1 INTRODUCTION

Relatively little attention has been paid to the prevalence and significance of a fetal configuration of the Circle of Willis in the context of ischemic stroke. In this normal variant of the primary collateral arterial network in the brain, the posterior cerebral artery (PCA) arises from the internal carotid artery (ICA) and the remnant PCA either completely lacks luminal connection to the basilar artery (BA), or its most proximal segment (P1) is hypoplastic and of smaller caliber than the ipsilateral posterior communicating artery (PcomA).1 The former variant is referred to as a complete FTP (cFTP) and the latter as a partial FTP (pFTP).2 An awareness of the prevalence of this variant of the arterial network is important in neurovascular and neurosurgical procedures,3 and also in the management of ischemic stroke, both in the early acute phase4 and when determining stroke etiology for adequate treatment and secondary prevention.5, 6 The prevalence of FTP in published reports varies widely (7%–46%) and may depend on the definition of FTP as well as method of investigation.1, 5, 7-10 The question of an increased risk of ischemic stroke—generally or in specific brain locations—in the presence of FTP has not been settled.11 A possible relationship between ischemic stroke topology and the presence of FTP has been investigated, but studies are scarce and results conflicting.5, 7, 12, 13 Hence, there is a need for studies in large samples of well-defined ischemic stroke patients. We determined the prevalence and laterality of FTP in a multicenter sample of patients with verified ischemic lesions on MRI-DWI, and evaluated if PCA territory involvement is related to the presence of ipsilateral FTP in the affected hemisphere. To investigate if FTP is more common in ischemic stroke populations than in other populations, we compared FTP prevalence and laterality in the ischemic stroke population to a mixed referral population investigated with computed tomography angiography (CTA) on any indication in a tertiary hospital setting.

2 METHOD AND MATERIALS 2.1 Ischemic stroke population

We reviewed all ischemic stroke cases in the neuroimaging repository of the MRI-GENetics Interface Exploration (MRI-GENIE) collaboration.14 At the time of review, MRI-GENIE contained image data of 3301 ischemic stroke patients. Data were contributed by 12 of the original National Institute of Neurological Disorders and Stroke Genetics Network (NINDS–SiGN) study sites. Information on ischemic stroke subtype according to the Causative Classification of Stroke (CCS)15 system was available for all patients in MRI-GENIE. Previous publications provide detailed descriptions of the NINDS-SiGN and MRI-GENIE studies, as well as collection periods and inclusion criteria for each site.14, 16

2.2 MRI/MRA evaluation of a multicenter ischemic stroke population

MRI scans were transferred to the MRI-GENIE neuroimaging repository in DICOM format and were available for review through a secure XNAT viewer. Two senior neuroradiologists (J.W., M.D.), who were blinded to the original evaluation at the enrollment site, centrally evaluated all MRI and MRA images regarding the presence of DWI lesions, vascular territory and location, and vessel anatomy and patency. Reasons for exclusion in the evaluation phase were poor image quality and lack of DWI sequences. We defined lesion(s) in the territory of the posterior cerebral artery (PCA) and its penetrating arteries, and in the territories of the vertebral (VA) or basilar arteries (BA) as posterior circulation ischemic stroke (PCiS). Lesions in the middle cerebral artery (MCA), anterior cerebral artery (ACA), or anterior choroidal artery (ACoA) vascular territories were defined as anterior circulation ischemic stroke (ACiS). The two neuroradiologists also recorded the side of lesion (left/right), cerebellar or brainstem location in infratentorial lesions, cortical/subcortical location for supratentorial lesions, and whether lesions were single or multiple. They also assessed vessel anatomy, patency, and evidence of vessel occlusion of the major intracerebral arteries related to lesion location.

2.3 Evaluation of CTA in a hospital-based population

Two neuroradiologists and one radiology resident blinded to the initial radiology report independently evaluated 546 CTAs of extra- and intracranial vessels, performed on any indication on referral from emergency or primary care physicians over a 12-month period in 2016 and 2017 at a tertiary teaching hospital, Skåne University Hospital, Lund, Sweden. Fifty-seven cases were excluded due to suboptimal image quality such as presence of artefacts making accurate assessment of target vessels difficult. A total of 489 cases were remained for evaluation of vessel anatomy and patency.

