Capillary Index Score and Correlation with Outcomes in Acute

Ischemic Stroke: a Meta-analysis

Manoj Jagani, Waleed Brinjikji, Mohammad H. Murad, Alejandro A. Rabinstein, Harry J. Cloft, and
David F. Kallmes

Abstract
Background and PurposeThe capillary index score (CIS) has been recently introduced as a metric
for rating the collateral circulation of ischemic stroke patients. Multiple studies in the last five years have
evaluated the correlation of good CIS with clinical outcomes and suggested the use of CIS in selecting
patients for endovascular treatment. We performed a meta-analysis of these studies comparing CIS with
Journal of Vascular and Interventional Neurology, Vol. 9

clinical outcomes.
MethodsWe conducted a computerized search of three databases from January 2011 to November 2015
for studies related to CIS and outcomes. A CIS = 0 or 1 is considered poor (pCIS) and a CIS = 2 or 3 is
considered favorable (fCIS). Using random-effect meta-analysis, we evaluated the relationship of CIS to
neurological outcome (modified Rankin scale score 2), recanalization, and post-treatment hemorrhage.
Meta-regression analysis of good neurological outcome was performed for adjusting baseline National
Institutes of Health Stroke Scale (NIHSS) between groups.
ResultsSix studies totaling 338 patients (212 with fCISs and 126 with pCISs) were included in the anal-
ysis. Patients with fCIS had higher likelihood of good neurological outcome [relative risk (RR) = 3.03; con-
fidence interval (CI) = 95%, 2.054.47; p < 0.001] and lower risk of post-treatment hemorrhage (RR =
0.38; CI = 95%, 0.190.93; p = 0.04) as compared with patients in the pCIS group. When adjusting for
baseline NIHSS, patients with fCIS had higher RR of good neurological outcome when compared with
those with pCIS (RR = 2.94; CI = 95%, 1.237, p < 0.0001). Favorable CIS was not associated with higher
rates of recanalization.
ConclusionsObservational evidence suggests that acute ischemic stroke patients with fCIS may have
higher rates of good neurological outcomes compared with patients with pCIS, independent of baseline
NIHSS. CIS may be used as another tool to select patients for endovascular treatment of acute ischemic
stroke.

Introduction
The degree of collateral circulation has been identified who will achieve a good clinical outcome following
as a major factor in determining the outcomes of acute endovascular treatment [8]. Several studies have been
ischemic stroke secondary to large vessel occlusion [1]. published in the last five years for evaluating the useful-
Adequate collateral blood flow mitigates infarct growth ness of the CIS and its correlation with clinical out-
and post-treatment hemorrhage and increases the likeli- comes [813]. However, the relatively small individual
hood of recanalization and reperfusion [27]. Many study sizes and the still small number of studies limit the
studies have shown that the presence of good collaterals ability of clinicians to determine the practicality of CIS
is associated with good clinical outcome [5,6,814]. in patient selection or as a predictor of outcomes. Due to
these pitfalls, we performed a systematic review and a
The capillary index score (CIS) has recently been pro- meta-analysis of studies that evaluated CIS in relation to
posed as a method of rating the collateral circulation of outcomes in patients undergoing intra-arterial therapy
acute ischemic stroke patients [9]. CIS was suggested to for acute ischemic stroke.
improve patient selection for endovascular treatment,
and it is thought to be a reliable predictor of patients

Vol. 9, No. 3, pp. 713. Published January, 2017.

All Rights Reserved by JVIN. Unauthorized reproduction of this article is prohibited
Corresponding Author:
 8
Journal of Vascular and Interventional Neurology, Vol. 9

Figure 1. Quantification of the Cis. in each image, the ischemic area is divided into three sectors. Sectors with no capil-
lary blush: 0. Sectors showing capillary blush: 1. Patient in (a) receives a Cis = 1. Patient in (b) receives a Cis = 2.
Patient in (c) receives a Cis = 3. Adapted from [11].

