INCREASED BLOOD BRAIN BARRIER PERMEABILITY ON PERFUSION CT MIGHT PREDICT MALIGNANT MIDDLE CEREBRAL ARTERY INFARCTION

Abstract

Background and Purpose

Perfusion computerized tomography (PCT) has been used to assess the extent of blood brain barrier (BBB) breakdown. The purpose of this study was to determine the predictive value of (BBB) permeability (BBBP) measured using PCT for development of malignant middle cerebral artery infarction (MMCA) requiring hemicraniectomy (HC).

Methods

We retrospectively identified patients from our stroke registry that had MCA infarction and were evaluated with admission PCT. BBBP and cerebral blood volume (CBV) maps were generated and infarct volumes calculated. Clinical and radiographic characteristics were compared between those who underwent HC versus those who did not undergo hemicraniectomy (NHC).

Results

122 patients (12 (HC), 110 (NHC)) were identified. 12 patients who underwent HC had developed edema, midline shift or infarct expansion. Infarct permeability area (IP_area), infarct CBV area (ICBV_area), and infarct volumes were significantly different (p<0.018, p<0.0211, p<0.0001, p<0.0014) between HC and NHC groups. Age (p=0.03) and admission National Institutes of Health Stroke Scale (NIHSS) (p=0.0029) were found to be independent predictors for HC. Using logistic regression modeling, there was an association between increased IP_area and HC. The odds ratio for HC based on a 5, 10, 15 or 20 cm² increase in IP_area were 1.179, 1.390, 1.638 or 1.932, respectively (95% CI 1.035-1.343, 1.071-1.804, 1.108-2.423, 1.146-3.255).

Conclusion

Increased IP_area is associated with an increased likelihood for undergoing HC. Since early HC for MMCA has been associated with better outcomes, the IP_area on admission PCT might be a useful tool to predict MMCA and need for HC.

INTRODUCTION

About 10–15% of all patients with cerebral infarction in the territory of the middle cerebral artery (MCA) suffer from progressive clinical deterioration because of increasing brain swelling, raised intracranial pressure, and brain herniation. This patient population constitutes a particularly difficult challenge for clinicians charged with their care.^1–3 Because of the limitations of medical therapies; decompressive surgery is an option for patients with neurological deterioration due to large hemispheric infarction and edema. The rationale of this therapy is to prevent brain tissue shifts and to normalize intracranial pressure, and thereby preserve cerebral blood flow and prevent secondary damage.⁴ Hemicraniectomy (HC) for malignant middle cerebral artery infarctions (MMCA) has been shown to be an effective treatment.⁵ Currently, clinical data and early cranial tomography (CT) does not reliably identify patients that will develop MMCA. Prognostic criteria that might permit early identification of patients at risk for such “malignant” infarction are important.^6-8

Perfusion CT (PCT) is a modern imaging technique that has been proposed for evaluating acute stroke patients at the time of their emergent evaluation. PCT involves the sequential acquisition of CT images performed in cine mode during the intravenous administration of iodinated contrast material and has been reported to allow for accurate quantitative assessment of cerebral blood flow (CBF) and cerebral blood volume (CBV).⁹ It also has been validated in determining final infarct volumes and measuring potentially salvageable tissue using different PCT maps.¹⁰ By extending the PCT acquisition time window, blood-brain-barrier permeability (BBBP) can also be obtained.¹¹ PCT techniques for measuring BBBP have been shown to be achievable in tumor models in animals^{12, 13} and in a small human series ¹⁴, but only recently has it been applied to acute stroke imaging.¹⁵

The aim of this retrospective study was to determine whether permeability maps, generated from admission PCT, can be used to predict MMCA. We hypothesized that patients with an increased area of permeability defect would be at higher risk for development of MMCA.

METHODS

We retrospectively reviewed our registry from 8/07 to 8/09 and identified 268 consecutive patients diagnosed with MCA infarctions (18 patients in the HC group and 250 patients in non-hemicraniectomy (NHC) group). We included in our analysis patients that underwent PCT on admission with interpretable permeability maps. Two patients in the HC and 31 patients in NHC group had uninterpretable PCT maps. Three patients in HC and 76 patients in the NHC groups did not have PCT. Patients who had their care withdrawn were also excluded from the analysis. In total, 122 patients fit our criteria: 12 underwent HC and 110 in the NHC group. Demographic characteristics and baseline clinical information including admission NIHSS, admission glucose level, HbA1c, lipids, the timing of MRI acquisition and PCT were extracted from the medical records. Clinical, laboratory and radiographic characteristics were analyzed between the two groups.

