Received: 1 July 2021 | Accepted: 4 July 2021

DOI: 10.1111/ene.15013

ORIGINAL ARTICLE

Circle of Willis variants and their association with outcome in

patients with middle cerebral artery-­M1-­occlusion stroke

Laura P. Westphal1 | Niklas Lohaus2 | Sebastian Winklhofer2 | Christian Manzolini1 |

Ulrike Held3 | Klaus Steigmiller3 | Janne M. Hamann1 | Mohamad El Amki1 |
Tomas Dobrocky4 | Leonidas D. Panos5 | Johannes Kaesmacher4,6 | Urs Fischer5 |
Mirjam R. Heldner5 | Andreas R. Luft1 | Jan Gralla4 | Marcel Arnold5 |
Roland Wiest4 | Susanne Wegener1

Abstract
1
Department of Neurology, Clinical
Neuroscience Center, University Hospital
Zurich and University of Zurich, Zurich, Background: An incomplete circle of Willis (CoW) has been associated with a higher risk
Switzerland of stroke and might affect collateral flow in large vessel occlusion (LVO) stroke. We aimed
2
Department of Neuroradiology, Clinical
to investigate the distribution of CoW variants in a LVO stroke and transient ischemic at-
Neuroscience Center, University Hospital
Zurich and University of Zurich, Zurich, tack (TIA) cohort and analyze their impact on 3-­month functional outcome.
Switzerland
Methods: CoW anatomy was assessed with time-­of-­flight magnetic resonance angiog-
3
Department of Biostatistics,
Epidemiology, Biostatistics and Prevention
raphy (TOF-­MRA) in 193 stroke patients with acute middle cerebral artery (MCA)-­M1-­
Institute, University of Zurich, Zurich, occlusion receiving endovascular treatment (EVT) and 73 TIA patients without LVO. The
Switzerland
4
main CoW variants were categorized into four vascular models of presumed collateral
Department of Diagnostic and
Interventional Neuroradiology, University flow via the CoW.
Hospital Berne and University of Berne, Results: 82.4% (n = 159) of stroke and 72.6% (n = 53) of TIA patients had an incomplete
Berne, Switzerland
5 CoW. Most variants affected the posterior circulation (stroke: 77.2%, n = 149; TIA: 58.9%,
Department of Neurology, University
Hospital Berne and University of Berne, n = 43; p = 0.004). Initial stroke severity defined by the National Institutes of Health
Berne, Switzerland
6
Stroke Scale (NIHSS) on admission was similar for patients with and without CoW vari-
Department of Diagnostic, Interventional
and Pediatric Radiology, University ants. CoW integrity did not differ between groups with favorable (modified Rankin Scale
Hospital Berne and University of Berne, [mRS]): 0–­2) and unfavorable (mRS: 3–­6) 3-­month outcome. However, we found trends
Berne, Switzerland
towards a higher mortality in patients with any type of CoW variant (p = 0.08) and a
Correspondence higher frequency of incomplete CoW among patients dying within 3 months after stroke
Susanne Wegener, Department of
Neurology, University Hospital Zurich, onset (p = 0.119). In a logistic regression analysis adjusted for the potential confounders
Frauenklinikstrasse 26, Zurich 8091, age, sex and atrial fibrillation, neither the vascular models nor anterior or posterior vari-
Switzerland.
Email: [Link]@[Link] ants were independently associated with outcome.
Conclusion: Our data provide no evidence for an association of CoW variants with clinical
Funding information
UZH Clinical Research Priority Program outcome in LVO stroke patients receiving EVT.
(CRPP) stroke; Swiss National Science
Foundation, Grant/Award Number: SNSF KEYWORDS
PP00P3_170683 anatomical variants, circle of Willis, collaterals, stroke, TOF-­MRA

[Correction added on 15 April 2022 after first online publication: CSAL funding statement has been added after print and online publication.]

This is an open access article under the terms of the Creat​ive Commo​ns Attri​bution-NonCo​mmercial License, which permits use, distribution and reproduction
in any medium, provided the original work is properly cited and is not used for commercial purposes.
© 2021 The Authors. European Journal of Neurology published by John Wiley & Sons Ltd on behalf of European Academy of Neurology

3682 | 
[Link]/journal/ene Eur J Neurol. 2022;28:3682–3691.
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
CIRCLE OF WILLIS VARIANTS IN LVO STROKE 3683