Methods artery or M1 branch of the middle cerebral artery occlu-

sions. To determine the CIS on a digitally subtracted
Literature Search angiogram, the bilateral common carotid arteries and the
To identify comparative studies on the CIS and out- dominant vertebral artery are injected and assessed
comes, an experienced librarian searched the three data- through the venous phase for capillary blush. The ische-
bases Ovid EMBASE, Ovid MEDLINE and Web of Sci- mic area is divided into three equal parts, and each part
ence for articles from January 2011 to November 2015. is given a score of 0 or 1. A score of 0 indicates
The keywords capillary index score, mechanical throm- no capillary blush, while a score of 1 indicates normal
bectomy, stroke, acute ischemic stroke, collaterals, and capillary blush. The scores for the three parts are added
outcome were used for the search. Both published up to give a CIS range of 03. A CIS = 0 or 1 is consid-
manuscript and abstract presentations were included in ered poor (pCIS), and a CIS = 2 or 3 is considered favor-
the search. After the initial search was complete, the ref- able (fCIS). The higher scores indicate that greater
erences of the identified articles were assessed to find ischemic area is potentially viable for recovery follow-
any other relevant articles. ing recanalization. An example of how CIS is calculated
is shown in Figure 1.
The initial studies were, then, further evaluated by two
authors for inclusion in the meta-analysis. The inclusion Outcomes
criteria were the following: 1) studies properly assigning Each study separated the patients into two groups of
CISs as described by Al-Ali et al. [9], 2) studies com- favorable CIS and poor CIS. Studied outcomes included
paring CISs with some relevant clinical outcomes rates of good neurological outcome, recanalization, and
between two separate groups (i.e., favorable CIS and post-recanalization hemorrhage. Good neurological out-
unfavorable CIS), 3) studies reporting separate outcome come was defined as a modified Rankin scale score
data for the two groups, and 4) studies with 20 or more (mRS) 2 at 90 days. Recanalization was defined as
patients. Favorable CIS was defined as a CIS of 2 or 3, thrombolysis in cerebral infarction 2b or thrombolysis
and unfavorable CIS was defined as a CIS of 0 or 1. in myocardial infarction 2. We also extracted data on
Studies were excluded if they: 1) did not stratify out- baseline National Institutes of Health Stroke Scale
comes by CIS or 2) had less than 20 patients. (NIHSS) scores and time to recanalization when availa-
ble for each group.
Capillary Index Score
The CIS was initially described by Al-Ali et al. as a Risk of Bias Assessment and Evaluation of
method to evaluate the adequacy of collateral flow in the Study Characteristics
ischemic area [9]. CIS assumes that capillary blush on a Risk of bias assessment was performed using the New-
cerebral angiogram indicates viable tissue. The CIS is castle-Ottawa Scale. This is a tool used for assessing the
used only in cases of intra-cerebral internal carotid quality of non-randomized studies included in system-
 Jagani et al. 9
Journal of Vascular and Interventional Neurology, Vol. 9

Figure 2. Flow chart of literature assessment.

atic reviews and/or meta-analyses. Each study is judged Results

on eight items categorized into three groups: 1) selection
of the study groups, 2) comparability of the study Literature Search
groups, and 3) ascertainment of the outcome of interest
Fifteen articles were found on the initial literature
[15]. Factors examined in our assessment of risk of bias
search. Of these articles, eight were excluded (53.3%)
were the following: 1) similar baseline NIHSS between
after reading the abstract because they did not use the
groups, 2) similar time to revascularization between
CIS or did not include any clinical outcome data. One
groups, 3) use of angiographic core labs, 4) patient-
additional article was excluded, because only percen-
reported mRS scores at 90 days, and 5) similar revascu-
tages of the patients in each group were given. In total,
larization techniques between groups.
six studies were included in the meta-analysis: four were
published in full text and two were abstract presenta-
Statistics
tions given at conferences. A flow chart that describes
We extracted a 2 2 table for each binary outcome from the study selection process is shown in Figure 2. Not all
each study. Random-effect meta-analysis was used for of the studies provided data on all outcomes. In addition,
pooling across studies [16]. The I2 statistic was used to data regarding good neurological outcome from [13]
express the proportion of inconsistency that is not attrib- were excluded because it was defined as mRS 3. The
utable to chance [17]. Meta-analysis results were largest study contained 146 patients (106 with fCISs and
expressed as relative risk (RR) and 95% of confidence 40 pCISs), and the smallest study included 26 patients.
interval (CI). We planned to explore the impact of publi- The studies and the available outcomes from each are
cation bias by constructing funnel plots and testing their summarized in Table 1. In total, 338 patients were inclu-
symmetry if a sufficient number of studies (>20) was ded in this study, 212 in the fCIS group and 126 in the
available. A meta-regression analysis for good neurolog- pCIS group.
ical outcome was performed for adjusting baseline
NIHSS between groups in studies which included such Baseline NIHSS and Time to Recanalization
data. Meta-analysis was conducted using Stata (Stata-
Corp. 2015. Stata Statistical Software: Release 14. Col- The baseline NIHSS scores were significantly different
lege Station, TX: StataCorp LP). between the two groups (p = 0.003). The mean baseline
NIHSS score in the fCIS group was 17.5 (16.118.9)
and 19.5 (18.220.8) in the pCIS group (p = 0.003). The
time to recanalization was not statistically different
between the two groups with the mean in fCIS group
 10

