Original Paper Received: January 21, 2014 Accepted: May 15, 2014 Published online: September 16, 2014

Cerebrovasc Dis 2014;38:39–45 DOI: 10.1159/000363619

Low Alberta Stroke Program Early CT Score (ASPECTS) Associated with Malignant Middle Cerebral Artery Infarction Caroline MacCallum a Leonid Churilov c Peter Mitchell b Richard Dowling b Bernard Yan a   

 

 

 

 

a

Melbourne Brain Centre at Royal Melbourne Hospital, b Department of Radiology at Royal Melbourne Hospital, The University of Melbourne, c The Florey Institute of Neuroscience and Mental Health, Melbourne, Australia  

 

 

Abstract Background: Early decompressive hemicraniectomy following malignant middle cerebral artery (MCA) infarction reduces mortality and improves clinical outcome. Imaging predictors of malignant infarction may serve as ‘red flags’, prompting intensive neurological monitoring and timely intervention. Our objective is to investigate whether lower ASPECTS (Alberta Stroke Program Early CT Score) is associated with malignant MCA infarction. Methods: A retrospective cohort study of all patients with MCA territory ischemic strokes who were admitted to the Royal Melbourne Hospital (RMH) between 1 January 2009 and 31 December 2009 (226 patients included). The main outcome measures were ASPECTS on admission for each patient and the development of malignant MCA infarction. Results: One-hundredand-eight patients out of 226 (48%) developed malignant MCA infarction. Good (>0.8) inter-rater agreement between observers scoring ASPECTS was observed using weighted kappa, intra-class correlation coefficient and Lin’s concordance coefficients. Using receiver operating characteristic (ROC) curve analysis, we validated that ASPECTS 7 was the optimal cut-off score to determine progression to malignant

© 2014 S. Karger AG, Basel 1015–9770/14/0381–0039$39.50/0 E-Mail [email protected] www.karger.com/ced

infarction, providing 50% sensitivity and 86% specificity. One hundred and fifty six patients had ASPECTS >7 (69%) and 70 patients had ASPECTS ≤7 (31%). Patients with ASPECTS ≤7 were significantly younger than those with ASPECTS >7, with the median age of each group being 72.5 and 78 respectively (p = 0.02); otherwise the groups were well-matched. With ASPECTS ≤7, 54 out of 70 patients (77%) developed malignant MCA infarction, compared with 54 out of 156 patients (35%) with ASPECTS >7 (age-adjusted OR = 0.12, 95% CI: 0.06, 0.25; p < 0.0001). If ASPECTS ≤7 is a positive result, then the positive predictive value is 77% and the negative predictive value is 65%. The median ASPECTS for developing malignant MCA infarction was 7.5 (IQR: 5 to 10), while the median ASPECTS for not developing MCA infarction was 10 (IQR: 8 to 10), resulting in a significant age-adjusted median difference of 2 (95% CI: 0.8, 3.2; p < 0.0001). We also found that coma on admission is associated with the development of malignant MCA infarction (OR = 22.63, 95% CI: 1.3, 393.7; p = 0.0323) and that a history of hypertension is not associated with the development of malignant MCA infarction (OR = 0.9707, 95% CI: 0.54, 1.75; p = 0.9213). Conclusions: ASPECTS ≤7 on initial brain CT in a patient with MCA infarction is associated with the development of malignant MCA infarction. We recommend close monitoring of, and early consideration of decompressive hemicraniectomy for, acute stroke patients with ASPECTS ≤7. © 2014 S. Karger AG, Basel

Bernard Yan, MBBS, FRACP Melbourne Brain Centre Royal Melbourne Hospital Grattan Street, Parkville, Vic. 3050 (Australia) E-Mail bernard.yan @ mh.org.au

Downloaded by: University of Utah 155.97.178.73 - 12/2/2014 1:08:05 PM

Key Words Stroke · CT · Malignant middle cerebral artery infarction · ASPECTS · Hemicraniectomy