2.4 Definition of FTP

We applied the definition proposed by van Raamt et al.2 by which a complete FTP (cFTP) lacks any luminal connection with the basilar artery, that is, there is no visible P1 segment on MRA/CTA and a partial FTP (pFTP) in which the P1 segment is of smaller caliber than the PcomA (Figure 1). Vessel calibers were visually gauged by the neuroradiologists.

Adult and fetal configurations of the circle of Willis. (A) Adult configuration with visible PcomA of smaller caliber than the P1 segment of PCA. (B) Right partial fetal configuration with a hypoplastic P1 segment of the ipsilateral PCA and a larger caliber PcomA. (C) Right complete fetal configuration with no luminal connection with the basilar artery (absent P1 segment)

2.5 Patient selection

In the analysis of infarct location in relation to the presence of FTP, we excluded patients with bilateral supratentorial stroke (n = 125) and infratentorial stroke only (n = 201). We then excluded patients with bilateral FTP (n = 110) leaving 971 patients for analysis.

2.6 Statistical analyses

Continuous variables were compared between groups using the independent samples t-test. Categorical variables were compared between groups using the Pearson chi-square test. For comparison of FTP lateralization, we performed a binomial test to evaluate if the proportion of FTP on either side was significantly different from 0.5. All analyses were performed using SPSS version 26.0.

3 RESULTS

The multicenter MRI-GENIE image repository contained 3301 ischemic stroke patients. There were MRAs available for 1596 patients. After exclusion of patients investigated with MRA but with no visible DWI lesion on MRI, 1407 patients remained in the analysis. There were 552 (39%) women and 855 (61%) men. Patient characteristics and FTP configurations in the ischemic stroke group are presented in Table 1. Prevalence of FTP configurations in the two study populations is presented in Table 2.

TABLE 1. Patient characteristics and FTP prevalence in ischemic stroke patients (MRA)

No FTP

n = 964

Unilateral pFTP

n = 222

Unilateral cFTP

n = 99

Bilateral FTP

n = 122

Age, median (IQR) 64 (52–74) 64 (52–76) 69 (58–79) 64 (49–77) Male (%) 595 (62) 131 (59) 67 (68) 62 (51) Female (%) 369 (38) 91 (41) 32 (32) 60 (49) Hypertension (%) 644 (67) 152 (69) 68 (70) 79 (65) Diabetes (%) 235 (24) 48 (22) 25 (26) 20 (17) Atrial fibrillation (%) 132 (14) 41 (18) 16 (16) 13 (11) CCS subtype (%) CE 146 (15) 40 (18) 20 (20) 21 (17) LAA 246 (26) 58 (26) 20 (20) 24 (20) SVO 121 (12) 26 (12) 14 (14) 21 (17) Undetermined 368 (38) 82 (37) 40 (40) 44 (36) Other 83 (8) 16 (7) 5 (5) 12 (10) Abbreviations: CCS, causative classification of stroke; CE, cardioembolic; cFTP, complete FTP; FTP, fetal type posterior cerebral artery; IQR, interquartile range; LAA, large artery atherosclerosis; pFTP, partial FTP; SVO, small vessel occlusion. TABLE 2. FTP prevalence on MRA in ischemic stroke patients and on CTA in an unselected hospital-based population FTP variants MRA CTA n = 1407 (%) n = 489 (%) None 964 (68) 333 (68) Any FTP 443 (31) 156 (32) Bilateral 122 (9) 40 (8) Unilateral 321 (23) 116 (24) pFTPa 222 (16) 73 (15) cFTPa 99 (7) 43 (9) Right-sideda 206 (15) 74 (15) pFTP 146 (10) 49 (10) cFTP 60 (4) 25 (5) Left-sideda 115 (8) 42 (8) pFTP 76 (5) 24 (5) cFTP 39 (3) 18 (4) Abbreviations: cFTP, complete FTP; CTA, computed tomography angiography; FTP, fetal type posterior cerebral artery; MRA, magnetic resonance angiography; pFTP, partial FTP. a Bilateral cases not included. 3.1 FTP prevalence and laterality on MRA in ischemic stroke patients

In 1407 acute ischemic stroke patients with DWI lesions, we detected any FTP configuration (partial/complete/uni-/bilateral) of the circle of Willis in 443 (31%) patients. There were 122 patients with bilateral FTP, and among these, we detected complete FTP in both hemispheres in 19 (16%). The most common bilateral configuration was partial FTP on both sides (64%). The most common such configuration was partial FTP on both sides, detected in 78 (64%) of the bilateral cases. All bilateral combinations are presented in Table S1.

