Article Text

Original research
CT perfusion and angiographic assessment of pial collateral reperfusion in acute ischemic stroke: the CAPRI study
  1. Arturo Consoli1,
  2. Tommy Andersson2,3,
  3. Ake Holmberg3,
  4. Luca Verganti4,
  5. Andrea Saletti5,
  6. Stefano Vallone4,
  7. Andrea Zini6,
  8. Alfonso Cerase7,
  9. Daniele Romano7,
  10. Sandra Bracco7,
  11. Svetlana Lorenzano8,
  12. Enrico Fainardi5,
  13. Salvatore Mangiafico1
  14. on behalf of the CAPRI Collaborative Group
    1. 1Interventional Neuroradiology Unit, Careggi University Hospital, Florence, Italy
    2. 2Departments of Radiology and Neurology, AZ Groeninge, Kortrijk, Belgium
    3. 3Department of Neuroradiology, Karolinska University Hospital, and Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden
    4. 4Department of Neuroradiology, Nuovo Ospedale Civile ‘S Agostino-Estense’, Modena, Italy
    5. 5Neuroradiology Unit, Azienda Ospedaliero-Universitaria of Ferrara, Cona, Ferrara, Italy
    6. 6Stroke Unit, Department of Neuroscience, Nuovo Ospedale Civile ‘S Agostino-Estense’, Modena, Italy
    7. 7Unit of Neuroimaging and Neurointervention, Policlinico ‘Santa Maria alle Scotte,’, Siena, Italy
    8. 8Department of Neurology and Psychiatry, Sapienza University of Rome, Rome, Italy
    1. Correspondence to Dr A Consoli, Interventional Neuroradiology Unit, Careggi University Hospital, Largo Palagi 1, 4th floor CTO, Firenze 50134, Italy; onemed21{at}


    Background The purpose of this study was to evaluate the correlation between a novel angiographic score for collaterals and CT perfusion (CTP) parameters in patients undergoing endovascular treatment for acute ischemic stroke (AIS).

    Methods 103 patients (mean age 66.7±12.7; 48.5% men) with AIS in the anterior circulation territory, imaged with non-contrast CT, CT angiography, and CTP, admitted within 8 h from symptom onset and treated with any endovascular approach, were retrospectively included in the study. Clinical, neuroradiological data, and all time intervals were collected. Careggi Collateral Score (CCS) was used for angiographic assessment of collaterals and the Alberta Stroke Program Early CT Score (ASPECTS) for semiquantitative analysis of CTP maps. Two centralized core laboratories separately reviewed angiographic data, whereas CT findings were evaluated by an expert neuroradiologist. Univariate and multivariate analysis were performed considering CCS both as an ordinal and a dichotomous variable.

    Results 37/103 patients (35.9%) received intravenous tissue plasminogen activator. Median (IQR) ASPECTS was 9 (6–10) for admission CT, 9 (5–10) for cerebral blood volume (CBV) maps, 3 (2–3) for mean transit time maps, 3 (2–4), for cerebral blood flow maps, and 5 (3–7) for CTP mismatch. Univariate analysis showed a significant correlation between CCS and ASPECTS for all CTP parameters. Multivariate analysis confirmed an independent association only between CCS and CBV (p=0.020 when CCS was considered as a dichotomous variable, p=0.026 with ordinal CCS).

    Conclusions A correlation between angiographic assessment of the collateral circulation and CTP seems to be present, suggesting that CCS may provide an indirect evaluation of the infarct core volume to consider for patient selection in AIS.

    • Stroke
    • Angiography
    • CT perfusion

    Statistics from


    All of the recent randomized controlled trials1–3 on the treatment of acute ischemic stroke (AIS), although with different results, highlighted a well known issue: not all patients with AIS benefit from endovascular treatment, and this may also be due to different selection criteria. The neuroimaging approach based on CT perfusion (CTP) maps is constantly developing in AIS as well as interest in the study of collaterals, and it has been postulated that CTP findings may be related to the collateral circulation.4 ,5 Conversely, although some authors have demonstrated an impact of collaterals on clinical outcome in AIS6–13 and proposed angiographic (digital subtraction angiography (DSA) or CT angiography (CTA)) scales, no grading system for the assessment of collateral circulation can be considered definite and/or commonly accepted.14

    The aim of this study was to evaluate the correlation between the collateral circulation, assessed with a recently proposed new angiographic score, the Careggi Collateral Score (CCS),12 ,13 and CTP findings in patients undergoing endovascular treatment after AIS.