Calculation of Infarct Volumes

All MRI scans were performed before decompression in the HC group. One of the investigators was blinded to clinical outcomes and independently reviewed all the DWI images to verify infarct volumes. Infarct volumes were measured by hand-drawn regions of interest around each area of infarct in every slice separately. The regions of interest area were calculated automatically by the Picture Archiving and Communication Systems software (General Electric Centricity workstation) and then multiplied by the slice thickness (5 mm for DWI). Seven patients (58.3%) in the HC group and 85 patients (77.3%) in the NHC group had DWI.

Calculation of Permeability and CBV volume

Permeability color maps were retrospectively generated from PCT data using a modified Patlak method. PCT was performed on a Siemens Somatom Sensation 40 scanner. After intravenous administration of 40 cc iodinated contrast at 8cc/sec by a power injector into an antecubital vein (Omnipaque, 350/40 mL ), images were acquired without delay. PCT imaging parameters were 80 kVp, 270 mAs, 1.2 mm section collimation. Slice thickness was 9.6 mm acquired 3 slices at a time for 75 seconds. The modified Patlak model (in Siemens Syngo Neuro PCT) calculates vascular permeability (PS) with CT enhancement values after the peak contrast enhancement has been reached and utilizing a built-in delay correction algorithm. This allows estimation of PS during the end of the first-passage of contrast but avoids erroneous elevation of PS values secondary to delayed arrival of contrast which would occur in a standard application of the Patlak model to PCT data in the setting of acute cerebral ischemia¹⁶. From the PCT data, permeability maps were calculated using Syngo Neuro PCT software which sets the threshold for generating the permeability maps at 0.5 ml/10 ml/min. We defined the perfusion abnormality as the region analyzed on the permeability maps, IP as the weighted mean of the permeability value, and IP_area as the area of the region that had perfusion abnormality. Each permeability map was analyzed by demarcating trace lines around the area of perfusion abnormality on each of the 3 slices. IP and IP_area values were measured by hand-drawn regions of interest in every slice separately. The weighted mean of IP and IP_area of the 3 slices were calculated, respectively and used for statistical analysis. Contra-lateral non-ischemic hemisphere was used as a control (CP).

Using the Siemens Syngo Neuro software, CBV was calculated along with CBF using the maximal slope model which has been shown to yield lesion volumes that are similar to delay-insensitive deconvolution techniques¹⁷.The hypoperfused areas on CBV maps were defined as volume abnormality. The infarct core was outlined on CBV maps as a severely hypoperfused area displayed by 2 colors in the color bar (eg, purple and blue). The CBV threshold was defined at 2.0 ml×100 g⁻¹. Each CBV map was analyzed by demarcating trace lines around the area of volume abnormality on each of the 3 slices by hand-drawn regions of interest. The area of volume abnormality was defined as CBV_area. The weighted mean of CBV and CBV_area were calculated in each of the 3 slices respectively and used for statistical analysis. Representative images of CBV and permeability are shown in fig 1 from a patient who developed malignant infarction and underwent HC.

Figure 1.

A 75-year-old right-handed man who presented with sudden onset of left hemiparesis and dysarthria. (A)Admission CT; (B)Admission CBV color maps, showing hypoperfusion with right MCA acute ischemic stroke (AIS). (C)Admission Permeability color maps, showed permeability abnormality with right MCA AIS. (D)Subsequent head CT before HC, showing malignant MCA infarction.

Statistical Analysis

Means with standard deviations or medians for continuous variables and proportions for categorical variables were used. The differences were assessed using t-tests, chi-square tests, Fisher exact test or Mann-Whitney U test. A significance level of 0.05 was used to assess statistical difference. We used multivariate analysis using logistic regression to obtain odds ratios to determine associations, controlling for potential confounders. The statistical analysis was performed using SAS 9.2.