I NTRO D U C TI O N Berne with approval of the local ethics committee of Berne (KEK:
231/14) and Zurich (KEK: 2014–­0304). All stroke patients under-
The circle of Willis (CoW) is a polygonal arterial anastomotic system went magnetic resonance imaging (MRI) as primary imaging diag-
located at the skull base connecting the anterior with the posterior nostic, whereas TIA patients received either computed tomography
circulation as well as both cerebral hemispheres, thereby ensuring the (CT) or MRI scan with MRI performed within 24 h after onset. Stroke
maintenance of cerebral blood flow (CBF) [1,2]. Anatomical variants patients were included if acute ischemic stroke due to an MCA-­M1-­
of the CoW are frequent with reported occurrence greater than 50% occlusion was confirmed by magnetic resonance angiography (MRA),
in the healthy population [2,3]. Most variants affect the posterior diffusion and time-­of-­flight magnetic resonance angiography (TOF-­
circulation with an average of 40% and more [1,3,4]. In patients with MRA) images were complete with sufficient quality, EVT was at-
an intact CoW and gradual progression of occlusive atherosclerotic tempted, and mRS score at 3 months’ follow-­up was available. Stroke
cerebrovascular disease, collateral flow via the CoW is provided via patients were excluded if prior territorial infarction or additional
the communicating anterior (Acom) and posterior arteries (Pcom) [2]. intra-­ or extracranial vessel occlusions other than extension of the
However, CoW variants may affect the risk of stroke as well as the col- M1-­occlusion to the internal carotid artery (ICA) could be detected
lateral flow via the CoW during stroke. A study from van Seeters et al. on MRI/MRA. In the TIA cohort, patients were excluded if MRI or
[4] found that in healthy individuals with atherosclerotic disease but MRA was missing or if the final diagnosis was other than TIA. Patient
no prior vascular events, an incomplete anterior CoW was associated characteristics for stroke patients included demographic informa-
with an increased risk of future anterior circulation stroke with the tion (age, sex, independent prior stroke), vascular risk factors (atrial
highest risk in those with combined incomplete posterior and anterior fibrillation, diabetes mellitus, arterial hypertension, dyslipidemia,
circulation, whereas a posterior circulation variant alone was not. Due current smoking, coronary heart disease, peripheral artery disease,
to the anastomotic system of an intact CoW, an acute intracranial ves- prior stroke or TIA), previous medication (antiplatelet therapy, oral
sel occlusion might be at least partially compensated by collateral flow anticoagulation, statin or antihypertensive therapy), baseline stroke
via the CoW, but less tolerated within an incomplete CoW. However, admission information (National Institutes of Health Stroke Scale
so far, there have been no studies analyzing the impact of ipsilateral [NIHSS] on admission, blood pressure, levels of glucose, glycosylated
and contralateral CoW variants on outcome in patients with middle hemoglobin [HbA1c], C-­reactive protein [CRP] and international
cerebral artery (MCA)-­M1-­occlusion stroke. normalized ratio [INR]), collateral status, stroke therapy and etiol-
In our study, we aimed to assess the frequency and distribution ogy. The percentages of missing values were as follows: age (0%),
of CoW variants in a LVO stroke and transient ischemic attack (TIA) sex (0%), independent prior stroke (10.4%), atrial fibrillation (0.52%),
cohort and to evaluate the role of CoW variants for clinical out- diabetes mellitus (0%), arterial hypertension (0%), dyslipidemia (1%),
come in stroke. Therefore, we analyzed a large homogenous cohort current smoking (9.3%), coronary heart disease (0%), peripheral ar-
of ischemic stroke patients with MCA-­M1-­occlusion, all subjected tery disease (9.3%), prior stroke or TIA (0%), NIHSS on admission
to mechanical thrombectomy within 6 h of symptom onset, and (0%), 3-­month mRS (0%), collateral status (1.6%), stroke etiology ac-
categorized them into groups with any, anterior or posterior CoW cording to TOAST (Trial of ORG 10172 in Acute Stroke Treatment)
variants and according to vascular models of presumed collateral criteria (0%), onset to groin puncture (2.6%), Thrombolysis In
flow via the CoW (see Methods section). Demographic and clinical Cerebral Infarction (TICI) scale (0%), general anesthesia (3.1%) and
parameters such as stroke severity, etiology and 3-­month modified CoW variants and vascular models (0%).
Rankin Scale (mRS) were compared between groups. Furthermore,
we assessed the distribution of CoW variants within 73 TIA patients
without intracranial vessel occlusion to evaluate the extent of vari- Circle of Willis assessment and vascular models
ants in a cohort with a similar vascular risk profile.
The anatomy of the CoW was assessed on TOF-­MRA acquired at the
time of hospital admission for stroke patients or on a follow-­up MRI
M E TH O D S within 24 h after symptom onset for TIA patients, respectively. The MRI
analyses of stroke patients were performed on 1.5T or 3T MRI systems
Patient data from one vendor (Siemens Healthineers, Erlangen, Germany) and for
TIA patients on 3T MRI systems (Skyra 3T Siemens Healthineers). The
In this retrospective data analysis, we used encrypted clinical and MRA of stroke patients consisted of 139 slices obtained with a three-­
imaging data from ischemic stroke patients treated with endovas- dimensional (3D) TOF-­MRA technique (flip angle 25 degrees, 1 signal
cular treatment (EVT) consecutively between January 2012 and acquired, slice thickness of 0.5 mm). The MRA of TIA patients com-
August 2017 at the Stroke Center of the University Hospital Berne, prised 200 slices obtained with a 3D TOF-­MRA technique (flip angle
Switzerland. Additionally, a cohort of TIA patients (n = 100) present- 20 degrees, 1 signal acquired, slice thickness of 0.6 mm, and six slaps
ing between June 2010 and October 2015 at the Stroke Center of with an overlap of 19% were used). Images were reconstructed and
the University Hospital Zurich, Switzerland, was analyzed. The study analyzed in the axial, coronal or sagittal plane with a maximum inten-
was performed according to the ethical guidelines of the Canton of sity projection (MIP) or source-­imaging algorithm.
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
3684 WESTPHAL et al.

