Anatomic Completeness, Variations, Patency, and Functional Assessment of Circle of Willis: Implications for Chronic Aortic Dissection and Non-Emergent Arch Surgery

Abstract

Circle of Willis (CoW) variations are present in a significant number of patients. In patients undergoing non-emergent surgery for aortic dissection (e.g., chronic dissection or planned complex aortic arch surgery), an extensive preoperative examination is warranted as anatomic completeness, variation, patency, and functional assessment of CoW is crucial to optimize neuroprotection strategies. In non-emergent clinical presentations, unilateral or bilateral selective cerebral perfusion should be tailored to patient’s anatomy to minimize intraoperative risk and improve outcomes. A new classification system provides a simple and comprehensive ontology of the described anatomical variations of the CoW. An accurate preoperative assessment might indicate the need for adjunctive perfusion in selected CoW configurations.

1. Introduction

Circle of Willis (CoW) or cerebral arterial circle (CAC) is an arterial anastomotic network formed by a symmetrical polygonal-shaped connection between internal carotid arteries (ICAs) and vertebral arteries (VAs). It is located in the subarachnoid space on the base of the brain, around the optic chiasm and infundibulum of the pituitary stalk in the suprasellar cistern.

The anastomotic circle may be divided into 2 parts:

• Anterior circulation: formed by left and right internal carotid arteries (ICAs), horizontal segments (A1s) of the left and the right anterior cerebral arteries (ACAs), and a single anterior communicating artery (AcomA)

• Posterior circulation: consisting of left and right posterior communicating arteries (PcomAs), the horizontal segments (P1s) of left and right posterior cerebral arteries (PCAs), and a single basilar artery formed by the union of right and left VAs.

This anastomotic network allows an equal vascular supply between the two cerebral hemispheres and provides adequate perfusion in case of impaired or decreased blood flow through one or more of its segment arteries. However, other functions may be attributed to CoW, as demonstrated by Burlakoti et al. [1] through the measurement of cross-sectional area of incoming, communicating, and outgoing arteries of 51 cadaveric brains. The significant differences in the cross-sectional area of incoming and outgoing arteries, together with the area of communicating arteries, may be relevant for lowering the peak pressures of arterial blood perfusion of the brain, reducing hemodynamic stress, and proving a better distribution of pressure waves.

Four criteria are established to define the “normal” (non-variant) anatomy of the CoW [2]:

• All segments are present (AComA, A1s, PComAs, and P1s);

• All segments arise from their natural origins;

• No accessory arteries are present;

• All segments have an external diameter of >1 mm.

It is estimated that 68.22% of individuals in the neurologically healthy population have CoW variations, more frequent than the considered non-variant configuration. Commonly recorded variation types include hypoplasia, absence, and duplication. Hypoplasia is usually defined as a diameter <1 mm. Variations are a risk factor for cerebrovascular diseases, above all variations of CoW increase the risk of stroke, as analyzed by De Caro et al. [3] in a study where the prevalence of CoW variants increased in patients with ischemic stroke when compared to those who have other illnesses or symptoms besides stroke, due to a less efficient collateral flow. Another important cerebrovascular disease involved is arterial aneurysms, which may develop from the altered hemodynamics that characterize variations of CoW, as a result of high and low wall shear stress around the bifurcation sites [1]. Larger aneurysms have an increased risk of rupture, leading to subarachnoid hemorrhage and hemorrhagic ischemia. Variations of CoW are also involved in therapeutic options in preoperative assessment for vascular and heart surgery to choose the most appropriate method for cerebral perfusion and protection.

The purpose of this chapter is to give an overview of how variations influence the selection of cerebral perfusion techniques in aortic arch surgery.

2. Classification of variations

The considerable heterogeneity of variants of CoW leads to the need to create an anatomical and functional classification system, due to the relevance of variants in the preoperative assessment to decide the best surgical approach and the requirement of precautions. The study of CoW variants and adequate cerebral protection techniques reduce surgical errors and postoperative complications, improving the clinical outcome of patients after surgery. The aim is to prevent neurological sequelae after arch surgery during total or partial arch replacement through the classification of variations in risk categories which may help to use appropriate techniques of cerebral perfusion.

The elaborate embryological evolution of CoW and the shifting configurations that precede its definitive form predispose to abnormal development that may influence its capacity as a shunt mechanism. In 1961, Riggs and Rupp [4] classified the anomalous formation of CoW in four groups, based on the hypoplasia of communicating arteries (AcomAs and PcomAs), divisional branches of basilar artery (P1s), proximal anterior cerebral arteries (A1s) and both divisional branches of basilar and anterior cerebral. Although it is an admirable work where anatomical description encounters functional consequence, they described 21 anomalous configurations only considering hypoplastic segments, according to the configuration accepted as typical by Padget, without including absence or duplication of the cerebral arteries segments.

