Incomplete variants of the circle of Willis and stroke outcome

Abstract

Background:

There is considerable variation in circle of Willis morphology among the general population, and these variations have been correlated with risk of aneurysms, cerebral ischemia, and other clinical events.

Purpose:

To investigate the relationship between circle of Willis variants and stroke outcome.

Materials and Methods:

We performed a retrospective study involving 297 patients from our institution’s acute stroke academic registry. All received MRA examinations of the head upon admission for acute strokes. All imaging was reviewed to assess for circle of Willis variants (particularly A1 and P1 aplasia or hypoplasia) along with vertebral artery aplasia or hypoplasia. Stroke outcome was defined as good (walking independently at the time of discharge) or poor (inability to walk at discharge, assistance needed to walk at discharge, or death). Severity of stroke was assessed using the National Institute of Health Stroke Scale.

Results:

An incomplete circle of Willis was seen in 34% of subjects. There was no significant association between age, gender, hypertension, or presence of arterial stenosis and circle of Willis completeness. Using logistic regression, we found that the presence of an incomplete circle of Willis decreased the odds of a stroke patient having a good outcome by 47% (p=0.046, OR 0.53, 95% CI 0.281– 0.988), after adjusting for age and severity of stroke at admission.

Conclusion:

This study suggests that an incomplete circle of Willis may be associated with a poorer prognosis for stroke patients.

Introduction

The circle of Willis (CoW) reflects a series of anastomotic connections between the anterior and posterior cerebral arterial circulations. Anteriorly, the circle is comprised of the bilateral internal carotid artery termini, the A1 segments of the anterior cerebral arteries, and the anterior communicating artery connecting the right and left circulation. The anterior and posterior circulation are connected by the posterior communicating arteries, with the P1 segments of the posterior cerebral arteries forming the posterior boundary of the circle.

The collateral blood flow enabled by the CoW helps maintain cerebral perfusion and protect against ischemia in the event of arterial disease. However, there is considerable morphological variation in cerebral arterial circulation among the general population, with numerous anatomic variants commonly seen in healthy patients.^1,2 For example, in one analysis of 525 healthy patients using 3D time-of-flight magnetic resonance angiography (MRA), a complete CoW was observed in only 20.9% of subjects.³ Similarly, in an examination of 1000 autopsy subjects, only 45.2% of subjects had typical anatomy along the CoW.⁴ Numerous other pathological and radiologic studies have confirmed that a complete CoW is only present in a minority of the population.^5,6 These variations in the CoW have also been shown to be relatively uniformly distributed across different races and genders.^7–9

While CoW variations are well established, the clinical significance of these variations remains a topic of research. Studies have suggested that variations in the CoW play a role in the development of cerebral aneurysms,^10,11 and an incomplete CoW is associated with a greater burden of white matter disease.^12–15 Incompleteness of the CoW has also been shown to increase the risk of intraoperative ischemia during endarterectomy,^16–19 while a meta-analysis demonstrated a positive association between CoW variants and development of ischemic stroke.²⁰ However, while CoW variations appear to increase the risk of strokes, it remains unclear if certain variants influence the prognosis of stroke patients.

The aim of this study is to retrospectively evaluate a group of stroke patients admitted to a comprehensive stroke center and to determine whether there is a significant difference in stroke outcomes for patients with CoW variants, after adjusting for two of the most important predictors of stroke outcome, age and severity of stroke at admission.²¹

Methods

A retrospective study was performed at an academic quaternary-care center with advanced certification as a Joint Commission Comprehensive Stroke Center. Using an academic registry of all acute stroke patients admitted to our institution between January 2011 and December 2016, we identified patients meeting the following inclusion criteria: (1) Patient experienced an acute cerebral ischemic infarction as confirmed by MR imaging, (2) patient underwent an MRA examination of the head during an index hospital admission for acute stroke, and (3) patient with initial NIHSS and stroke outcome information available in medical records (Figure 1). Patients were excluded if they had a hemorrhagic stroke or if there were concomitant intracranial comorbidities such as venous sinus thrombosis or intracranial tumors.