The CoW of stroke patients was assessed by two independent summarized as counts and percentages of total. The Mann−Whitney
raters experienced in neuroradiologic image analysis (LPW and U test was applied for two-­group exploratory comparisons of continu-
NL) after prior training of the readout by a board-­certified senior ous variables and the Fisher`s exact test for comparisons of categorical
neuroradiologist (SWi). In diverging cases, a consensus decision variables. Two different logistic regression models were fitted to model
was reached after a second imaging analysis, if necessary with the favorable outcome defined as 3-­month mRS 0–­2 and to estimate the
assistance of a board-­certified senior neuroradiologist (SWi). The strength of association of the four vascular models as well as anterior,
anterior circulation was considered as differing if the Acom or the posterior and all CoW variants on primary outcome with adjustment for
A1 segment of the anterior cerebral artery (ACA) was hypoplastic clinical confounders (age, sex and atrial fibrillation). In the multivariate
(<0.8 mm with rounded values, <0.75 mm in absolute values) and logistic regression analysis, we adjusted for the clinically most relevant
considered as incomplete if the Acom or one A1 segment was unde- risk factors, which were significant in the univariate analysis (Table 1)
tectable (aplastic). The posterior circulation was categorized as dif- and added sex as a common clinical confounder. NIHSS on admission
fering if one or both Pcom arteries or P1-­segments of the posterior was excluded from the logistic regression analysis as it could be influ-
cerebral artery (PCA) were hypoplastic (<0.8 mm with rounded val- enced by collateral flow and CoW variants, thereby potentially mediat-
ues, <0.75 mm in absolute values) [3,4] and classified as incomplete ing outcome. Statistical model I included all vascular models, anterior
if one or both Pcom arteries or PCAs were undetectable (aplastic) and posterior variants, and statistical model II only the vascular models
including cases with a full fetal PCA origin. simultaneously. An interaction term between the vascular models 1–­4
After assessing CoW variants, we categorized the CoW of pa- was included in each model, but removed if there was no evidence for
tients into four vascular models, which we developed according to an interaction (i.e., the corresponding p value of the interaction term was
previously reported hypotheses of collateral flow via the CoW [5,6]. >0.05). Missing values were considered as missing completely at ran-
A schematic visualization of the models and the presumed collat- dom, therefore a complete case analysis was conducted.
eral blood flow dynamics via the CoW is shown in Figure 1a with All calculations were performed using STATA 14.1 and R 3.6.0
exemplary findings of these models on TOF-­MRA images demon- (R Core Team, 2019). The study was reported according to the
strated in Figure 1b. Model 1 has a missing or hypoplastic ACA-­A1-­ STROBE (Strengthening the Reporting of Observational studies in
segment ipsilateral to the side of MCA-­M1-­occlusion with presumed Epidemiology) guidelines for observational studies [8].
collateral flow via the Acom towards the differing A1-­segment and
stroke side thereby likely to provide better collateral flow via lep-
tomeningeal collaterals (LMC) [6,7]. Model 2 describes a missing or R E S U LT S
hypoplastic A1-­segment contralateral to stroke and M1-­occlusion
with an expected reverse flow via the Acom to the impaired A1-­ Stroke and TIA patient data and baseline
segment possibly leading to a reduction of collateral flow via the characteristics
CoW to the stroke side [7]. Model 3 has a fetal PCA ipsilateral to
stroke and M1-­occlusion side with a presumed collateral blood flow We screened 274 acute ischemic stroke patients with MCA-­M1-­
via the contralateral Pcom, A1 and Acom. Model 4 corresponds to occlusion and included 193 into the analysis (Figure 2). A total of
an impaired posterior circulation including a-­or hypoplastic Pcom or 52 patients were excluded either because of insufficient image
P1-­segments ipsilateral to stroke and M1-­occlusion with a hypothe- quality due to excessive head movement or other imaging artifacts
sized worse clinical outcome due to reduced collateral flow from the (n = 32), additional LVO other than extension of the occlusion to
ipsilateral Pcom to the anterior circulation [7]. the ICA (n = 9), prior territorial infarction (n = 7) or missing angi-
ography (n = 4). Furthermore, 29 cases with extra-­ or intracranial
ICA-­occlusion as a possible confounder for outcome analyses were
Outcome measures excluded, leading to an overall dataset of 193 stroke patients avail-
able for the analysis of CoW variants. Additionally, we screened
Outcome measures were assessed for the cohort of stroke patients. a cohort of 100 patients who presented with suspicion of a TIA.
The mRS for 3-­month functional impairment after stroke was de- Twenty-­seven cases were excluded due to a final diagnosis other
fined as the primary outcome, whereas mortality defined as death than TIA or missing MRI data, leading to a set of 73 patients diag-
occurring within the first 3 months after stroke onset was set as the nosed with a TIA without diffusion restriction or intracranial vessel
secondary outcome. Favorable outcome was defined as mRS 0–­2, occlusion on follow-­up MRI.
unfavorable outcome as mRS 3–­6. As presented in Table 1, the median age of stroke patients was
73.4 (IQR: 61–­82) years, 61.1% (n = 118) were females and 92.5%
(n = 160) were independent prior to stroke. Among the vascular risk
Statistical analysis factors, arterial hypertension with 66.3% (n = 128) and dyslipidemia
with 58.6% (n = 112) were the most frequent. The median NIHSS on
For the descriptive analyses, median and interquartile range (IQR) were admission was 12 (IQR: 8–­17). 48.7% (n = 94) of patients underwent
used for continuous variables, whereas categorical variables were intravenous thrombolysis, all patients were referred to EVT. Most
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
CIRCLE OF WILLIS VARIANTS IN LVO STROKE 3685

F I G U R E 1 Scheme of vascular models

of circle of Willis (CoW) variants (a) and
visualization in time-­of-­flight magnetic
resonance angiography (TOF-­MRA).
(b) Axial magnetic resonance images of
acute ischemic stroke patients performed
on admission showing the CoW in
3D-­multi-­slab TOF-­angiography in the
maximum intensity projection (MIP).
CoW variant indicated with a white
arrow on each image. 1: Hypoplastic
A1-­segment ipsilateral to M1-­occlusion
on the right side; 2: aplastic A1-­segment
contralateral to M1-­occlusion on the right
side; 3: full fetal PCA ipsilateral to M1-­
occlusion on the left side; 4: hypoplastic
Pcom ipsilateral to M1-­occlusion on
the right side. Abbreviations: Acom,
anterior communicating artery; ICA,
internal carotid artery; MCA, middle
cerebral artery; PCA, posterior cerebral
artery; Pcom, posterior communicating
artery [Colour figure can be viewed at
[Link]]