    Study subjects

    Between January 2010 and June 2012, we retrospectively evaluated 431 consecutive patients with AIS in four experienced centers. Of these, 103 patients with AIS in the anterior circulation territory due to a major occlusion imaged with non-contrast CT (NCCT), CTA, and CTP within 8 h from symptom onset, not eligible for or non-responders to intravenous tissue plasminogen activator, and treated with any endovascular approach, were included in the study. A fifth site had the role of study coordinator center. Exclusion criteria are reported in online supplementary figure S1. The study received the approval of the local institutional review board of each participating center.

    Clinical assessment

    National Institutes of Health Stroke Scale (NIHSS) score at admission, 24 h after symptom onset, and at discharge was assessed in all patients. Clinical outcome at 3 months was measured using the modified Rankin Scale (mRS) by in-person visit or telephone interview; mRS ≤2 and >2 were defined as good and poor outcomes, respectively.

    CT acquisition, processing, and analysis

    NCCT, CTA, and CTP were performed at admission, on a 64 slice CT scanner (GE Medical System, Milwaukee, Wisconsin, USA) (for details, see online supplementary material). NCCT was repeated at 24–48 h. The extension of early ischemic changes was evaluated on NCCT at onset using the Alberta Stroke Program Early CT Score (ASPECTS) methodology. ASPECTS was also used to measure the final infarct extension on follow-up NCCT at 24–48 h after symptom onset. All CTP scans were assessed using a commercially available delay sensitive deconvolution software (CT Perfusion 3; GE Healthcare, Waukesha, Wisconsin, USA), which ran on an imaging workstation (Advantage Windows; GE Healthcare). As described elsewhere,15–18 CTP ASPECTS was calculated using cerebral blood flow (CBF??), cerebral blood volume (CBV) and mean transit time (MTT) maps (for details, see online supplementary material). CTP ASPECTS mismatch was considered as CBV ASPECTS minus MTT ASPECTS.19 The type of hemorrhagic transformation was classified according to the European Cooperative Acute Stroke Study II criteria. The extent of intracranial thrombus was assessed on admission CTA by the Clot Burden Score.20

    Endovascular treatment

    Patients were treated by means of thrombectomy with stent-like retrievers and loco-regional fibrinolysis alone or in combination. All time intervals were recorded (onset to intravenous treatment, onset to groin puncture, time of procedure, and onset to reperfusion) and considered for statistical analysis. Procedure related adverse events were classified as follows: (a) subarachnoid hemorrhage secondary to arterial perforation; (b) migration of thrombus in areas not previously involved; and (c) arterial dissection.

    Angiographic assessment

    Angiographic assessment of leptomeningeal collaterals was performed with a complete study of the anterior circulation, in anteroposterior and laterolateral projections, by using prolonged injections (including late venous phases until visualization of venous sinuses) through both internal carotid arteries and, when possible, at least one of the two vertebral arteries, according to the methodology adopted in CCS development (see online supplementary figure S2).12 For details, see online supplementary material. In this study, we used CCS as both a dichotomous (0–2: poor collaterals; 3–5: good collaterals) and an ordinal variable. Last anteroposterior and laterolateral runs were used to evaluate the final reperfusion, assessed with the Thrombolysis in Cerebral Infarction grading system.21

    Core radiological laboratories

    Two different core radiological laboratories were identified to analyze separately DSA studies (angiographic core laboratory) and an expert neuroradiologist with 20 years of experience analyzed NCCT, CTA, and CTP. For details, see online supplementary material.

    Statistical analysis

    Results were expressed as percentage for categorical variables, with proportions calculated by dividing the number of events by the total number of patients, excluding missing or unknown cases, and as median (IQR) or mean (±SD) for continuous variables. As appropriate, the c2 or Fisher's exact test, and the Student's t test or the Mann–Whitney U test were used to compare categorical and continuous variables between the two groups. Spearman's correlation coefficients (Spearman's r) were derived to quantify the association between CCS as an ordinal variable and CTP parameters. Multivariate linear regression modeling was used to adjust for the effects of potential confounders and to evaluate whether CCS independently predicted CTP parameters. Variables with a univariate association with CCS at a p value of ≤0.10 were included in the relative multivariate models plus other potential predictors of perfusion deficit, independent of their univariate p value, such as age. Multivariate binary and ordinal logistic regression analyses were used to evaluate whether CCS was an independent predictor of functional outcome at 3 months. Receiver operating characteristic (ROC) curve analysis was performed to determine the area under the curve (AUC) and hence cut-offs for CCS and ASPECTS parameters for predicting poor outcome (mRS >2). The independent effect of both radiological and clinical variables on mRS was calculated using multivariate logistic regression analyses, incorporating all covariates with p≤0.10 on univariate analysis. Finally, in order to verify whether a particular combination of poor collaterals and low ASPECTS was associated with poor outcome, interaction terms between CCS and each of the ASPECTS cut-offs were created and included in a binary logistic regression model for outcome to evaluate which had the highest predictive power. A two tailed p value of <0.05 was considered significant. All statistical analyses were performed with SPSS statistical package (SPSS V.22 Inc, Chicago, Illinois, USA).


