummer R; Hamann GF. Ischaemic damage of
intravenous rtPA. Stroke 1999; 30: 1528–33. brain microvessels: inherent risks for thrombolytic treatment in
9 Wardlaw JM, Dorman PJ, Lewis SC, Sandercock PAG. Can stroke stroke. J Neurol Neurosurg Psychiatry 1998; 65: 1–9.
physicians and neurologists identify signs of early cerebral infarction 16 Baird AE, Warach S. Magnetic resonance imaging of acute stroke.
on CT? J Neurol Neurosurg Psychiatry 1999; 67: 651–53. J Cereb Blood Flow Metab 1998; 18: 583–609.
10 The NINDS rt-PA Stroke Study Group. Intracerebral hemorrhage 17 Baron JC, von Kummer R, del Zoppo GJ. Treatment of acute
after intravenous t-PA for ischemic stroke. Stroke 1997; 28: ischemic stroke: challenging the concept of a rigid and universal time
2109–18. window. Stroke 1995; 26: 2219–21.

1674 THE LANCET • Vol 355 • May 13, 2000

For personal use only. Not to be reproduced without permission of The Lancet.