This study was a retrospective analysis of data prospectively collected from acute ischemic stroke patients who presented at our institution within 6 hours of symptom onset from April 2015 to January 2016. Patients were included if they met the following inclusion criteria: 1) baseline imaging data, including non-contrast computed tomography and mpCTA prior to endovascular treatment, were available, 2) there was occlusion of the intracranial segment of the internal carotid artery (ICA) or middle cerebral artery (MCA) (M1 or M2 segment), and 3) they had a National Institutes of Health Stroke Scale (NIHSS) score of 5 or higher. We excluded patients with poor-quality mpCTA data that prevented the assessment of pial collateral circulation and those who were lost to follow-up.

All patients underwent a standard scanning protocol at presentation using a 256-slice multidetector CT scanner (Brilliance iCT 256; Philips Medical Systems, Cleveland, OH, USA). The mpCTA scanning parameters used were 80 kVp, 175 mAs, and 128×0.625 mm detector collimation. A total of 90 mL of contrast was injected for the scan at a rate of 4–5 mL/sec, followed by a 30-mL normal saline chase at a rate of 3 mL/sec.

The first phase of imaging involved an aortic arch-to-vertex CTA scan that was initiated by the bolus tracking method with a trigger level of 150 Hounsfield units at the aortic arch, minimum scan delay, and an average dose length product of 1,100 to 1,200 mGy∙cm. This was acquired in less than 3 seconds. The second phase started after a delay of 4 seconds and was acquired in 3 seconds from the vertex to skull base. The third phase was initiated after another 4-second delay with scanning in the opposite direction from the skull base to vertex. The scan was performed in this direction for the third phase because our scanner is not able to be repositioned within the time limit of 7 seconds between phases. Images from the first phase (i.e., the aortic arch-tovertex scan) were then reconstructed as 1.5-mm-thick axial sections. In addition, 24-mm-thick slab multi-planar reformations of axial first to third phase images from the skull base to vertex were performed to assess the degree of pial collateral filling.

This mpCTA protocol does not require additional contrast medium, and there is minimal extra radiation (approximately 177 mGy∙cm per phase) as the second- and third-phase images are acquired from the skull base to vertex only [11,12].

Radiologic data were independently assessed by a neurologist (H.G.W.) and 2 neuroradiologists (C.J. and L.S.) who were blinded to patient clinical symptoms. Pial arterial filling score was measured in accordance with a previously published report [9]. Although there was excellent interrater reliability between the two radiologists (n=30, κ=0.81, P<0.001) in a previous report, interrater reliability in our study was moderate on first interpretation (n=38, κ=0.46, P<0.001). Hence, we needed to clarify the three determinants of pial arterial filling score, i.e., the degree of phase delay, prominence, and extent of pial collaterals. The degree of phase delay was defined as the difference in the phase of pial collateral filling between thick axial maximum intensity projections of the head in the index territory and that in the contralateral territory [12]. Reductions in the prominence and extent of pial collaterals were defined as when the number of pial vessels in the index territory was smaller than that in the contralateral territory, with adjustment for different phases of collateral filling between the index and contralateral territories. Ischemic regions with no vessels were defined as regions where one third or more of the lesions in the index territory had no collateral filling in any phase (Table 1).

Table 1 and Fig. 1 show an example template for each score (i.e., 0–6) based on the three determinants above. Our three raters then revised their pial collateral assessments according to this template, with a consensus required in cases of disagreement.

A neurointerventionalist (C.J.) performed intraarterial treatment (IAT) of all the cases on a biplane angiography machine (Integris Allura, Philips, the Netherlands). We used a 9 French balloon guiding catheter. In case of a tortuous cervical ICA, a second guiding catheter was used for the coaxial system. The primary device for IAT was a stent retriever (Solitaire FR; Covidien, Irvine, California; Trevo-XP; Stryker, Kalamazoo, MI, USA). Mechanical aspiration device, thrombolytics, or permanent stenting were also used at the discretion of operator. The reperfusion grading was measured by using modified treatment in cerebral infarction before and after IAT [13].