    Of 431 consecutive patients, 328 were excluded and 103 patients were included in the analysis. No significant differences in baseline characteristics were found between patients excluded and those included in the study. Demographic and baseline characteristics of the included 103 patients, overall and by CCS groups, are reported in table 1.

    Table 1

    Baseline and follow-up characteristics of all patients and association with CCS 0–2 vs CCS 3–5

    Median CCS was 3 (IQR 2–4) and good collaterals, as defined by CCS 3–5, were observed in 75/103 (72.8%). Interobserver agreement for the assessment of CCS performed by the angiographic core laboratory was kappa 0.83. Median ASPECTS was 9 for admission CT, 9 for CBV maps, 3 for MTT maps, 3 for CBF maps, and 5 for CTP mismatch. Recanalization, as indicated by Thrombolysis in Cerebral Infarction 2b-3, was achieved in 64.1% of cases. Favorable outcome (mRS 0–2) at 3 months and mortality were observed in 47.6% and 6.8% of cases, respectively. Overall, patients with good collaterals (CCS 3–5), compared to those with poor collaterals (CCS 0–2), had a lower NIHSS at baseline (p=0.002) and at discharge (p<0.001), a higher proportion of middle cerebral artery M1 proximal/distal segment occlusions, a lower proportion of internal carotid artery occlusions (p=0.001), a higher admission CBV (p=0.010), and mismatch (p=0.021) ASPECTS (table 1). CCS, used as an ordinal variable, showed a significant correlation with all CTP parameters measured with ASPECTS, indicating that good collaterals were related to the presence of higher CBV (Spearman's r=0.25; p=0.013), CBF (0.22; p=0.027), MTT (0.21; p=0.039), and CTP mismatch (0.21; p=0.032) ASPECTS and, hence, with a smaller brain tissue perfusion deficit. When we included CCS in the multivariate linear logistic regression models for each of the CT perfusion parameters (CBV, MTT, CBF, and CTP mismatch ASPECTS), CCS was a significant independent predictor only for CBV ASPECTS as both a dichotomous (model 1) and ordinal variable (model 2), indicating that good collaterals predicted a higher CBV ASPECTS (table 2).

    Table 2

    Multivariate linear regression analysis for cerebral blood volume ASPECTS with CCS as the dichotomous variable (0–2 vs 3–5) (model 1) and as the ordinal variable (model 2)

    The r2 and adjusted r2 were 0.52 and 0.50, respectively, for model 1, and 0.52 and 0.49, respectively, for model 2. More patients with good collaterals achieved a favorable outcome (mRS 0–2) compared with those with poor collaterals (60.0% vs 14.3%, p<0.001) (table 1, figure 1). In multivariate binary logistic regression analysis, CCS was an independent predictor of favorable outcome (OR 9.78, 95% CI 2.55 to 37.45; p=0.001). The shift analysis confirmed CCS as an independent predictor of functional outcome at 3 months as both a dichotomous (estimate: −1.389, 95% CI −2.291 to −9.487; p=0.003) and a categorical (estimate: −1.104, 95% CI −1.741 to −0.466; p=0.001) variable. Mortality was significantly lower for patients with CCS 3–5 (2.7% vs 17.9%; p=0.007) (table 1).

    Figure 1

    Distribution of the modified Rankin Scale (mRS) score at 3 months by the Careggi Collateral Score (CCS). Note the shift to a favorable outcome with good collaterals. The numbers at the lines and in the table indicate differences in the proportion of patients (%) who attained a particular score on the mRS between good (CCS 3–5) and poor (CCS 0–2) collaterals.