patients (54.4%, n = 105) were assigned to vascular model 4 (ipsilat- hypoplastic vessels of the anterior (ACA-­A1 or Acom) or posterior
eral a-­ or hypoplastic PCA or Pcom) with 49.2% (n = 95) of all CoW circulation (PCA-­P1 or Pcom) as shown in Table 2. From these, 39.9%
variants being an ipsilateral a-­ or hypoplastic Pcom (see Table 1). In (n = 77) had variants of the anterior circulation and 77.2% (n = 149)
the anterior circulation, an a-­ or hypoplastic A1-­segment (vascular a differing or incomplete posterior circulation (Table 2). Among an-
model 1) was observed in 7.8% (n = 15) on the ipsilateral side and in terior CoW variants, a differing Acom was the most frequent find-
6.2% (n = 12) on the contralateral side of the M1-­occlusion (Table 1). ing (n = 59, 30.6% of all CoW variants), whereas a hypoplastic P1 or
Pcom was most often observed in posterior CoW variants (n = 114,
59%) (see Table 2).
Distribution of CoW variants in stroke and In TIA patients without intracranial vessel occlusion, we found
TIA patients a total amount of 72.6% (n = 53) of CoW variants as also presented
in Table 2 (vs. 82.4% [n = 159] in stroke patients, p = 0.088). From
Some 82.4% (n = 159) of ischemic stroke patients with MCA-­M1-­ these, 45.2% (n = 33) were anterior circulation variants (vs. 39.9%
occlusion showed a differing or incomplete CoW including a-­ or [n = 77] in stroke patients, p = 0.486). Posterior circulation variants
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
3686 WESTPHAL et al.

TA B L E 1 Baseline characteristics of stroke patients

Outcome Mortality

Favorablea Unfavorableb
Characteristic All (n = 193) (n = 104) (n = 89) p Alive (n = 162) Dead (n = 31) p

Demographic factors
Age, median (IQR) 73.4 (61–­82) 68.7 (57–­77) 79.7 (68–­86) <0.001 70.9 (60–­79) 83.4 (79–­88) <0.001
Female sex, n (%) 118 (61.1) 62 (59.6) 56 (62.9) 0.66 100 (61.7) 18 (58.1) 0.69
Independent prior stroke, 160 (92.5) 93 (96.9) 67 (87) <0.05 139 (95.2) 21 (77.8) <0.01
n (%)
Vascular risk factors, n (%)
Atrial fibrillation 84 (43.8) 37 (35.9) 47 (52.8) <0.05 66 (41.0) 18 (58.1) 0.11
Diabetes mellitus 33 (17.1) 11 (10.6) 22 (24.7) <0.05 26 (16.1) 7 (22.6) 0.43
Arterial hypertension 128 (66.3) 58 (55.8) 70 (78.7) <0.01 106 (65.4) 22 (71) 0.68
Dyslipidemia 112 (58.6) 64 (61.5) 48 (55.2) 0.38 100 (62.1) 12 (40) <0.05
Current smoking 40 (22.9) 27 (27.6) 13 (16.9) 0.11 37 (24.8) 3 (11.5) 0.20
Coronary heart disease 30 (15.5) 17 (16.4) 13 (14.6) 0.84 24 (14.8) 6 (19.4) 0.59
Peripheral artery disease 6 (3.4) 1 (1) 5 (6.3) 0.09 4 (2.7) 2 (7.4) 0.23
Prior stroke or TIA 27 (13.9) 15 (14.2) 12 (13.5) 1.00 22 (13.6) 5 (16.1) 0.78
Previous medication, n (%)
Antiplatelets 61 (31.6) 27 (26.0) 34 (38.2) 0.09 48 (29.6) 13 (41.9) 0.21
Oral anticoagulants 20 (10.4) 10 (9.6) 10 (11.2) 0.81 15 (9.3) 5 (16.1) 0.33
Statins 41 (21.4) 21 (20.4) 20 (22.5) 0.73 36 (22.4) 5 (16.1) 0.63
Antihypertensives 111 (57.5) 52 (50.0) 59 (66.3) <0.05 91 (56.2) 20 (64.5) 0.43
Clinical parameters, median (IQR)
NIHSS on admission 12 (8–­17) 11 (8–­15) 15 (10–­19) 0.001 12 (8–­16) 16 (8–­20) <0.05
Systolic blood pressure 152 (133–­168) 146 (131–­161) 158 (136–­173) <0.05 150 (133–­167) 160 (144–­175) 0.1
Diastolic blood pressure 82 (70–­95) 81 (70–­92) 82 (69–­96) 0.76 81 (70–­93) 85 (75–­100) 0.25
Onset to groin puncture 217 (163–­391) 210 (160–­316) 240 (165–­4 47) 0.34 212 (163–­364) 245 (159–­476) 0.57
in min
TICI score after EVTc , n (%)
0 10 (5.2) 2 (1.9) 8 (9) <0.05 7 (4.3) 3 (9.7) 0.076
1 8 (4.2) 2 (1.9) 6 (6.7) 4 (2.5) 4 (12.9)
2a 24 (12.4) 13 (12.5) 11 (12.4) 21 (13) 3 (9.7)
2b 68 (35.2) 34 (32.7) 34 (38.2) 58 (35.8) 10 (32.3)
3 83 (43) 53 (51) 30 (33.7) 72 (44.4) 11 (35.5)
General anesthesia 88 (46.6) 45 (45.5) 41 (46.6) 0.88 71 (45.5) 15 (48.4) 0.85
LMC status, n (%)
Weak or no (0) 20 (10.4) 9 (8.7) 11 (12.4) 0.43 18 (11.1) 2 (6.5) 0.71
Moderate (1) 68 (35.2) 35 (33.7) 33 (37.1) 58 (35.8) 10 (32.3)
Good (2) 102 (52.9) 60 (57.7) 42 (47.2) 84 (51.9) 18 (58.1)
CoW incomplete/differing, n (%) 159 (82.4) 85 (81.7) 74 (83.2) 0.85 130 (80.3) 29 (93.6) 0.119
Vascular models, n (%)
1 (ipsilateral a-­/hypoplastic 15 (7.8) 9 (8.7) 6 (6.7) 0.79 13 (8) 2 (6.5) 1.00
A1)
2 (contralateral a-­/ 12 (6.2) 7 (6.7) 5 (5.6) 1.00 10 (6.2) 2 (6.5) 1.00
hypoplastic A1)
3a (ipsilateral full fetal PCA) 11 (5.7) 8 (7.7) 3 (3.4) 0.23 9 (5.6) 2 (6.5) 0.69
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
CIRCLE OF WILLIS VARIANTS IN LVO STROKE 3687