RESULTS

Patients

We included 122 patients in total, 12 HC and 110 NHC. Of the baseline characteristics analyzed, mean age (56.08+ 13.20 vs 65.40+13.66; p<0.03) was statistically different between the HC and NHC groups. Patients in the HC group were significantly younger than the NHC group. Other baseline characteristics such as gender, race and risk factors were not statistically different between the 2 groups (Table 1). Laboratory parameters (mean serum glucose at admission, Hb A1c level, serum total cholesterol level, triglyceride, HDL, LDL) assessed at baseline were not significantly different between HC and NHC groups.

Table 1.

Demographics and Preexisting Conditions

	Hemicraniectomy group (n=12)	Non-Hemicraniectomy group (n=110)	P Value
Age in yrs (mean ± SD)	56.08 ± 13.20	65.40±13.66	0.03 *
Gender			0.66 †
Female (%)	7 (58.33)	57(51.82)
Male (%)	5 (41.67)	53 (48.18)
Race			0.11‡
Black (%)	4 (33.33)	27 (25.71)
Hispanic (%)	4 (33.33)	12 (11.43)
White (%)	4 (33.33)	63 (60)
Others (%)	0 (0)	3 (2.86)
Risk factors
Hypertension (%)	9(75)	77(70)	0.71†
Diabetes Mellitus (%)	2( 16.67)	29(26.36)	0.72‡
Coroner disease (%)	3 (25)	24 (21.82)	0.72‡
Atrial Fibrillation (%)	4 (33.33)	18 (16.36)	0.22‡
Smoking (%)	3 (25)	37 (33.64)	0.74‡

^*TT=Student’s t test;

^†CS= chi-square test;

^‡FET=Fisher’s Exact Test

Clinical Factors

Baseline stroke severity, as measured by median NIHSS scores, was significantly higher in the HC group than the NHC group (19 (11-29) vs 11 (0-40); p<0.0029). The mean time from stroke onset to PCT in minutes in the HC and NHC groups were 310.55±306.75 and 502.56 ±845.90, respectively. There was no significant difference in time to PCT scanning between two groups (p=0.19). Seven (58.3%) patients in the HC group and 85 (77.3%) patients in the NHC group had DWI (p<0.0035). Onset to acquired DWI images was not statistically different (877.6+ 712.7 vs 975.5 + 709.4 minutes, p=0.7680) (Table 2).

Table 2.

Clinical Characteristic

Clinical Features	Hemicraniectomy group (n=12)	Non-Hemicraniectomy group (n=110)	P Value
Admission NIHSS median (min-max)	19 (11-29)	11( 0- 40)	0.0029§
Onset to PCT in minutes (mean ± SD)	310.55 ± 306.75	502.56 ± 845.90	0.19*
Onset to DWI in minutes (mean ± SD)	877.6± 712.7	975.5 ± 709.4	0.76*
iv t-PA given (%)	4 (33.33)	60(54.55)	0.22‡

NIHSS=National Institutes of Health Stroke Scale;

^*TT=Student’s t test;

^§TST= Two –sample test,

^‡FET=Fisher’s Exact Test

Radiographic Characteristics

Mean DWI infarct volume was significantly larger in the HC group compared with the NHC group (162.206 + 81.20 vs 43.62 + 55.33 cc; p< 0.0014). Patients with infarct volumes greater than 145 cc were more likely to receive HC (62.5% vs 37.5%; p<0.0001).