Malignant middle cerebral artery (MCA) infarction refers to an MCA territory infarction in which there is extensive cerebral edema leading to increased intracranial pressure and brain herniation [1, 2]. The conversion to malignant MCA infarction occurs in up to 50% of all MCA infarctions [3–6]. The consequences of malignant MCA infarction are dire, with a mortality rate of up to 80% without surgical intervention [3, 7], and survivors suffering severe disabilities. The pooled analysis of three European randomized controlled trials demonstrated that decompressive hemicraniectomy to treat malignant MCA infarction reduces mortality and improves functional outcomes [5, 8–10]. The general consensus is that early decompression is important for optimal outcomes [11–15]. To ensure prompt treatment, it is critical to determine the predictors, both clinical and radiological, of whether an MCA infarction will become malignant. Previous studies have investigated predictors but have reported poor positive predictive value [16–19]. ASPECTS (Alberta Stroke Program Early CT score) is a scoring system that provides a methodical and reproducible way to identify the extent of dead tissue and cytotoxic edema on CT following a stroke [20, 21]. The original ASPECTS study showed that ASPECTS of 7 or below differentiated patients who were highly unlikely to achieve independent functional outcome [22–28], although this study focused on stroke patients who received thrombolytic therapy. A later study similarly showed that ASPECTS ≤7 for an early CT predicts a high risk for parenchymal hematoma following acute ischemic stroke, in patients without thrombolysis [29]. ASPECTS may be used as a diagnostic tool for subsequent malignant MCA infarction. Our study aims to investigate the relationship between ASPECTS and malignant MCA infarction. We hypothesize that (a) ASPECTS 7 provides an optimal diagnostic cut-off point for subsequent malignant MCA infarction and (b) lower ASPECTS is associated with malignant MCA infarction.

Methods Patients This is a retrospective study of all patients with MCA territory ischemic strokes who were admitted to the Royal Melbourne Hospital (RMH) between 1 January 2009 and 31 December 2009. The inclusion criteria for this study were (i) acute MCA ischemic stroke, (ii) clinical data availability, and (iii) imaging data available,

40

Cerebrovasc Dis 2014;38:39–45 DOI: 10.1159/000363619

specifically data obtained from an initial CT scan on admission and a follow-up CT scan or MRI scan during the admission. The exclusion criterion was a hemorrhagic stroke. Data We collected clinical data for each patient by using a specific stroke registry maintained at RMH, and through extraction from medical records. These data include patient demographics; cardiovascular risk factors; stroke type using the TOAST classification; neurological features at onset and deterioration; Glasgow Coma Scale (GCS) at onset and deterioration; treatment received including intravenous or intra-arterial thrombolysis, intra-arterial clot retrieval or decompressive hemicraniectomy; the modified Rankin Scale on admission, on discharge and three months after onset of stroke; and mortality status one month after presentation. Imaging Baseline non-contrast CT scans were performed on admission. Follow-up non-contrast CT scans or MRI scans were performed during admission; we only used the next-in-time follow-up CT scan or MRI scan. This neuroimaging data were acquired through the Pictures and Archiving System (PACS) and were retrospectively reviewed. All CT scans were independently reviewed by three independent observers (PJM, RJD and BY) who were blinded to the clinical information. The first two observers are neuroradiologists, one with twenty years (PJM) and the other with ten years (RJD) neuroradiology experience respectively. The third observer is a neurologist and neurointerventionist (BY), with ten years neuroradiology experience. The Alberta Stroke Program Early CT Score (ASPECTS) system was used to analyze the admission CT scans. ASPECTS is a 10-point quantitative topographic CT scan score, which is determined by evaluating two standardized regions of the MCA territory, the basal ganglia level and the supraganglionic level, for ischemic changes [22]. These areas are divided into 10 distinct regions. Evidence of early ischemic change in one of these areas amounts to 1 point. In order to compute ASPECTS, a patient starts with a maximum score of 10 (no ischemic changes) and 1 point is subtracted for each area involving ischemic change (see fig. 1 for examples). Definition of Malignant MCA Infarction The primary endpoint of malignant MCA infarction was defined as clinical signs of large middle cerebral artery territory infarction with either (a) secondary neurological deterioration including decline of consciousness (GCS fall of 1 or more points) or new neurological deficit (anisocoria, Babinski reflex, worsening hemiparesis, dysarthria, dysphasia) (‘neurological deterioration’), or (b) GCS ≤12 based on GCS on admission (‘low admission GCS’). Statistical Analysis Weighted kappa, intra-class correlation coefficient and Lin’s concordance coefficient were used to estimate the inter-rater agreement between the ASPECTS scorers. Youden’s index (J = sensitivity+specificity-1) for the receiver operating characteristic (ROC) curve was used to validate ASPECTS 7 as the optimal cut-off level of ASPECTS to allow for differentiation between malignant and nonmalignant MCA infarctions. Baseline characteristics, such as demographics, cardiovascular risk factors, stroke type, GCS at onset and deterioration, thrombolysis, and mortality status at one

MacCallum/Churilov/Mitchell/Dowling/ Yan

Downloaded by: University of Utah 155.97.178.73 - 12/2/2014 1:08:05 PM

Introduction

Results

month were compared between patients with ASPECTS ≤7 and ASPECTS >7 based on data from admission CT scan using either Wilcoxon-Mann-Whitney ranksum test or Fisher’s exact test depending on the nature of the underlying distribution. The association between dichotomized ASPECTS (≤7 vs. >7) and malignant MCA infarction was investigated using Fisher’s exact test and logistic regression. In addition, the association between ASPECTS on the full scale and malignant MCA infarction was investigated using Wilcoxon-Mann-Whitney ranksum test and median regression. All statistical analysis was performed using Stata v12 IC software. p values less than 0.05 were considered indicative of statistical significance.