TA B L E 1 (Continued)

Outcome Mortality

Favorablea Unfavorableb
Characteristic All (n = 193) (n = 104) (n = 89) p Alive (n = 162) Dead (n = 31) p

3b (two-­sided full fetal PCA) 2 (1) 2 (1.9) 0 (0) 0.50 2 (1.2) 0 (0) 1.00
4 (ipsilateral a-­/hypoplastic 105 (54.4) 60 (57.7) 45 (50.6) 0.39 86 (53.1) 19 (61.3) 0.44
PCA/Pcom)
4a (ipsilateral hypoplastic 73 (37.8) 40 (38.5) 33 (37.1) 0.88 61 (37.7) 12 (38.7) 1.00
PCA/Pcom)
4b (ipsilateral aplastic PCA/ 32 (16.6) 20 (19.2) 12 (13.5) 0.33 25 (15.4) 7 (22.6) 0.31
Pcom)
4c (two-­sided a-­/hypoplastic 89 (46.1) 48 (46.2) 41 (46.1) 1.00 73 (45.1) 16 (51.6) 0.56
PCA/Pcom)
4d (ipsilateral a-­/hypoplastic 95 (49.2) 55 (52.9) 40 (44.9) 0.31 78 (48.2) 17 (54.8) 0.56
Pcom)

Abbreviations: CoW, circle of Willis; DSA, digital subtraction angiography; EVT, endovascular treatment; IQR, interquartile range; LMC,
leptomeningeal collateral; mRS, modified Rankin scale; NIHSS, National Institutes of Health Stroke Scale; PCA, posterior cerebral artery; Pcom,
communicating posterior artery; TIA, transient ischemic attack; TICI, Thrombolysis In Cerebral Infarction. Bold values mark statistically significant
values.
a
(mRS ≤ 2).
b
(mRS ≥ 3–­6).
c
LMC status assessed during initial DSA; TICI scale: 0, no perfusion; 1, penetration with minimal perfusion; 2a, partial perfusion <2/3 of the entire
vascular territory; 2b, complete filling of the expected vascular territory, but with a perceptibly slower filling rate; 3, complete perfusion.

Association of CoW variants with clinical parameters

Ischemic stroke patients with
MCA-M1-occlusion screened in stroke and TIA patients
between 2012-17 in Berne with
MR perfusion data (n=274)
When comparing stroke patients with a complete CoW to patients
Multiple occlusion sites (n=9) with any CoW variant or only anterior or posterior variants, the median
Insufficient image quality (n=32) age of all groups with CoW variants was higher compared to the group
Prior territorial infarction (n=7) without variants (normal CoW: 68.7 [IQR: 56–­80] years; all CoW vari-
Missing angiography (n=4) ants: 74.5 [IQR: 62–­82] years, p = 0.22). Within the group of posterior
variants, there was evidence for a higher median age of stroke patients
MRA data (n=222) with variants (74.7 [IQR: 64–­82] years vs. 67.7 [IQR: 55–­79] years) (see
Table S1). In TIA patients, we found a median age of 74 years (IQR:
Additional ICA occlusion/stenosis
(n=29) 65–­80) in patients with any CoW variant compared to 69 years (IQR:
55–­77) in patients with a normal CoW (p = 0.19) as shown in Table
Final data set (n=193) S1). Regarding stroke severity, there was no evidence for a differing
NIHSS on admission between groups with different CoW variants.
F I G U R E 2 Flow chart of study population. Abbreviations: ICA, Furthermore, there was no evidence for a difference between the
internal carotid artery; MRA, magnetic resonance angiography groups with and without variants for neither the LMC status assessed
during the initial digital subtraction angiography (DSA), nor success of
recanalization assessed by the TICI scale (Table S1).
were detected less frequently in TIA patients (58.9% [n = 43] vs.
77.2% [n = 149] in stroke patients, p = 0.004). A detailed list of the
distribution of the different CoW variants of both patient cohorts Outcome analysis of CoW variants in stroke patients
can be found in Table 2. The median age of TIA patients was 73 years
(IQR: 64–­78 years, compared to stroke patients, p = 0.582). There In patients with any CoW variant there was a trend towards a higher
were more women in the stroke group (61.1% [n = 118] vs. 39.7% mortality compared to patients with an intact CoW (18.2% vs. 5.9%,
[n = 29] in TIA patients, p = 0.002). When adjusting stroke and TIA p = 0.08), in particular in patients with posterior circulation variants
patients for sex, we found no evidence for a differing amount of all (p = 0.058) (see Table S1). When stratifying the group of all CoW var-
CoW variants or anterior variants, while the difference in posterior iants (n = 159) within the stroke cohort for outcome and mortality,
variants between stroke and TIA patients remained (p = 0.0076). respectively, there was a tendency towards more incomplete CoW
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
3688 WESTPHAL et al.