3.1.1 Partial and complete FTP

Partial (pFTP) configurations with a visible P1 segment of smaller caliber than the ipsilateral PcomA were more prevalent than cFTP. There were 403 partial FTP vessels and 162 complete FTP vessels including those observed on either or both sides of a bilateral configuration. A unilateral pFTP was present in 222 (16%) of the patients. A unilateral cFTP was detected in 99 (7%) patients.

3.1.2 Laterality of FTP

In the IS population, the configuration was bilateral in 122 (9%) patients and unilateral in 321 (23%) patients. Bilateral FTP was significantly more common in women than in men (p = 0.03, OR1.6, 95%CI 1.02–2.37) (Figure 2).

Prevalence of FTP on MRA in ischemic stroke patients and on CTA in unselected patients (left), and on MRA in women versus men with ischemic stroke (right)

In the ischemic stroke population overall (n = 1407), the prevalence of unilateral right-sided FTP was 15% versus 8% unilateral left-sided FTP. (Table 2). A unilateral FTP was found on the right side in 206/321 and on the left side in 115/321 patients of the FTP patients. The difference between the proportion of FTPs on the right versus left side was significant (p < .001) (Table 3). The number of pFTP/cFTP on each side is presented in Figure 3.

TABLE 3. Proportion of unilateral right/left side FTP among ischemic stroke patients (examined with MRA) and unselected patients (examined with CTA) Right FTP Left FTP Proportion right versus left FTPa p-valueb MRA 206 115 0.64 (95%CI 0.59–0.70) <.001 CTA 74 42 0.64 (95%CI 0.54–0.72) <.005 Subjects with bilateral FTP not included. Abbreviations: CTA, computed tomography angiography; FTP, fetal type posterior cerebral artery; MRA, magnetic resonance angiography. a Clopper-Pearson confidence interval (CI). b Binominal test of proportions.

Right versus left partial and complete FTP in ischemic stroke patients (MRA) and unselected patients (CTA)

3.1.3 Ischemic stroke etiology

The prevalence of FTP was highest in patients with cardioembolic stroke (36%), and lowest in patients assigned the CCS subtypes large artery atherosclerosis (29%) and Other (28%). The difference in prevalence between etiologies overall was not significant (p = .52). The difference in prevalence between large artery atherosclerosis and cardioembolic stroke was not significant (p = 0.33).

3.2 FTP prevalence and laterality on CTA in an unselected hospital population

We also investigated the prevalence and laterality of FTP in a mixed referral hospital-based population without an ischemic stroke diagnosis. After exclusion of low-quality examinations, we evaluated 489 patients investigated with CTA on any referral indication. We found any fetal configuration (partial/complete/uni-/bilateral) of the circle of Willis in 156 patients (32%). There were 40 patients with bilateral FTP, and among these, we detected complete FTP in both hemispheres in 9 (22%). The most common such configuration was partial FTP on both sides, detected in 20 (50%) of the bilateral cases. All bilateral combinations are presented in Table S1.

3.2.1 Partial and complete FTP

Partial (pFTP) configurations were more prevalent than cFTP (16% vs. 9%) and were unilateral in 73 (15%) of the patients. A unilateral pFTP was found on the right side in 49 (10%) patients and on the left side in 24 (5%) patients. Unilateral cFTP was present in 43 (9%) patients.

3.2.2 Laterality of FTP

The configuration was bilateral in 40 (8%) patients, and bilateral FTP was proportionally more common in women than in men (11% vs. 7%) (Figure 2). Unilateral right-sided FTP was present in 74 (15%) and unilateral left-sided FTP in 42 (8%) of all patients. In the unselected population overall (n = 489), the prevalence of unilateral right-sided FTP was 15% versus 8% unilateral left-sided FTP. (Table 2). The proportion of right versus left FTP differed significantly (p < .005) (Table 3). The number of pFTP/cFTP per side is presented in Figure 3.