Patient demographics, as well as clinical and radiological findings, were compared between patients with favorable (modified Rankin Scale [mRS] at 3 months ≤2) and unfavorable (mRS at 3 months >2) outcomes using the Pearson chi-square test, independent t test, and Mann-Whitney U test as appropriate. Univariate and multivariate logistic regression analyses were then performed to identify factors associated with a favorable functional outcome. Variables with a P-value of ≤0.1 in univariate analysis were included in multivariate analyses. We then determined the area under the receiver-operating characteristic curve (AUC) to assess model fit and compared AUC values between pial arterial filling scores dichotomized at various cut-off scores (i.e., >4 vs. ≤4, >3 vs. ≤3, and >2 vs. ≤2). Areas under the receiver operating characteristic curves for these 2 scoring systems were compared using DeLong’s test. All statistical analyses were performed using SPSS Statistics 22 (IBM, Armonk, NY, USA) and Stata 13.0 software (Stata Corporation, College Station, TX, USA). P values <0.05 were considered statistically significant. Results are expressed as mean±standard deviation where appropriate.

Of the 838 patients in our database with acute ischemic stroke, 257 patients underwent mpCTA within 6 hours of symptom onset. Among these, 208 patients who were not candidates for endovascular revascularization were excluded due to mild symptoms (NIHSS <4, 152 patients), no occlusion identified on mpCTA (46 patients), and hemorrhagic diathesis (10 patients). In total, 11 patients underwent endovascular treatment for occlusion in the basilar artery or vertebra artery, and 38 underwent endovascular treatment using a stent retriever for ICA/MCA occlusion identified on mpCTA. The mean age of enrolled patients was 72.4±10.2 years, and 55.3% were male. Favorable outcomes were observed in 20 patients (52.6%), whereas unfavorable outcomes were observed in 18 (47.4%). The most common stroke subtype was cardioembolism (63.2%). Thirty-one patients (81.6%) were treated with intravenous tissue plasminogen activator before endovascular treatment, and 34 (89.5%) had successful reperfusion. The mean pial arterial filling score was 3.0±1.2.

Interrater reliability of the pial arterial filling score between a neurologist and two neuroradiologists, as assessed using unweighted k statistics with 95% confidence intervals (CIs), indicated substantial agreement (n=38, κ=0.64, P<0.001).

Patient demographics and clinical and radiological parameters are shown in Table 2. Mean age was lower in the favorable (67.1±10.4 years) than in the unfavorable (78.3±6.0 years) outcome groups (P<0.001), while pial arterial filling score was higher in the favorable (3.6±0.8) than in the unfavorable (2.4±1.2) outcome groups (P=0.002).

In our univariate analysis, younger age (odds ratio [OR], 0.825; 95% CI, 0.720–0.945; P=0.006) and lower NIHSS score at admission (OR, 0.839; 95% CI, 0.720–0.977; P=0.024) were significantly associated with favorable outcomes at 3 months. Higher pial arterial filling score (OR, 2.917; 95% CI, 0.321–6.443; P=0.008) was also significantly associated with favorable outcomes. Because of the ambiguity of the score threshold for predicting clinical outcomes in endovascular treatment, we dichotomized the pial arterial filling score at >4 vs. ≤4, >3 vs. ≤3, and >2 vs. ≤2 and used each dichotomization as a predictor variable in our analyses. Similar trends to those above were shown when the pial arterial filling score was dichotomized at >2 vs. ≤2 (OR, 14.143; P=0.003). In contrast, pial arterial filling scores dichotomized at >3 vs. ≤3 and >4 vs. ≤4 were not significantly associated with clinical outcome. After adjusting for age and NIHSS score at admission, higher pial arterial filling scores and pial arterial filling scores dichotomized at >2 vs. ≤2 were found to be independently associated with favorable clinical outcomes (Table 3). The pial arterial filling scores dichotomized at >2 vs. ≤2 (AUC, 0.756; 95% CI, 0.621–0.89; P=0.249) also effectively predicted favorable clinical outcomes at 3 months, with higher AUC than dichotomization at pial arterial filling scores dichotomized at >3 vs. ≤3 and >4 vs. ≤4 (AUC, 0.636; 95% CI, 0.482–0.791; P=0.364, and AUC, 0.550; 95% CI, 0.483–0.617; P<0.001, respectively) (Fig. 2). The difference was statistically significant in favor of the pial arterial filling scores dichotomized at >2 vs. ≤2 compared with the pial arterial filling scores dichotomized at >4 vs. ≤4 (DeLong’s test, P-value=0.006). However, the difference was not statistically significant between the pial arterial filling scores dichotomized at >3 vs. ≤3 compared with the pial arterial filling scores dichotomized at >4 vs. ≤4 (DeLong’s test, P-value=0.274) and between the pial arterial filling scores dichotomized at >2 vs. ≤2 compared with the pial arterial filling scores dichotomized at >3 vs. ≤3 (DeLong’s test, P-value=0.073).