TA B L E 2 Distribution of circle of Willis variants in stroke and transient ischemic attack patients

All All
Stroke cohort (n = 193) TIA cohort (n = 73)

CoW, n (%)
Incomplete/differinga 159 (82.4) Incomplete/differinga 53 (72.6)
Anterior CoW, n (%)
A1 or Acom incomplete/differing 77 (39.9) A1 or Acom incomplete/differing 33 (45.2)
A1 or Acom hypoplastic 65 (33.7) A1 or Acom hypoplastic 30 (41.1)
A1 or Acom aplastic 15 (7.8) A1 or Acom aplastic 3 (4.1)
A1-­asymmetry 26 (13.5) A1-­asymmetry 10 (13.7)
A1-­asymmetry (ipsilateral) 15 (7.8) A1-­asymmetry (ipsilateral) NA
A1 hypoplastic (ipsilateral) 9 (4.7) A1 hypoplastic 8 (11)
A1 aplastic (ipsilateral) 6 (3.1) A1 aplastic 2 (2.7)
Acom differing 59 (30.6) Acom differing 23 (31.5)
Acom hypoplastic 54 (28.0) Acom hypoplastic 22 (30.1)
Acom aplastic 5 (2.6) Acom aplastic 1 (1.4)
Azygos variant 2 (1) Azygos variant 1 (1.4)
Posterior CoW, n (%)
P1 or Pcom incomplete/differing 149 (77.2) P1 or Pcom incomplete/differing 43 (58.9)
P1 or Pcom incomplete/differing (ipsilateral) 105 (54.4) P1 or Pcom incomplete/differing (ipsilateral) NA
P1 or Pcom hypoplastic 114 (59.1) P1 or Pcom hypoplastic 32 (43.8)
P1 or Pcom hypoplastic (ipsilateral) 74 (38.3) P1 or Pcom hypoplastic (ipsilateral) NA
P1 or Pcom aplastic 55 (28.5) P1 or Pcom aplastic 12 (16.4)
P1 or Pcom aplastic (ipsilateral) 31 (16.1) P1 or Pcom aplastic (ipsilateral) n.a.
P1 or Pcom one-­sided differing 60 (31.1) P1 or Pcom one-­sided differing 29 (39.7)
P1 or Pcom two-­sided differing 89 (46.1) P1 or Pcom two-­sided differing 14 (19.2)
Pcom differing (ipsilateral) 95 (49.2) Pcom differing 39 (53.4)
Pcom hypoplastic (ipsilateral) 64 (33.2) Pcom hypoplastic 28 (38.4)
Pcom aplastic (ipsilateral) 31 (16.1) Pcom aplastic 12 (16.4)
Full fetal PCA 21 (10.9) Full fetal PCA 4 (5.5)
Full fetal PCA (ipsilateral) 11 (5.7) Full fetal PCA (ipsilateral) NA
Partial fetal PCA 54 (28.0) Partial fetal PCA 16 (21.9)
Partial fetal PCA (ipsilateral) 38 (19.7) Partial fetal PCA (ipsilateral) NA

Abbreviations: Acom, anterior communicating artery; CoW, circle of Willis; NA, not applicable; PCA, posterior cerebral artery; Pcom, posterior
communicating artery; TIA, transient ischemic attack.
aIncluding all CoW variants.

among patients dying within 3 months after stroke onset (93.6% that either one of the vascular models or the variants had a reliable ef-
[n = 29] vs. 80.3% [n = 130], p = 0.119) (see Table 1). However, there fect on outcome (the estimated change in odds ratio and 95% confidence
was no evidence for a differing outcome after 3 months. An illustra- intervals are shown in Figure 4). We additionally performed the analysis
tion of the distribution of 3-­month mRS stratified for patients with a including only the four vascular models (statistical model II) and again
complete CoW, all, anterior or posterior variants and the four vascu- there was no evidence for an association of the included variables with
lar models is presented in Figure 3. favorable outcome. In both statistical models, there was no evidence for
Since we assumed that an imbalance of clinical variables such as an interaction between the vascular models 1–­4.
slightly older age in patients with CoW variants might affect our out-
come comparison, we conducted a logistic regression analysis in order to
evaluate the association of CoW variants on favorable clinical outcome DISCUSSION
adjusted for clinical confounders. When analyzing the four vascular
models and also adding anterior and posterior variants as exploratory CoW variants are frequent and their presence and shape have been
variables (statistical model I), we did not find enough evidence to claim associated with ischemic stroke previously [4,6,9]. However, it is
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
CIRCLE OF WILLIS VARIANTS IN LVO STROKE 3689

unknown if CoW variants affect outcome in MCA-­M1-­occlusion or early outcome in a heterogenic cohort including different stroke
stroke after EVT. etiologies.
Our aim was to assess the frequency and distribution of CoW Similar to other studies, we found that older age was associated
variants in a MCA-­M1-­occlusion stroke and TIA cohort without in- with a higher frequency of CoW variants in patients with stroke
tracranial vessel occlusion and to investigate the potential impact of [3,9,11,12] which is likely to affect outcome analyses, if not adjusted.
CoW variants on stroke outcome. Our hypothesis was that an acute Results from a twin study by Forgo et al. [1] suggest that environ-
intracranial vessel occlusion might be partially compensated by col- mental rather than genetic effects determine CoW variants after
lateral flow via the CoW in the setting of a complete CoW, but less demonstrating a high rate of discordance of CoW variants in mono-
tolerated in an incomplete CoW. However, some conditions such zygote twins. Another anatomical study demonstrated that some hy-
as a missing A1-­segment ipsilateral to an acute MCA-­M1-­occlusion poplastic arteries have inward vascular remodeling consistent with
(model 1) could be beneficial by providing additional collateral flow atherosclerotic arterial occlusion [13] which might explain to some
via a well-­established Acom and LMC towards the missing A1-­ extent the higher frequency of CoW variants in older individuals.
segment and stroke side [6,7]. This led to the development of four Except for fewer posterior variants in the TIA cohort, we found
vascular models of presumed flow compensation in cases of MCA-­ no evidence for a difference in frequency and distribution of CoW
M1-­occlusion stroke (Figure 1) according to previously described variants in TIA compared to stroke patients. The difference in poste-
flow dynamics in patients with LVO [5–­7]. We found that a high rior variants remained after adjusting both cohorts for sex. Although
(82.4%) proportion of patients with MCA-­M1-­occlusion stroke had some imbalance in the rather small sample size of TIA patients might
an incomplete or hypoplastic CoW affecting predominantly the pos- contribute to this finding, we cannot rule out that certain CoW vari-
terior circulation (77.2%), which is in line with previous data [10]. ants, such as posterior variants, render patients more susceptible to
Despite a trend towards a higher mortality in patients with any MCA-­M1-­occlusion stroke.
CoW variant in our study, the presence of any type of CoW vari- The strengths of our study are the clinically well-­characterized
ant was not independently associated with 3-­month functional out- patient cohort with a homogenous intracranial vessel status, includ-
come after adjusting for potential clinical confounders, particularly ing only patients with stroke due to MCA-­M1-­occlusion and no addi-
age. Furthermore, none of the vascular models of presumed collat- tional extra-­or intracranial vessel stenosis or occlusion. Furthermore,
eral flow in different conditions of missing CoW segments could be the CoW was assessed by two blinded and independent physicians
confirmed in our outcome analysis. In line with our results, a recent with neuroradiological expertise trained and supervised by a board-­
study analyzing ipsilateral CoW variants in patients with MCA-­ or certified neuroradiologist. Additionally, we analyzed data from a TIA
ICA-­occlusion found no evidence for an effect of these CoW vari- cohort in order to assess the distribution of CoW variants in a similar
ants on clinical outcome [11]. However, in that analysis, contralateral vascular patient cohort without any intracranial vessel occlusion.
CoW variants as well as Acom variants were not included [11]. In Limitations of the study are the small sample size of subgroups
line with our results, a study by Shaban et al. [12] could not find an among the different CoW variants and classification into vascular
association between the presence of a fetal PCA and stroke severity models and subgroups of variants, which affect the power of our