3.3 Infarct location

In the analysis of infarct location in relation to the presence of FTP, we excluded patients with bilateral supratentorial stroke (n = 125) or infratentorial stroke only (n = 201). We then excluded remaining patients with bilateral FTP (n = 110) leaving 971 patients for analysis. Among these, there were 464 right hemisphere strokes and 507 left hemisphere strokes. The proportion of patients with ipsilateral FTP to their stroke was 71/464 in the right versus 51/507 in the left hemisphere. The proportion of patients with ipsilateral FTP and PCA territory involvement was 5/71 in the right hemisphere and 10/51 in the left hemisphere. The presence of ipsilateral FTP to stroke side was not associated with PCA territory involvement (p = 0.89, [OR 0.96, 95%CI 0.54–1.71]).

4 DISCUSSION

FTP is a prevalent finding in both ischemic stroke patients and in unselected hospital populations.2, 7, 12 We observed a preferential lateralization of FTP to the right side with a right-sided FTP almost twice as common in both the ischemic stroke group and in the unselected hospital-based group. Other studies have reported similar findings,7, 12 but the difference was not specifically addressed. Notably, another congenital variant of the cerebral vascular tree—vertebral artery hypoplasia—has also been found to occur more frequently on the right side in both healthy subjects and ischemic stroke patients.17-19 A recent study of 923 healthy subjects found that vertebral artery hypoplasia was associated with the presence of FTP, and that the proportion of patients with FTP increased with the degree of vertebral artery hypoplasia.10 The same study found that FTP was more often ipsilateral to the hypoplastic vertebral artery. The lateralization to the right of vertebral artery hypoplasia has been attributed to arterial flow conditions under which the left vertebral artery assumes dominance due to its origin directly from the left subclavian artery causing higher shear stress.20 In the presence of a hypoplastic vertebral artery during fetal development, a persisting FTP rather than the “normal” regression to an adult configuration, may represent a compensatory mechanism to ensure adequate perfusion to the posterior brain.10 Although speculative, the higher proportion of right-sided FTP observed in this and other studies could be related to the same developmental conditions that lead to lateralization of vertebral artery hypoplasia.

In our study, the overall prevalence of any FTP in the ischemic stroke group was similar among men and women, but bilateral FTP was significantly more common in women than in men among patients with FTP. In the unselected CTA group women also had a higher proportion of bilateral FTPs than men (10% vs. 6%), but in this smaller sample, the finding did not reach statistical significance. Few studies report sex-specific prevalence data for FTP, but one other previous study reported that complete FTP was more common among women.12 We have been unable to find any biological explanation in the literature for the difference between men and women regarding FTP configurations. The influence on stroke risk and stroke location by the presence of FTP in the absence of other vascular risk factor is still not settled. Results from other studies are conflicting and difficult to compare due to heterogeneity between studies.5, 7, 13, 21 In our analyses of infarct location,5, 13 we did not find evidence that PCA territory involvement is more common in patients with FTP in the stroke-affected hemisphere versus patients without FTP in the same hemisphere. In certain settings, such as steno-occlusive disease of the ipsilateral internal carotid artery, it is clear that an FTP-supported PCA territory is at risk of ischemic injury in addition to the MCA territory.1, 11, 22 The PCA territory may also be the only arterial territory involved in an anterior circulation compromise. In the setting of atherosclerotic and occlusive ICA disease, ischemic injury may affect the ipsilateral PCA territory, either through failing anastomotic perfusion or artery to artery embolism via FTP.23 We were not able to evaluate the status of the extracranial portion of the carotid system in this study.

Finding an FTP variant in almost one third of ischemic stroke patients carries implications for stroke management. For example, patients with high-grade stenosis of the ICA, in combination with ipsilateral FTP and ischemic stroke or TIA in the PCA territory should be considered for carotid endarterectomy as a secondary prevention measure.1, 24 Such individuals are also at risk of developing hemispheric infarctions involving both ipsilateral MCA and PCA territories because of their common vascular supply by the ICA.11, 13 In addition, leptomeningeal collaterals may fail to develop in the absence of connections (cFTP) between the vertebrobasilar and carotid system,2 putting a larger brain volume at risk of hypoperfusion and ischemic injury in the presence of occlusive arterial disease. In acute stroke imaging, the presence of FTP may lead to perfusion patterns with left-right asymmetry in the posterior territory, mimicking cerebrovascular compromise. Knowledge of cerebral perfusion patterns associated with the presence of FTP is important to avoid a misdiagnosis of ischemic stroke.4, 25