Table 1. Studies Included in Meta-analysis

Authors, Year fCIS (n) pCIS (n) Good Neu- Recan- Post-Treat- Baseline Baseline Time to Time to Risk of Bias
rologi- alization ment Hem- NIHSS NIHSS Recanaliza- Recanaliza-
cal Outcome orrhage in fCIS in pCIS tion in fCIS tion in
group group group (min) pCIS group
(min)
Jagani et al., 2015 13* 14* 18 6 16 6 336 176 355 121 H
[11]
Al-Ali et al., 2013 [9] 11 15 18 9 22 7 380* 399* M
Al-Ali et al., 2014 13 15 19 4 20 3 L
[10]
Al-Ali et al., 2015 [8] 48 30 18 5 20 4 237 49 252 57 L
Labeyrie et al., 2015 106 40 16 5 19 6 340 79 304 93 M
[12]
Teleb et al., 2014 21 12 H
[13]
*
No standard deviation available.
Journal of Vascular and Interventional Neurology, Vol. 9

Figure 3. Forest plot for good neurological outcome.

being 332 min (264.8397.2) and the mean in the pCIS those with pCIS (RR = 2.94; CI = 95%, 1.237, p <
group being 326.5 min (246.5406.6). 0.0001).
There was lower risk of post-treatment intra-cerebral
Outcomes
hemorrhage within the fCIS group (RR = 0.38; CI =
Favorable CISs were associated with higher RR of good 95%, 0.190.93; p = 0.04). Favorable CIS was not asso-
neurological outcome (RR = 3.03; CI = 95%, 2.054.47; ciated with higher rates of recanalization (RR = 1.18; CI
p < 0.001) compared with pCIS [Figure 3 (forest plot)]. = 95%, 0.991.41; p = 0.06). These data are summarized
in Table 2. Forest plots for hemorrhage and recanaliza-
Fifty-nine percent (112 of 191) of patients with fCIS had
tion are provided in Figures 4 and 5.
good neurological outcome compared with 18% (21 of
114) of patients with pCIS. On meta-regression adjust-
Study Heterogeneity
ing for baseline NIHSS, patients with fCIS had higher
The I2 values were <50% for the three main outcomes:
RR of good neurological outcome when compared with
1) good neurological outcome (I2 = 38%), 2) post-treat-
 Jagani et al. 11

Table 2. Meta-analysis Results: Outcomes

Outcome RR (fCIS versus pCIS) 95% CI P Value I2
Good Neurological Outcome (mRS 2) 3.03 2.054.47 <0.001 38%
Post-Treatment Hemorrhage 0.38 0.160.93 0.04 0%
Recanalization 1.18 0.991.41 0.06 0%
Journal of Vascular and Interventional Neurology, Vol. 9

Figure 4. Forest plot for post-treatment hemorrhage.

Figure 5. Forest plot for recanalization.