17.6% 82.4% 39.9% 77.2% 7.8% 6.2% 6.7% 54.4%

(n=34) (n=159) (n=77) (n=149) (n=15) (n=12) (n=13) (n=105)

no V all V ant. V post. V M1 M2 M3 M4

1.00

0.75

F I G U R E 3 Distribution of 3-­month
modified Rankin scale (mRS) among circle
percentage

of Willis (CoW) variants. No V = no CoW 0.50

variants, all V = all CoW variants, ant.
V = all anterior circulation variants, post.
V = all posterior circulation variants,
M1 = model 1 (ipsilateral a-­or hypoplastic 0.25
A1), M2 = model 2 (contralateral
a-­or hypoplastic A1), M3 = model 3
(ipsilateral fetal posterior cerebral artery),
M4 = model 4 (ipsilateral a-­or hypoplastic 0.00
P1 or posterior communicating artery)
[Colour figure can be viewed at
mRS 3 months: 0 1 2 3 4 5 6
[Link]]
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
3690 WESTPHAL et al.

F I G U R E 4 Estimated changes in odds

Model 1 (present)
(ipsilateral a− or hypoplastic A1)
ratio of favorable outcome and 95%
1.84 (95% CI from 0.52 to 6.88)
confidence intervals for statistical models
Model 2 (present) I and II. Odds ratios >1 indicate that the
(contralateral a− or hypoplastic A1)
2.39 (95% CI from 0.62 to 9.77)
association increases the chance for

Statistical model I
Model 3 (present) favorable outcome. Abbreviations: Ant.,
(ipsilateral fetal PCA)
2.74 (95% CI from 0.66 to 14.23) anterior; PCA, posterior cerebral artery;
Model 4 (present) Pcom, posterior communicating artery;
(ipsilateral a− or hypoplastic P1 or Pcom)
1.43 (95% CI from 0.65 to 3.15) Post., posterior

Ant. Variants (present)

0.51 (95% CI from 0.23 to 1.07)

Post. Variants (present)

1.32 (95% CI from 0.51 to 3.45)

Model 1 (present)
(ipsilateral a− or hypoplastic A1)
1.16 (95% CI from 0.36 to 3.85)

Model 2 (present)

Statistical model II
(contralateral a− or hypoplastic A1)
1.55 (95% CI from 0.45 to 5.69)

Model 3 (present)
(ipsilateral fetal PCA)
2.36 (95% CI from 0.58 to 12.07)

Model 4 (present)
(ipsilateral a− or hypoplastic P1 or Pcom)
1.58 (95% CI from 0.84 to 3.03)

0 1 2 4 6 8 10 12 14
Odds ratio

study. Furthermore, MRA may underestimate the size of vessels C O N FL I C T O F I N T E R E S T S