Figure 1:

Flow chart detailing sample selection

Overall, 1250 subjects met the eligibility criteria. MRA examinations for these patients were performed within 0–13 days of admission, but with the large majority (n=1089) of patients obtaining scans on either the day of admission (n=656) or within 1 day of admission (n=433). 103 patients received MRA examinations between days 2–3 from admission. Overall, 96% of patients had MRA exams of the head within 3 days of hospital admission. Out of 1250 patients meeting eligibility criteria, we selected a group of 400 subjects using the random sample selection option in SPSS. No clinical, demographic, or imaging information was incorporated in the selection process or known to the person performing the selection. Within the randomly selected group of 400 patients, 221 had MRA scans on the day of admission and 130 had scans within 1 day of admission. Overall, 96% of patients in the randomly selected group had MRA with 3 days of hospital admission. We first assessed the completeness of CoW for the group of 400 patients and then correlated results with clinical data.

Data regarding patient demographics and clinical characteristics were obtained from our institution’s acute stroke academic registry. Severity of stroke was assessed using the National Institute of Health Stroke Scale (NIHSS) at the time of admission. Stroke etiology was classified based on Trial of ORG 10172 in Acute Stroke Treatment (TOAST) classification. Stroke outcomes were also obtained from CAESAR and were defined in a binary fashion as either good (walking independently at the time of discharge) or poor (inability to walk at discharge, assistance needed to walk at discharge, or death). Here we report on 297 subjects for whom both admission NIHSS and outcome data were available. The institutional review board at our institution approved this study and waived the requirement for informed consent.

All imaging was performed on 1 of 3 MRI scanners: SIGNA Architect (3T), Discovery 750 (3T), or SIGNA Artist (1.5T) (GE Healthcare). All MRA examinations were performed using 3D time-of-flight technique with maximum intensity projection images. MRAs of all subjects were reviewed by a neurologist blinded to patient clinical data, with equivocal determinations resolved by a CAQ-certified neuroradiologist with 12 years of experience. The A1 segment of the anterior cerebral arteries, P1 segments of the posterior cerebral arteries, and the V4 segments of the vertebral arteries were categorized as either aplastic/hypoplastic or normal for each patient (Figure 2). Aplasia/hypoplasia (A/H) of the A1 and V4 segments was defined as a vessel diameter less than 50% of the contralateral side, while A/H of the P1 segment was defined as a vessel diameter less than 50% of the ipsilateral P2 segment. Absence of the anterior communicating artery occurred at a negligible rate and thus was not included in analysis. The posterior communicating artery was also excluded from analysis due to inverse correlation between vessel calibers of the P1 segment and posterior communicating artery, along with the propensity for slower or more turbulent flow within this vessel resulting in suboptimal delineation on MRA examinations, particularly on examinations performed on the 1.5 T magnet. The presence or absence of atherosclerotic arterial stenosis was qualitatively assessed and confirmed with original neuroradiologist reporting for the MRA examinations.

Figure 2:

MRA images demonstrating aplasia/hypoplasia of the left A1 segment (A), both left P1 and A1 segments (B), and left vertebral artery (C). A patient without one of the vascular variants considered in this study is also shown for comparison (D).

Statistical Analysis

Categorical variables were compared with χ² tests or Fisher exact test, depending on the number of cases. Group means for continuous variables were compared using t-test or - when appropriate - U Mann Whitney test. The relationship between stroke outcome and the presence of an incomplete CoW or vertebral artery A/H was examined using logistic regression. The models were adjusted for age and severity of stroke at admission (measured with NIHSS) as in Weimar et al.¹ Variables were retained only if they significantly contributed to the model. The most parsimonious models were considered final. Statistical significance was defined as a p value <0.05. SPSS (version 25, SPSS, Inc., Chicago, IL) software was used for all analyses.

Results

Of the 297 patients in this study, the mean age was 76 ± 12.2 years and 157 (53%) were female. Men (75 ± 12.6 years) and women (76 ± 12.2 years) did not differ significantly in age. Hypertension was present in 68% of subjects. Hypertensive subjects were older (77±11.2 vs. 73±13.8 years, U Mann-Whitney p=0.02). Distribution of hypertensive subjects did not differ significantly between men and women. At least one site of intracranial arterial stenosis/occlusion was present in 43% of subjects. Subjects with stenosis/occlusion were older (78±10.5 vs. 74±13.2 years, U Mann-Whitney p=0.02) and more often had hypertension (χ²=6.1, p=0.01), but the percentage of men and women did not significantly differ between those with and without stenosis/occlusion.