Two hundred and fifty nine patients with MCA infarction who presented to Royal Melbourne Hospital between 1 January 2009 and 31 December 2009 were eligible for the final analysis. Twelve were excluded due to lack of imaging and 21 were excluded due to incomplete data. Overall, 226 patients with complete imaging and data were included in our analysis. In order to exclude selection bias, we compared the baseline characteristics of the included patients as against the characteristics of the excluded patients. We found that these results were comparable, except for the gender factor where there was a significant difference with more males than females in the included group (55%), but fewer males in the excluded group (28%) (p = 0.01). The demographics for our patients are recorded in table 1. Out of 226 patients, 124 patients (55%) were male and the median age was 74.20 (IQR 69 to 83.8). In terms of cardiovascular risk factors, 71 (31%) were smokers, 85 (38%) had atrial fibrillation, 70 (31%) had ischemic heart disease, 55 (24%) had diabetes mellitus, 166 (73%) had hypertension, 56 (25%) had suffered a previous stroke, and 19 (8%) had suffered a previous TIA. We found that the median ASPECTS for the whole sample was 9 (IQR 7 to 10). One hundred and eight patients out of 226 (48%) developed malignant MCA infarction. Out of 108 patients with malignant MCA infarction, 69 (64%) patients were classified as being malignant due to GCS ≤12 on presentation, 28 (26%) due to a fall in GCS of 1 point or more, 8 (7%) due to a new neurological deficit, and 3 (3%) due to a simultaneous fall in GCS and a new neurological deficit. The median time of the first CT scan for each patient after ischemic onset was 3 h and 32 minutes (IQR: 1 hour 28 minutes, 10 hours). For patients who developed malignant MCA infarction, the median time of malignant MCA infarction after ischemic onset was three hours and 45 minutes (IQR: 1 hour 57 minutes, 9 hours 28 minutes). The scans were analyzed by three independent observers. There was good inter-rater agreement between the scores, with agreement above 0.8 using three methods – weighted kappa (0.82, 95% CI: 0.724, 0.89), intra-class coefficient (0.82, 95% CI: 0.77, 0.86), and Lin’s concordance correlation coefficient (0.819, 95% CI: 0.77, 0.861). Therefore, for the sake of consistency, only one observer’s scores were used for the analysis. Using the receiver operating characteristic curve analysis, we validated that ASPECTS 7 was the most appropriate cut-off score to allow for differentiation between malignant and nonmalignant MCA infarctions, as described earlier in the litera-

Low ASPECTS Associated with Malignant MCA Infarction

Cerebrovasc Dis 2014;38:39–45 DOI: 10.1159/000363619

b

c

Fig. 1. Examples of CT scans scored with ASPECTS. a CT scan scored as ASPECTS 10. b CT scan scored as ASPECTS 7. c CT scan

scored as ASPECTS 3. Arrows show areas of hypodensity representing ischemia.

41

Downloaded by: University of Utah 155.97.178.73 - 12/2/2014 1:08:05 PM

a

Table 1. Patient demographics

Characteristic

All patients (n = 226)

ASPECTS ≤7 (n = 70)

Age (median, IQR) Sex (male), n (%) Cardiovascular risk factors Smoking AF IHD Diabetes Previous stroke Previous TIA Hypertension Etiology Atherothrombotic Cardioembolic Undetermined Other Treatment IV thrombolysis

74.20 (69, 83.8) 124 (55%)

72.5 (61, 84) 41 (58%)

ASPECTS >7 (n = 156) 78 (71, 83.5) 83 (53%)

p* 0.02 0.47

71 (31%) 85 (38%) 70 (31%) 55 (24%) 56 (25%) 19 (8%) 166 (73%)

20 (29%) 24 (34%) 23 (33%) 13 (19%) 15 (21%) 4 (6%) 49 (70%)

51 (33%) 61 (39%) 47 (30%) 42 (27%) 41 (26%) 15 (10%) 117 (75%)

0.64 0.55 0.75 0.24 0.5 0.44 0.52

31 (14%) 81 (36%) 115 (51%) 4 (2%)

12 (17%) 25 (36%) 35 (50%) 3 (4%)

19 (12%) 56 (36%) 80 (51%) 1 (1%)

0.4 >0.99 0.86 0.09

60 (27%)

21 (30%)

39 (25%)

0.42

AF = Atrial fibrillation; IHD = ischemic heart disease; TIA = transient ischemic attack; IV = intravenous; Other = stroke of other etiology, examples including carotid dissection and cerebral vasculitis. * p values are from Fisher’s exact test for all comparisons apart from the age, where p value is from Wilcoxon-Mann-Whitney test.