 12

ment hemorrhage (I2 =0%), and (3) recanalization (I2 = 0.001) [14]. Of the 19 studies, the majority graded col-
0%). These values indicate relatively low heterogeneity laterals using DSA imaging, primarily on the American
among the studies. Society of Interventional and Therapeutic Neuroradiol-
ogy and the Society of Interventional Radiology scale
Dicussion [20]. One potential advantage of the CIS scoring method
over traditional DSA-based collateral scoring methods is
We conducted a systematic review and meta-analysis of the fact that the scale is easily quantifiable with high
five studies reporting on 338 patients and found that rates of inter-observer agreement [11]. Furthermore, the
patients with favorable CISs had higher rates of good scale evaluates the functionality of collaterals by assess-
neurological outcome (mRS 2) following endovascu- ing capillary perfusion rather than the presence of collat-
lar therapy for acute ischemic stroke when compared erals alone.
with those with a poor CIS, even when adjusting for
baseline NIHSS. Patients with fCIS had significantly In addition, several studies have used collateral grading
lower rates of post-treatment hemorrhage and trended methods based on CT angiography and CT perfusion
toward higher recanalization rates as well. These find- techniques [2124]. One advantage of CT-based collat-
Journal of Vascular and Interventional Neurology, Vol. 9

ings are important as they suggest that CIS is a poten- eral scores is the fact that they rely on non-invasive
tially useful tool in predicting outcomes following endo- imaging and can, thus, be used for patient selection prior
vascular treatment of acute ischemic stroke. to conventional angiography. These scores may be more
advantageous because they truly evaluate the enhance-
Proposed applications of the CIS are in patient selection
ment and extent of collateral vessels in the ischemic ter-
and prognostication. Many of the individual studies and
ritory. To date, no studies have been performed compar-
this meta-analysis found fCIS and good neurological
ing the accuracy of CTA-based scoring systems and the
outcome to be strongly correlated. The relationship of
CIS in predicting good neurological outcome following
collateral grade to clinical outcome supports the hypoth-
endovascular therapy.
esis that fCIS can identify viable ischemic tissue that
may be amenable to revascularization; however, there is
Limitations
no guarantee that the tissue will recover with or without
endovascular intervention. In addition, identifying This study has several limitations. First, there were a rel-
patients with pCIS can be valuable. Reperfusing patients atively small number of studies included in analysis.
with pCIS can be harmful because reperfusing nonviable This may be due to the relatively short time since the
tissue may lead to hemorrhagic transformation and introduction of the CIS. In addition, not all studies had
edema, as demonstrated in our study. Our findings reaf- each of the investigated outcomes, further adding to the
firm the importance of collaterals and CIS in predicting small size and limiting the power of the meta-analysis.
clinical outcomes, independent of baseline NIHSS. It is Each of the studies included was retrospective and non-
important to point out, however, that fCIS cannot be the randomized, further limiting certainty in the observed
sole determinant in patient selection for endovascular estimates. Baseline NIHSS scores were lower in the
therapy. The degree of collateral flow can help predict fCIS group and likely played a role in the association of
the likelihood of good outcomes, but the absence of the score results and outcome. It is important to point
good flow should not prevent endovascular therapy[6]. out, however, that in meta-regression of good neurologi-
In patients with pCIS, 18% still recovered with good cal outcomes adjusting for baseline NIHSS, fCIS was
neurological outcome. still significantly association with good neurological
outcome. Differences in grading the CIS, quality of
A number of studies have been published examining angiogram scored, and rater experience among the inclu-
efficacy of other collateral grading systems in predicting ded studies could not be evaluated. Publication bias was
outcomes of endovascular therapy for treatment of acute not assessed because our meta-analysis incorporated less
ischemic stroke. In general, poor collaterals are associ- than 10 studies. Overall, the quality of the evidence (i.e.,
ated with lower rates of good neurological outcomes and confidence in the estimates of effect) is low.
higher rates of hemorrhage, similar to those seen with
the CIS [3,18,19]. In a large meta-analysis of 19 studies Conclusion
over the last 15 years that compared collateral grades
with outcomes, Leng et al. found that good pre-treat- Observational evidence suggests that acute ischemic
ment collateral status almost doubled the rate of good stroke patients with favorable CISs had higher rates of
neurological outcome as compared with patients with good neurological outcomes, independent of baseline
poor collaterals (RR = 1.98; CI = 95% ,1.642.38; p < NIHSS. These findings suggest that CIS can be used as
 Jagani et al. 13