    The optimal accuracy and cut-off points for CCS and baseline ASPECTS parameters (CT, CBF, CBV, MTT, and CBV–MTT mismatch) regarding 3 month mRS were defined in the ROC curve analyses. As reported in figure 2, CCS ≤2 had the highest AUC (0.70, 95% CI 0.60 to 0.80; p=0.001) with an overall performance significantly superior to other parameters, followed by MTT ASPECTS ≤2 (0.62, 95% CI 0.51 to 0.73; p=0.037). CT ASPECTS ≤7, CBF ASPECTS ≤2, and CBV ASPECTS ≤5 had an AUC of 0.60, but they did not reach the nominal statistical significance. Univariate analyses for poor outcome performed using the cut-offs individuated by ROC curve analysis, showed that low ASPECTS parameters, suggesting a larger ischemic brain injury, and poor collaterals, were associated with mRS >2 (CCS ≤2, p<0.001; MTT ASPECT ≤2, p=0.021; CT ASPECTS ≤7, p=0.049; CBF ASPECTS ≤2, 0.077; and CBV ASPECT ≤5, p=0.065). However, after multivariate adjustment, only CCS ≤2 remained as an independent predictor of mRS score 3–5 (OR 9.58, 95% CI 2.49 to 36.88; p=0.001). Clinical independent predictor was baseline NIHSS score (OR, 1.21; 95% CI 1.10 to 1.33; p<0.001). In order to verify whether a particular combination of poor collaterals and low ASPECTS was associated with poor outcome (mRS >2), interaction terms between CCS ≤2 and each of the ASPECTS parameter cut-offs were created. All were significantly associated with poor outcome on univariate analysis, but after adjustment in a binary logistic regression analysis, only the combination of CCS ≤2 and CBF ASPECTS ≤2 was an independent predictor of poor outcome (OR 6.49, 95% CI 1.14 to 36.84; p=0.035).

    Figure 2

    Receiver operating characteristic (ROC) curves analysis of the Careggi Collateral Score (CCS) and the Alberta Stroke Program Early CT Score (ASPECTS) parameters as predictors of poor outcome (modified Rankin Scale score >2). AUC, area under the curve; CBF, cerebral blood flow; CBV, cerebral blood volume; MTT, mean transit time.


    Revascularization of the occluded vessel(s) currently represents the most important goal in patients with AIS. Nevertheless, in approximately 20–25% of patients who achieve recanalization, it is found to be futile.22 Therefore, clarifying the association between collaterals and CTP findings could improve our ability to select patients for stroke treatment, given the value of the collateral circulation6–11 and CTP ASPECTS15 ,17 ,18 in predicting outcome. In addition, CTP seems to be able to differentiate between infarct core and ischemic penumbra as, according to the classical ‘penumbral hypothesis’, areas with low CBV and low CBF or high MTT (CBF or MTT/CBV match) correspond to the core, while those with normal CBV and low CBF or high MTT (CBF or MTT/CBV mismatch) refer to the penumbra.23 ,24 Although this assumption has been challenged by the demonstration that CBF better approximates the extent of the infarct core than CBV,25 the potential of the MTT/CBV mismatch model in the selection of patients for reperfusion therapies was recently confirmed.26

    In addition, although infarct core and ischemic penumbra are commonly identified using specific CTP thresholds,23 ,24 it has been repeatedly demonstrated that the application of ASPECTS to CTP maps can be useful to accurately describe the extent of irreversible and reversible ischemic damage, which are indicated by CBV and CBF or MTT ASPECTS, respectively.15 ,16 ,18 ,19 Of relevance, a connection between MTT and CBV CTP maps and collaterals has also been found.4 ,5 On the other hand, most of the existing angiographic classifications have mainly focused on the quantitative extension of the collateral circulation,21 ,27 ,28 without taking into account its qualitative features. Only the ASITN classification21 differentiated between ‘slow’ and ‘rapid’ collateral flow according to the time of collateral filling.

    For these reasons, we sought to investigate the relationships between CTP data and collaterals using CTP ASPECTS methodology and a novel angiographic grading scale (CCS) that considers an immediate and easily applicable semiquantitative and, overall, qualitative assessment of the collateral circulation. The correlation between collaterals and CTP results should be based on the concept of pial reperfusion: the more extended and efficient the retrograde leptomeningeal reperfusion, the higher the number of areas that may be maintained in a condition of potentially reversible critical hypoperfusion. In our study, univariate analysis showed a significant correlation between CCS, considered as an ordinal variable, and all perfusion parameters (CBV, MTT, CBF ASPECTS as well as CTP ASPECTS mismatch). As a higher ASPECTS corresponds to smaller ischemic brain damage and, conversely, a lower ASPECTS is related to a larger brain tissue perfusion defect, these findings suggest that good collateral flow is associated with better perfusion at the level of ischemic tissue and vice versa. However, multivariate analysis demonstrated a significant association only between pure CCS and CBV ASPECTS, while no correlation was observed with the other perfusion parameters, indicating that CCS is an independent predictor of CBV ASPECTS.