MMI NMMI

≤7 (n = 70) (positive result)

>7 (n = 156) (negative result)

p

54 (77%) 16 (23%)

54 (35%) 102 (65%)

7 (69%) and 70 patients had an ASPECTS score ≤7 (31%). Patients with ASPECTS ≤7 were significantly younger than those with ASPECTS >7, with the median age of each group 42

Cerebrovasc Dis 2014;38:39–45 DOI: 10.1159/000363619

being 72.5 and 78 respectively (p = 0.02). The groups were well-matched in terms of the other demographics recorded. Table  2 compares the distribution of MCA infarction  in the groups of ASPECTS ≤7 and ASPECTS >7. There were 70 patients with an ASPECTS score ≤7, out of which 54 (77%) developed malignant MCA infarction and only 16 (23%) did not. Out of 156 patients with an ASPECTS score >7, 54 (35%) patients developed malignant MCA infarction and 102 (65%) patients did not. This is a statistically significant difference: OR = 0.16 (95% CI: 0.08, 0.324, p < 0.0001), indicating that a patient with ASPECTS >7 is much less likely to develop malignant MCA infarction than a patient with ASPECTS ≤7. This difference remains statistically significant when adjusted for age (OR = 0.12, 95% CI: 0.06, 0.25, logistic regression p < 0.0001). On multivariable analysis, this difference also remains statistically significant when adjusted for other variables (other variables are the baseline characteristics listed in table 1). If ASPECTS ≤7 is considered a positive test result, then the positive predictive value of the test for a malignant MCA infarction is 77% and the negative predictive value is 65%. When analyzed on full ASPECTS scale, we also found that the median ASPECTS relating to the initial CT scan MacCallum/Churilov/Mitchell/Dowling/ Yan

Downloaded by: University of Utah 155.97.178.73 - 12/2/2014 1:08:05 PM

Table 2. Fisher’s Exact test – ASPECTS dichotomized to ≤7 and >7

We found that ASPECTS ≤7 on initial CT scan was associated with the development of malignant MCA infarction. In this study, we investigated the use of ASPECTS as a tool for predicting malignant MCA infarction. We hypothesized that lower ASPECTS on admission is associated with malignant MCA infarction, and aimed to validate that an ASPECTS score of 7 is an optimal diagnostic cut-off point for malignant MCA infarction. The original ASPECTS study showed that 7 is the optimal cut-off point for predicting functional outcome. We used the ROC curve to validate the most appropriate cutoff point in the 10-point ASPECTS system. We performed statistical analysis on our data on the basis that ASPECTS ≤7 is a positive test result and ASPECTS >7 is a negative test result, and that malignant MCA infarction is a dependent outcome and nonmalignant MCA infarction is a negative outcome (see table 2). On this basis, the sensitivity of this test is poor, at 50%. The implication is that if the test is positive, namely ASPECTS ≤7, then in 50% of cases the patient will develop malignant MCA infarction. The specificity of this test is better, with a value of 86%. It follows that if the test is negative (ASPECTS >7), then 86% of cases the patient will not develop malignant MCA infarction. The positive predictive value of this diagnostic test is 77%, supporting a 77% probability that a patient will develop malignant MCA infarction in the setting of ASPECTS ≤7. The lack of sensitivity means that a negative test does not effectively rule out the risk of malignant infarction. Therefore, there is a significant group of pa-