a tool to prognosticate outcomes of endovascular treat- 12. Labeyrie PE, et al. The capillary index and aspect scores are strong
ment of acute ischemic stroke. It is important to point predictors of patient outcome in the interventional management of
ischemic strokes. Stroke 2015;46
out that nearly 20% of patients with a pCIS had good
13. Teleb M, et al. Validation of a new modified capillary index score
neurological outcome following endovascular treatment,
angiographic real time assessment of dead vs salvable tissue. J Neu-
thus suggesting that CIS should not be used as the only roInterv Surg 2014;6:A71.
tool for patient selection at this time. 14. Leng X, et al. Impact of collaterals on the efficacy and safety of
endovascular treatment in acute ischaemic stroke: a systematic
References review and meta-analysis. J Neurol Neurosurg Psychiatry
2016;87(5):537544.
1. Liebeskind DS. Collateral circulation. Stroke 2003;34(9):227984. 15. Deeks JJ, et al. Evaluating non-randomised intervention studies.
2. Bang OY, et al. Impact of collateral flow on tissue fate in acute Health Technol Assess 2003;7(27):iiix. 1173.
ischaemic stroke. J Neurol Neurosurg Psychiatry 2008;79(6):625 16. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control
629. Clin Trials 1986;7(3):177188.
3. Bang OY, et al. Collateral flow predicts response to endovascular 17. Higgins JP, et al. Measuring inconsistency in meta-analyses. BMJ
therapy for acute ischemic stroke. Stroke 2011;42(3):693699. 2003;327(7414):557560.
4. Bang OY, et al. Collateral flow averts hemorrhagic transformation
Journal of Vascular and Interventional Neurology, Vol. 9

18. Elijovich L, et al. CTA collateral score predicts infarct volume and
after endovascular therapy for acute ischemic stroke. Stroke
clinical outcome after endovascular therapy for acute ischemic
2011;42(8):22352239.
stroke: a retrospective chart review. J Neurointerv Surg 2015
5. Christoforidis GA, et al. Angiographic assessment of pial collaterals
as a prognostic indicator following intra-arterial thrombolysis for 19. Nambiar V, et al. CTA collateral status and response to recanaliza-
acute ischemic stroke. AJNR Am J Neuroradiol 2005;26(7):1789 tion in patients with acute ischemic stroke. AJNR Am J Neuroradiol
1797. 2014;35(5):884890.
6. Liebeskind DS, et al. Collaterals at angiography and outcomes in the 20. Higashida RT, et al. Trial design and reporting standards for intra-
Interventional Management of Stroke (IMS) III trial. Stroke arterial cerebral thrombolysis for acute ischemic stroke. Stroke
2014;45(3):759764. 2003;34(8):e109e137.
7. Marks MP, et al. Effect of collateral blood flow on patients undergo- 21. Menon BK, et al. Multiphase CT angiography: a new tool for the
ing endovascular therapy for acute ischemic stroke. Stroke imaging triage of patients with acute ischemic stroke. Radiology
2014;45(4):10351039. 2015;275(2):510520.
8. Al-Ali F, et al. Relative influence of capillary index score, revascu- 22. Seeta Ramaiah S, et al. The impact of arterial collateralization on
larization, and time on stroke outcomes from the interventional outcome after intra-arterial therapy for acute ischemic stroke. AJNR
management of stroke III trial. Stroke 2015;46(6):15901594. Am J Neuroradiol 2014;35(4):667672.
9. Al-Ali F, et al. The capillary index score: rethinking the acute ische- 23. Sung SM, et al. Functional outcome after recanalization for acute
mic stroke treatment algorithm. Results from the Borgess Medical pure M1 occlusion of the middle cerebral artery as assessed by col-
Center Acute Ischemic Stroke Registry. J Neurointerv Surg lateral CTA flow. Clin Neurol Neurosurg 2015;131:7276.
2013;5(2):139143.
24. van den Wijngaard IR, et al. Impact of collateral status evaluated
10. Al-Ali F, et al. Capillary index score in the interventional manage- by dynamic computed tomographic angiography on clinical out-
ment of stroke trials I and II. Stroke 2014;45(7):19992003. come in patients with ischemic stroke. Stroke 2015;46(12):3398
11. Jagani M, et al. Capillary index score in acute ischemic stroke: 3404.
inter-observer reliability and correlation with neurological out-
comes. J Neurosurg Sci 2015