    Our findings are partially concordant with recent publications reporting a strong correlation between high relative CBV and good collaterals assessed by CTP source images (CTP-SI), confirming that CBV is a predictor of excellent collateral status.4 However, to our knowledge, no studies focusing on the correlation between angiographic assessment of collaterals and CTP parameters have been reported in the literature.

    Based on these observations, we are tempted to speculate that areas judged to have absent/incomplete and complete filling by CCS may correspond to regions of reduced and preserved CBV visualized on CTP, respectively. The extent of poor or no opacification identified by CCS and the volume of CBV deficit may reflect the infarct core size. Conversely, the low correlation between CCS and CTP ASPECTS mismatch suggests that the collateral grading system is not able to describe the amount of penumbral tissue. Collectively taken, our findings seem to suggest that the combination of CBV ASPECTS and CCS could be useful in the selection of AIS patients for endovascular therapy. In fact, a low CBV ASPECTS with poor collaterals (CCS 0–2) may identify ‘critical’ patients who do not benefit from revascularization and in whom recanalization could prove to be futile. The value of the association between CTP ASPECTS and CCS as a promising tool in the selection of AIS patients for endovascular treatment was further stressed by the demonstration that the only interaction which independently predicted poor outcome was between CBF ASPECTS ≤2 and CCS ≤2.

    This study was affected by several limitations. First, the retrospective nature, the relatively small sample size, and the preliminary selection of patients could weaken the consistency of our data. Second, our CTP studies were performed with a delay sensitive deconvolution algorithm and one phase acquisition protocol, which lead to incorrect estimation of perfusion parameters due to the effects of bolus delay and truncation of time density curves, respectively.29 Third, the actual correlation between CCS and CTP parameters remains to be elucidated because we did not use CBF to define infarct core and Tmax to delineate total hypoperfusion, as recently recommended.30 Fourth, the specific limitations of the CCS should also be considered, such as the need for a contralateral injection in the case of internal carotid artery occlusion which implies an additional delay during procedural DSA performance, lack of assessment of a potential supply from the posterior circulation and, finally, lack of a comparative assessment with other angiographic scales.12 ,13


    In this study, we demonstrated a correlation between the grade of collateral circulation evaluated by a novel angiographic scale (CCS) and CTP CBV ASPECTS, indicating that CCS may be considered as a surrogate indicator of infarct core size. As CCS provides a fast and easily applicable assessment of collaterals as well as being an independent predictor of clinical outcome, the combination of CCS and CBV ASPECTS may be useful for the selection of AIS patients as candidates for endovascular reperfusion therapy, particularly for those referred from peripheral hospitals (stroke centers) to hub centers. Further prospective and multicenter studies in a larger patient population are warranted to verify this hypothesis.


    The authors thank Dr Elizabeth Jenkins for helpful corrections to the manuscript.


    View Abstract
    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


    • Collaborators Andrea Rosi, Leonardo Renieri, Sergio Nappini, Nicola Limbucci, Guido Bigliardi, Maria Luisa Dell’Acqua, Onofrio Marcello, Luca Borgatti, Massimo Borrelli, Stefano Ceruti, Andrea Bernardoni, Marina Padroni, Carmine Tamborino, Alessandro De Vito, Cristiano Azzini, and Leonardo Capaccioli.

    • Contributors ACo and SM were responsible for the concepts of the study, coordinated the study group, and drafted and reviewed the different versions of the paper. EF, TA, and SM were involved in the core laboratories for CT and angiographic images. SL performed the statistical analysis. EF, TA, AH, AS, LV, SV, ACe, SB, DR, AZ, and all the collaborators of the CAPRI Collaborative Group, collected the cases and reviewed all versions of the drafted papers.

    • Competing interests None declared.

    • Ethics approval The study was approved by the local institutional review board of each participating center.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement All supplementary data, including specifications about the study protocol and further analyses, have been submitted and are available as supplementary material.

    Request permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.