tients who may need to be monitored, but who are not identified as requiring monitoring by the ASPECTS ≤7 cut-off test. While the ASPECTS ≤7 cut-off test is useful in identifying one group of patients who are at risk of developing malignant infarction, clinicians should still regularly assess patients with ASPECTS >7 and look for the development of malignant infarction. Our results also showed that the median ASPECTS for patients who developed malignant MCA was 7.5 (IQR: 5 to 10), while the median ASPECTS for patients who did not was 10 (IQR: 8 to 10) (p < 0.001). In terms of demographics, it is interesting to note that patients with ASPECTS ≤7 are on average younger than patients with ASPECTS >7. The aim of this project was to identify a diagnostic tool for the development of malignant MCA infarction in order to guide treatment decisions regarding hemicraniectomy and to determine which patients to monitor closely. There have been many studies with a similar aim. Clinical predictors include high NIHSS on admission [19], coma on admission [16], early nausea and vomiting [17], history of hypertension [30], and history of heart failure [30], but these provide poor predictive value. Serum markers including high white cell count [20] and the presence of S100B [31, 32] are also predictive of poor outcome in MCA infarction, but again the predictive value is low. There are many studies relating to imaging predictors, which include an infarct size of >50% of the MCA territory on early CT [17, 30, 33–35], the ratio of infarct volume to CSF volume using perfusion CT and routine CT [6], the lesion volume of >145 ml on diffusion-weighted imaging (DWI) [18, 19], and the volume of severe perfusional deficit is more predictive of malignant edema than the total perfusional deficit, as measured by 11Cflumazenil PET [36]. The first two predictors are problematic because measuring the infarct size on CT is a crude test, which is open to inter-observer variability. Another problem with the latter three predictors is that DWI-MRI, perfusion CT and PET are not widely available imaging tools. Further, while the specificity relating to the infarct size predictors using CT and DWI-MRI is high, both these predicting systems have low sensitivity. It is noted that using the ratio of infarct volume to CSF volume as a predictor provides both good specificity and sensitivity (97% for both). Our method offers an alternative to these challenges. First, ASPECTS is a system which attempts to provide good inter-rater agreement between users. This was confirmed in our study where agreement was excellent using three different statistical tests (weighted kappa = 0.819,

Low ASPECTS Associated with Malignant MCA Infarction

Cerebrovasc Dis 2014;38:39–45 DOI: 10.1159/000363619

Discussion

43

Downloaded by: University of Utah 155.97.178.73 - 12/2/2014 1:08:05 PM

for a patient who did develop malignant MCA infarction was 7.5 (IQR: 5, 10), while the median ASPECTS for a patient who did not develop malignant MCA infarction was 10 (IQR: 8, 10), resulting in the median difference of 3 (95% CI: 2, 4; p < 0.0001). This result is statistically significant using both Wilcoxon-Mann-Whitney test and a median regression model, with a lower ASPECTS indicating that a patient is more likely to develop malignant MCA infarction. This difference also remains statistically significant when adjusted for age (median difference = 2, 95% CI: 0.8, 3.2, median regression p < 0.0001). We also found that coma on admission is associated with the development of malignant MCA infarction (OR = 22.63, 95% CI: 1.3, 393.7; p = 0.0323) and that a history of hypertension is not associated with the development of malignant MCA infarction (OR = 0.9707, 95% CI: 0.54, 1.75; p = 0.9213).

CI 0.724–0.890, intra-class coefficient = 0.82, 0.77–0.86, and Lin’s concordance correlation coefficient = 0.819, CI 0.77–0.861). Second, CT is widely available in clinical practice, meaning that the ASPECTS system can be used in more remote facilities where more advanced imaging technologies may not be available. Third, as with the other imaging predictors, an ASPECTS score ≤7 provides good specificity (86%) but poor sensitivity (50%), which makes it a practical test to determine which patients to monitor closely for the development of malignant infarction, but it does not reliably predict that patients with ASPECTS >7 will not develop malignant MCA infarction. There are strengths and limitations in this study. One strength is that CT is a widely available imaging modality. Another strength is that we used three independent investigators who were blinded for the ASPECTS scoring. One limitation is that this is a retrospective study, and selection bias is a problem. The study is limited by the population of patients included in the analysis, in that there may be selection bias because some patients were not included on the basis of lack of imaging and lack of clinical data. We compared the baseline demographics of these patients and found that the only significant difference between the included and excluded patients was

gender, because there was a higher proportion of males in the included patients (55 vs. 28%, p = 0.01). The prediction of whether an MCA infarction becomes malignant is important for the ongoing treatment and close monitoring of the patient, and specifically to understand whether that patient will be a candidate for decompressive hemicraniectomy. It is generally understood that there are better outcomes associated with early surgery; therefore, early predictors are clinically useful. We found that there is an association between low ASPECTS on admission CT and development of malignant MCA infarction, and specifically found that ASPECTS

Low Alberta Stroke Program Early CT score (ASPECTS) associated with malignant middle cerebral artery infarction.

Early decompressive hemicraniectomy following malignant middle cerebral artery (MCA) infarction reduces mortality and improves clinical outcome. Imagi...
674KB Sizes 0 Downloads 4 Views