4.1 Limitations

This study has several limitations. Compared to our ischemic stroke populations, the individuals investigated with CTA may not be representative of a “non-stroke” control group since their investigation with CTA in a number of cases likely was associated with hospital visits or referrals brought about by symptoms that were assessed as potentially vascular in nature. We did not access medical records to ascertain final diagnosis in the CTA group. However, CTA is commonly ordered on a wide range of indications and our sample may be seen as representative of a much broader segment of the population than confirmed acute ischemic stroke patients. Another potential limitation is the difference in the proportion of sex representation between the groups (MRI: male 61%, female 39%; CTA: male 50%, female 50%). However, both the overall proportionate prevalence of FTP configurations and the sex-specific prevalence were very similar in the two groups, making it unlikely that the slight difference in sex representation influenced the results. In the ischemic stroke group, patients with large infarctions involving both MCA and PCA territories may have been selected not to undergo MRI investigation leading to selection bias and under-representation of this patient group in our study.

The MRA investigations included in this study were intracranial only. For this reason, we were not able to evaluate extracranial portions of the vertebral arteries or explore the presence and laterality of VAH, nor its possible association with specific FTP configurations.

Finally, we excluded cases of bilateral FTP from the analysis of PCA territory involvement in relation to the presence/absence of FTP. Bilateral FTP occurred in 9% of ischemic stroke patients, but accounted for 28% of FTP cases (122/443), which may have influenced the results. However, we consider this unlikely given that the proportion of PCA territory involvement in bilateral FTP was lower than in unilateral FTP patients (8% vs. 14%). We further suggest that the presence of bilateral FTP may effect cerebral hemodynamics and collateral compensation in ways not well studied, thus influencing the risk of ischemic stroke, stroke evolution, and location in a different way than unilateral FTP, warranting future studies in which this group is investigated separately.

4.2 Strengths

The strengths of this study include the large sample of patients with MRI verified ischemic stroke diagnosis and the detailed information on stroke location available. We made very similar findings regarding the prevalence of different FTP configurations and laterality in two different populations with different modalities (MRA and CTA), which strengthens the reliability of the results. The study groups were comparable in age. Median age in the hospital-based group (CTA) was slightly higher than in the ischemic stroke group (69, IQR 53–75 vs. 64, IQR 52–75), in our view limiting the risk that FTP prevalence in the unselected hospital group represents that of a disproportionate number of future stroke patients. However, the study is not longitudinal, and we therefore do not know if some patients in the CTA group later developed ischemic stroke.

We believe that any potential differences in detection rate between the modalities due to issues such as flow conditions influencing MRA findings are unlikely to have introduced systematic errors that would significantly alter the study results.

5 CONCLUSIONS

The prevalence of FTP on MRA is similar in patients with ischemic stroke and unselected patients investigated with CTA, which does not support the idea that FTP per se influences the risk of stroke. There is a preferential lateralization of FTP to the right hemisphere, a finding that needs to be confirmed in prevalence studies of both VA and FTP in healthy individuals. Our study does not provide evidence that FTP alone is a risk factor for ipsilateral PCA involvement compared to patients without FTP. An awareness of the high prevalence of FTP is nevertheless of importance in the acute phase of ischemic stroke evolution, in the workup of ischemic stroke subtype assignment and decision-making regarding secondary prevention measures.

ACKNOWLEDGMENTS

We gratefully acknowledge all patients who participated in the study. The authors would like to acknowledge Andrea Dahl Sturedahl, Clinical studies Sweden-Forum South, for her assistance with statistical analyses.

CONFLICT OF INTEREST

All authors have declared no conflict of interest.

AUTHOR CONTRIBUTIONS

PF was involved in the study design, interpretation of data, and drafted the manuscript. JW, MD, and OW in the acquisition, analysis, and interpretation of radiology data. JP, NSR, and AL in the study design, interpretation of data, and revision of the manuscript.

ETHICS CONSIDERATIONS

This study was conducted in accordance with the Helsinki Declaration as revised in 2013. The MRI-GENIE study has been approved by the institutional review board at Massachusetts’ General Hospital, Boston, MA; MRI-GENIE: IRB #2001P001186. All participants provided informed consent either directly or through surrogate authorization at the time of enrollment at the original sites. The CTA study was approved by the Regional Ethical Review Board, Lund University, Sweden: # 2018/411.