No significant differences in IP were observed between the two groups (7.26 + 2.95 mL .min ⁻¹. [100 g] ⁻¹) vs 8.44+ 7.42 mL .min ⁻¹ .[100 g] ⁻¹), p=0.2950). Similarly, no significant differences in mean CBV were observed between HC and NHC groups (13.72+ 12.89 mL. [100 g] ⁻¹) vs. 16.55+ 12.15 mL. [100 g] ⁻¹, p= 0.4487) (Table 3). The mean IP_area was significantly different between the two groups (61.97+ 27.10 vs 41.485 + 32.270; p<0.018). We used a logistic regression model to determine whether IP_area is associated with the probability of having a HC, controlling for potential confounders. We found that the odds of receiving a HC was higher for patients with larger IP_area (OR: 1.018(1.001, 1.036); p=0.0432). The odds ratios and 95% CI for HC based on a 5, 10, 15 or 20 cm² increase in IP_area were 1.179(1.035, 1.343), 1.390(1.071, 1.804), 1.638(1.108, 2.423) and 1.932 (1.146, 3.255), respectively). Thus, a patient with IP_area of 50 would have a 39% higher odds of having HC compared to a patient with IP_area 40 (Fig 2). The positive predictive value of IP_area for HC was 10.4% and the negative predictive value 100%. There was a significant association between receiving HC and DWI volume (OR: 1.019(1.007, 1.031); p=0.0016) but not with CBV lesion volume (OR: 0.976(0.916, 1.040); p=0.4513). However, when controlling for the DWI volume, there was no association between receiving HC and IP_area (OR: 1.008(0.975, 1.041); p=0.6391). There was no correlation between DWI volume and IP_area (p=0.5219). We also did not find a correlation between IP_area and admission NIHSS (p=0.0817), HbA1c levels (p=0.9072) or admission glucose levels. (p=0.6542).

TABLE 3.

Radiographic measurements

Radiographic measurements	Hemicraniectomy (n=12)	Non- hemicraniectomy (n= 110)	P value
DWI infarct volumes ( mean ± SD) cm3	162.20 ± 81.20 (n=7)	43.62 ± 55.33 (n=85)	0.0014§
DWI infarct volumes ≥ 145 cm3 (%)	5 (83.33%)	1(16.67%)	0.0001†
Infarct permeability	7.26mL.min−1.[100 g]−1 ±2.95	8.44mL.min−1.[100 g]−1±7.42	0.29*
Infarct permeability area	61.97 cm2±27.10	41.48 cm2±32.27	0.018§
Contralateral permeability	2.80mL.min1.[100g]−1±1.88	2.71mL.min−1.[100 g]−1±3.11	0.92*
Infarct cerebral blood volume	13.72 mL. [100 g ]−1±12.89	16.55 mL. [100 g ]−1±12.15	0.44*
Infarct cerebral blood volume area	48.36 cm2±24.23	29.14 cm2±27.33	0.0211*
Contralateral cerebral blood volume	29.80 mL. [100 g ]−1±10.04	28.86 mL. [100 g ]−1±10.72	0.77*

^†CS= chi-square test,

^§TST= Two –sample test

^*TT=Student’s t test,

Figure 2.

Graphic 1: Change in Infarct Permeability Area vs Odds Ratio

DISCUSSION

Progressive deterioration of clinical status due to massive hemispheric edema occurs in 10% of patients and is described as the “MMCA.” Ischemic edema occurs within hours after stroke onset ¹⁸ and is initially cytotoxic characterized by intracellular water accumulation, and later vasogenic, in which water moves across the blood-brain barrier (BBB) into the extracellular interstitial space. The disruption of the BBB can be demonstrated as early as 20 minutes after brain ischemia in rats.^{18, 19}

Over the past 15 years, several studies have shown that decompressive surgery is a possible treatment strategy for increased ICP after severe hemispheric stroke and can reduce the mortality to less than 50%.^3,20-23 The clinical course of many patients with severe MCA stroke is predictable. Therefore, waiting for pupillary dilatation causes an unnecessary delay.^24,25 However, there is little information to guide clinicians in selecting which patients are most suitable for this aggressive intervention.

The aim of this study was to define radiographic predictors of malignant MCA infarct in retrospectively evaluated MCA stroke patients and to provide early prognostic factors that may affect treatment decisions of clinicians. Patients that underwent HC were significantly younger than the NHC group. Wijdicks and colleagues²⁶ suggested that ICP in younger patients might rise rather quickly due to less reserve to compensate for the increase in volume. Certain degrees of age-related cerebral atrophy might protect older patients from space-occupying brain swelling.^2,7,8,27 Age is an important factor for deciding to pursue surgery and clinicians focus on younger patients before signs of herniation appear, such as stroke severity. Krieger et al reported that patients with malignant MCA strokes had a higher NIHSS score at admission and no single item or cluster of items in the NIHSS was a predictor of fatal brain edema than the total NIHSS score.⁸ Higher NIHSS scores at admission were found in our HC group as well.