Overall, an incomplete CoW was seen in 34% of subjects (n=102), while 30% had vertebral artery A/H (n=89). The frequencies of assessed cerebral arterial variants are detailed in Table 1. Vertebral artery A/H was more common among subjects with an incomplete CoW (38% vs 26%, χ²= 4.2, p=0.04). Clinical and demographic characteristics on study subjects are detailed in relation to CoW completeness in Table 2.

Table 1.

Frequency of Assessed Cerebral Arterial Variants

A1 A/H		13%
P1 A/H		24.5%
	Bilateral	5%
	Unilateral	19.5%
Incomplete CoW (P1 or A1 A/H)		34%
	Both A1 and P1	3%
	A/H	31%
	P1 or A1 A/H
Vertebral Artery A/H		30%

Table 2.

Demographic and clinical characteristic of subjects with complete and incomplete CoW.

	Complete CoW (n=195)	Incomplete CoW (n=102)	p-value
Age	75.1 ± 12.8	76.5 ± 11.0	0.45
Sex (n, % Female)	103, 53%	54, 53%	0.98
Hypertension (n, %)	131, 67%	72, 71%	0.55
Any stenosis (n, %)	87, 45%	42, 41%	0.57
Vertebral artery A/H (n, %)	51, 26%	38, 38%	0.04
NIHSS at admission	4.0 ± 5.5	5.0 ± 7.2	0.88
Positive stroke outcome (n, %)	150, 77%	65, 64%	0.02
30 day mortality (n, %)	7, 3.5%	4, 4%	1.00
Stroke etiology (n, %)
Small vessel occlusion	30, 15%	18, 18%	0.71
Large artery atherosclerosis	24, 12%	14, 14%
Cryptogenic	65, 33%	27, 26%
Cardioembolic	53, 27%	30, 29%
Multiple	12, 6%	10, 10%
Other	2, 1%	0, 0%
Incomplete workup	9, 4.5%	3, 3%

Vascular variants – relationship with demographic and clinical characteristics.

There was no significant association between age, gender, hypertension or arterial stenosis and CoW completeness or vertebral artery A/H.

CoW Variants and stroke outcome

Using logistic regression, we found that the presence of an incomplete CoW decreased the odds of a stroke patient having a good outcome by 47% (p=0.046) after adjusting for age and severity of stroke at admission (Table 3). Vertebral artery A/H was not related to stroke outcome.

Table 3.

Logistic regression models predicting stroke outcome.

Variable	β	S.E	Significance	OR	95% CI
age	−0.05	0.015	0.001	0.96	0.927 – 0.982
NIHSS at admission	−0.23	0.04	<0.001	0.79	0.734 – 0.853
Incomplete CoW	−0.64	0.32	0.046	0.53	0.281 – 0.989

Discussion

The reported prevalence of CoW variants may vary greatly depending on assessment methods.^{4, 22, 23} In this study, 34% of patients had an incomplete CoW, a lower rate than reported in many other studies. However, this discrepancy is expected given exclusion of the anterior and posterior communicating arteries from our analysis, with prior studies suggesting that posterior communicating artery A/H may be the most frequent CoW variant.^23,24

This study demonstrated a significant correlation between CoW completeness and stroke prognosis, with an incomplete CoW being associated with significantly lower odds of good short-term outcomes, even after adjusting for two of the most important predictors of stroke outcome: age and stroke severity at admission. The clinical relevance of CoW variants in cerebrovascular disease is not surprising, as variations in CoW anatomy have been shown to significantly affect cerebral arterial flow rates, flow distributions, and other cerebral perfusion parameters.^14,25–28

Prior studies have also correlated an incomplete CoW with increased risk of intraoperative ischemia in patients undergoing endarterectomy,^16–19 along with an increased risk of stroke in non-operative settings, particularly for patients with atherosclerotic disease.^20,29–33 Nonetheless, there is a relative dearth of literature regarding the effect of an incomplete CoW on prognosis following a stroke, with the available studies often demonstrating incongruent results.