within the CoW compared to DSA as imaging gold standard [4]. Due SWi has received a research grant from the Promedica Foundation,
to the small size of CoW segments, the true number of patients with travel honoraria from Bayer and has received speaker honoraria
a complete CoW might be higher. Nevertheless, the amount of dif- from Siemens. JMH has received a travel grant from Pfizer. JK
ferent CoW variants we found was in accordance with other studies has received travel grants from Pfizer and Stryker and academic
using different methods [9,10]. Furthermore, the restriction to an grants from the SAMW/Bangerter Foundation and the Swiss Stroke
LVO stroke cohort might have influenced or overlapped the effect Society. SWe has received speaker honoraria from Amgen, travel
of CoW variants on outcome. honoraria from Bayer, a research grant from Boehringer Ingelheim
In conclusion, our study shows that CoW variants are a frequent and academic grants from the Swiss National Science Foundation,
finding in ischemic stroke patients with MCA-­M1-­occlusion. There the UZH (Clinical Research Priority Program Stroke), the Swiss Heart
was no evidence for an association of CoW variants with 3-­month Foundation and the Olga Mayenfish Foundation. LPW, NL, CM, UH,
functional disability. However, the level of evidence indicates a trend KS, MEA, TD, LDP, UF, MRH, ARL, JG, MA and RW report no disclo-
towards a higher mortality in patients with any type of CoW variant sures. All authors do not report any conflicts of interests.
and a higher frequency of incomplete CoW among patients dying
within 3 months after stroke onset represent valuable hints to val- AU T H O R C O N T R I B U T I O N S
idate these results in future studies. In addition, the role of CoW Laura P. Westphal: Data curation (equal); Formal analysis (lead);
variants for intracranial vessel occlusions other than MCA-­M1 and Investigation (lead); Methodology (equal); Visualization (equal);
the association of the LMC status among the different subgroups of Writing-­original draft (lead). Niklas Lohaus: Data curation
CoW variants with outcome could be addressed in further studies (equal); Writing-­review & editing (equal). Sebastian Winklhofer:
with larger patient cohorts. Methodology (equal); Supervision (equal); Validation (equal);
Writing-­review & editing (equal). Christian Manzolini: Data curation
AC K N OW L E D G M E N T S (equal); Writing-­review & editing (equal). Ulrike Held: Methodology
Support from the Swiss National Science Foundation (SNSF (lead); Supervision (equal); Writing-­review & editing (equal). Klaus
PP00P3_170683) and the UZH Clinical Research Priority Program Steigmiller: Methodology (equal); Writing-­review & editing (equal).
(CRPP) Stroke is acknowledged. Open Access Funding provided by Janne M. Hamann: Data curation (equal); Writing-­review & editing
Universitat Zurich. (equal). Mohamad El Amki: Visualization (equal); Writing-­review &
 |

14681331, 2021, 11, Downloaded from [Link] by CAPES, Wiley Online Library on [31/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
CIRCLE OF WILLIS VARIANTS IN LVO STROKE 3691

6. Hendrikse J, Hartkamp MJ, Hillen B, et al. Collateral ability of the

editing (equal). Tomas Dobrocky: Data curation (equal); Writing-­ circle of Willis in patients with unilateral internal carotid artery
review & editing (equal). Leonidas D. Panos: Data curation (equal); occlusion: border zone infarcts and clinical symptoms. Stroke.
Writing-­review & editing (equal). Johannes Kaesmacher: Data 2001;32:2768-­2773.
7. Zhu G, Yuan Q, Yang J, Yeo JH. The role of the circle of Willis in
curation (equal); Writing-­review & editing (equal). Urs Fischer:
internal carotid artery stenosis and anatomical variations: a compu-
Writing-­review & editing (equal). Mirjam R. Heldner: Data cura-
tational study based on a patient-­specific three-­dimensional model.
tion (equal); Writing-­review & editing (equal). Andreas R. Luft: Biomed Eng Online. 2015;14:107.
Writing-­review & editing (equal). Jan Gralla: Writing-­review & 8. Von Elm E, Altman DG, Egger M, et al. The Strengthening the
editing (equal). Marcel Arnold: Writing-­review & editing (equal). Reporting of Observational Studies in Epidemiology (STROBE)
statement: guidelines for reporting observational studies. Lancet.
Roland Wiest: Writing-­review & editing (equal). Susanne Wegener:
2007;370:1453-­1457.
Conceptualization (lead); Formal analysis (equal); Funding acquisi- 9. Zhou H, Sun J, Ji X, et al. Correlation between the integrity of the
tion (lead); Supervision (lead); Writing-­review & editing (lead). circle of Willis and the severity of initial noncardiac cerebral infarc-
tion and clinical prognosis. Medicine (Baltimore). 2016;95:e2892.
10. Hamming AM, Van Walderveen MAA, Mulder IA, et al. Circle of
DATA AVA I L A B I L I T Y S TAT E M E N T
Willis variations in migraine patients with ischemic stroke. Brain
Anonymized data will be shared on request from any qualified Behav. 2019;9:e01223.
investigator. 11. Seifert-­Held T, Eberhard K, Christensen S, et al. Circle of Willis vari-
ants are not associated with thrombectomy outcomes. Stroke Vasc
Neurol. 2021;6:310-­313.
ORCID
12. Shaban A, Albright KC, Boehme AK, Martin-­Schild S. Circle of Willis
Laura P. Westphal [Link] variants: fetal PCA. Stroke Res Treat. 2013;2013:105937.
Mirjam R. Heldner [Link] 13. Alpers BJ, Berry RG, Paddison RM. Anatomical studies of the
Susanne Wegener [Link] circle of Willis in normal brain. AMA Arch Neurol Psychiatry.
1959;81:409-­418.

REFERENCES
1. Forgo B, Tarnoki AD, Tarnoki DL, et al. Are the variants of the circle S U P P O R T I N G I N FO R M AT I O N
of Willis determined by genetic or environmental factors? Results Additional supporting information may be found online in the
of a twin study and review of the literature. Twin Res Hum Genet.
Supporting Information section.
2018;21:384-­393.
2. Krishnaswamy A, Klein JP, Kapadia SR. Clinical cerebrovascular
anatomy. Catheter Cardiovasc Interv. 2010;75:530-­539.
3. Krabbe-­Hartkamp MJ, Van Der Grond J, De Leeuw FE, et al. Circle How to cite this article: Westphal LP, Lohaus N, Winklhofer S,
of Willis: morphologic variation on three-­dimensional time-­of-­flight et al. Circle of Willis variants and their association with
MR angiograms. Radiology. 1998;207:103-­111. outcome in patients with middle cerebral artery-­M1-­occlusion
4. Van Seeters T, Hendrikse J, Biessels GJ, et al. Completeness of the
stroke. Eur J Neurol. 2022;28:3682–­3691. [Link]
circle of Willis and risk of ischemic stroke in patients without cere-
brovascular disease. Neuroradiology. 2015;57:1247-­1251. org/10.1111/ene.15013
5. Liebeskind DS. Collaterals in acute stroke: beyond the clot.
Neuroimaging Clin N Am. 2005;15:553-­573.