DWI is increasingly being used for the early management of acute stroke.²⁸ Oppenheim et al ²⁸ had used DWI to study patients with impending malignant MCA infarction and found a mean DWI volume of 244 cm³ in patients with malignant MCA infarction and a cutoff value of 145 cm³ to predict massive brain edema within the first 14 hours after stroke.²⁸ We identified a mean volume of infarct of 162.20 cm³ and a cut off value of 100 cm³ in the HC population. Quantitative measurement of DWI volumes is a reliable factor to decide for HC.

Other radiographic factors may play a role in determining the population at risk for subsequent fatal ischemic brain swelling.⁸ PCT is an imaging technique currently used to evaluate acute stroke patients at the time of their emergency evaluation.⁹ Dittrich et al have even reported that the blood flow and volume parameters of PCT might permit identifying patients at risk for malignant infarction.²⁹ More recently, PCT has been used to characterize the BBBP.^{15, 30, 31} A relatively simple and frequently applied model to calculate BBBP is the Patlak model.^32,33 Applying this model to PCT data ^15,34-36 requires using arterial and parenchymal contrast enhancement curves to calculate the rate of contrast transfer from the intravascular to the extravascular compartment, which is a measure of BBBP. Microvascular permeability (expressed as the transendothelial transfer constant [kPS] or permeability surface area product [PS]) is a metric of BBB integrity. Similar to the standard perfusion metrics, PS can also be calculated using dynamic imaging by measuring the leakage of an intravascular tracer into the extravascular (interstitial) space.^{33, 37, 38}

In the normal brain parenchyma, PS is 0 for relatively large hydrophilic molecules (such as a peripherally injected iodinated contrast agent), which reflects the tight regulation of the BBB.^39-42 Evidence from animal and human studies suggests that increased permeability can occur in the first 2–4 hours of acute ischemia.^43-45 In our study, the infarct permeability area and infarct cerebral blood volume area of the ischemic region determined by PCT were shown to be significantly larger in patients in the HC group. This result can be explained by the fact that the patients with larger infarct permeability developed malignant brain edema. Recently PCT data were also used to characterize the BBBP, in an effort to predict which stroke patients develop hemorrhagic transformation. ¹⁵ Lin et al showed that elevated PS had been detectable during the hyperacute period using first pass dynamic CT data and was predictive for hemorrhagic transformation in patients who received t-PA.¹⁵ Aviv et al evaluated the admission PS measurements of acute stroke patients and reported that a PS threshold of 0.23mL-(100 g) ⁻1 had high sensitivity and specificity for predicting hemorrhagic transformation⁴⁶ Taken together, these studies along with our results underscore the fact that a compromise in the BBB is a necessary but not sufficient criterion for hemorrhagic transformation. If the permeability of the BBB is not large enough for blood cellular elements (RBC, platelets) to pass, hemorrhage will not occur, but the BBB only needs to be permeable to much smaller molecules such as albumin in order for water to follow it into the interstitial space and cause edema. BBB permeability maps may therefore have even more utility than just predicting hemorrhage by showing which patients may develop malignant edema.

This study is limited by its retrospective design and small sample number in the HC group. There could also have been variations in the gravity of the MCA infarcts (e.g. ICA infarct with involvement of the ACA) which may have favored HC and thus may have introduced a systematic difference between the HC and NHC groups. In addition, we found that in multivariate regression, IParea was no longer predictive of HC when controlling for DWI lesion volume. While DWI lesion size may be more predictive for HC, MRI is not always available or feasible for all patients in the initial management period, particularly for patients with large strokes who are susceptible to malignant edema (as shown by our data where patients with deterioration were less likely to have DWI follow-up). Thus, the BBB permeability map might be useful in predicting patients who will develop MMCA especially in light of the result that the CBV (core) lesion area was not predictive of deterioration. Thus, the modified Patlak permeability map uniquely offers novel information over the conventional perfusion CT maps and PCT can be obtained rapidly when the patient first presents to the emergency department.

In conclusion, the IP_area on admission PCT might be a useful tool to predict MMCA and need for HC. Given that studies suggest early rather than delayed HC for MMCA is associated with better outcomes,^20,46,47 further prospective studies are needed to validate our findings.