Chuang et al found that an incomplete CoW was associated with an increased risk of hemorrhage following tissue plasminogen activator administration in acute stroke patients, and both Chuang and Zhou found that acute stroke patients with a complete CoW were more likely to have early improvement in NIHSS and good functional outcome at 3 months.^34,35 On the other hand, there are also studies demonstrating no significant correlation between CoW variations and stroke outcomes, with Seifert-Held et al finding no association between an incomplete CoW and functional stroke outcome after endovascular therapy.³⁶ De Caro et al also found no significant differences between admission and discharge NIHSS, in-hospital mortality, or hemorrhagic transformation between stroke patients with and without CoW variants.²¹

The results of this study lend further support to the hypothesis that CoW completeness has an influence on stroke prognosis. The discrepancy between our study results and those of prior aforementioned studies may largely be attributable to methodology. For example, Seifert-Held et al focused only upon stroke patients undergoing endovascular therapy, thus having less generalizability.³⁶ Seifert-Held et al also acknowledged that the effect size in their study suggests that a significant association between an incomplete CoW and stroke outcomes at 90 days may be demonstrated in larger cohorts. De Caro et al also used different methodology; most notably, their study did not control for initial stroke severity.²¹ De Caro et al also defined stroke outcomes in a different manner than in this study (using NIHSS at time of discharge – information that was not recorded for much of our sample).

Among the few prior studies demonstrating an association between an incomplete CoW and poor stroke prognosis, the study with methodology most similar to ours was that of Zhou et al.³⁵ However, an important distinction is that in their patient sample, Zhou et al reported more severe neurological deficits at admission among their patient sample with an incomplete CoW when compared with patients with a complete CoW; since initial stroke severity is one of the strongest predictors of outcome, it stands to reason that initial stroke severity in the patient sample for their study may have been a confounding variable. In this study, we demonstrated differences in stroke outcome between groups with complete and incomplete CoW despite controlling for initial NIHSS. Therefore, based on the results in this study, one can infer that CoW completeness affects outcome independently of initial neurological deficit, possibly also in the phase of recovery. We also did not exclude subjects with cardiac embolism as opposed to Zhou et al, who studied non-cardiac strokes.

There were several limitations to this study. The most notable of these limitations is our exclusion of the anterior and posterior communicating arteries from analysis. The posterior communicating artery has been reported as the most frequent location for CoW variants.^23,24 The posterior communicating artery may also arguably be a more clinically significant route of collateral flow given that it is usually the sole connection between the anterior and posterior intracranial arterial circulations. However, given the propensity for slower or more turbulent flow in this vessel which compromises assessment on time-of-flight imaging techniques, we did not feel that a reliable assessment of the posterior communicating arteries could be performed, and exclusion of the posterior communicating arteries from analysis was necessary to compensate for inherent limitations in assessment of cerebral vasculature on MRA examination. The anterior communicating artery was also excluded for similar reasons, with prior studies showing that this vessel is not reliably detected on 1.5 T time-of-flight MRA imaging.³⁷ While many patients in our stroke registry had CT angiography or digital subtraction angiography images available, focusing on these patients was felt to introduce a greater degree of selection bias as these imaging techniques tend to be reserved for more specific subsets of stroke patients at our institution. Anatomic variations of the aortic arch and other arteries outside the CoW also were not fully considered in this assessment.

Conclusion

This study suggests that an incomplete CoW may be associated with a poorer prognosis for stroke patients. Future studies are warranted to validate our data in both prospective and retrospective manners. It is essential to further establish whether the relationship between CoW incompleteness and outcome is limited to (or amplified in) a subgroup of patients with certain clinical characteristic. Should our results be replicated, they can potentially lead to adjustments in stroke management such as more aggressive monitoring or treatment of subjects with incomplete CoW.

Abbreviations

BP

Blood pressure

CoW

Circle of Willis

A/H

Aplasia/hypoplasia

MRA

Magnetic resonance angiography

NIHSS

National Institute of Health Stroke Scale

TOAST

Trial of ORG 10172 in Acute Stroke Treatment