Later, in 1979, Lazorthes [5] described 22 different patterns of CoW, where type I is the typical CoW according to Padget’s definition, and type II is the “fetal type”, characterized by hypoplasia of all arterial segments. With regards to other configurations, they are based on the dependence or independence of the three main arterial axes (carotid arteries and basilar artery), but only segment hypoplasia is included and the lack of anatomical or functional distribution of variants in groups can generate confusion and misunderstanding about the functional consequence and the main feature of each variation. Moreover, the illustrations of CoW patterns are unclear and not exactly descriptive of hypoplasia.

In 1998, Krabbe-Hartkamp [6] outlined 18 variants using three-dimensional time-of-flight magnetic resonance angiography, dividing them into 2 groups based on the anatomical position, whether the hypoplasia or absence of segment is in the anterior or posterior circle. Hypoplasia was defined as a diameter smaller than 0.8 mm, therefore it may underestimate the frequency and the number of variations.

Afterward, Merkkola et al. [7] studied casts and angiographies made during forensic medicine autopsies to assess in theory weather perfusion through the right axillary artery is sufficient in providing uniform distribution of blood to both hemispheres of the brain (Table 1). They defined hypoplasia as a diameter below 0.5 mm but also repeated the analysis using 1 mm as threshold. In addition, they considered that critical arteries to examine are the AcomA and the left PcomA, because AcomA transports blood through the anterior circulation of CoW to the left cerebral hemisphere and PcomA establishes communication between the anterior and the posterior circulation. Then their absence is associated with hypoperfusion of the left hemisphere of the brain. They defined communicating arteries as normal, compromised, or missing depending on the diameter of the vessel, if >0.5 mm, <0.5 mm, or not present in either the permanent cast or the angiography. Then, they created four groups: the first one was referred to as excellent because both communicating arteries were normal; the second group called “good” was characterized by only one normal communicating artery; the patients who belonged to the compromised group, the third one, had both arteries compromised or one compromised and the other one missing; the last group was characterized by the absence of both the communicating arteries. This classification is limited because only considers two segments of CoW and does not look at vertebral arteries, right A1 and left P1 abnormalities. The proposed classification included in the same group a missing and a hypoplastic segment, which may generate confusion and may not represent an accurate analysis of the hypoperfusion risk of unilateral cerebral perfusion technique.

Study (year)	Patients	CoW variants, details, and outcomes
Merkkola et al. [7]	87	Exellent: AcomA+AcomPsn N 35%; Good: AcomA N+ AcomPsnM 32%, AcomA M+ AcomPsn N 8%, AcomA C + AcomPsn N 8%; Compromised: AcomA m + AcomPsn C 1%, AcomA C + AcomPsn M 2%; Bad: AcomA M+ AcomPsn M 14%
Papantchev et al. [8]	99	CoW type Ia Left PCoA or both PCoA Left MCA 27.3% (27/99) 26.6% (22/83); CoW type Ib AcoA Left ACA 0% (0/99) 2.4% (2/83); CoW type IIa ACoA and left PCoA Left ACA; left MCA 3.1% (3/99) 3.6% (3/83); CoW type IIb Right VA or left PCA (P1 segment) Left MCA; left PCA 9.1% (9/99) 7.2% (6/83); CoW type III Right ACA (A1 segment) Right ACA; left ACA; left MCA 3.1% (3/99) 7.2% (6/83); CoW type IV Right VA and right ACA (A1 segment) Right ACA; left ACA; left MCA; left PCA 0% (0/99) 1.2% (1/83); Total 42.4% (42/99) 48.2% (40/83)
Papantchev et al. [9]	500	Type IA Left PComA Left middle cerebral artery 35.6% (178/500); Type IB AComA Left anterior cerebral artery 2% (10/500); Type IIA Left PComA + AComA Left anterior cerebral artery + left middle cerebral artery 4.8% (24/500); Type IIB Left P1 or right VA Left middle cerebral artery + left posterior cerebral artery 9.2% (46/500); Type IIIA Right A1 + right anterior cerebral artery + left anterior cerebral artery + left middle cerebral artery 6% (30/500); Type IIIB AcomA + right VA Left anterior cerebral artery + left middle cerebral artery + left posterior cerebral artery 0.2% (1/500); Type IV Right A1 + right VA or right A1 + left P1 Right anterior cerebral artery + left anterior cerebral artery + left MCA + left PCA 0.8% (4/500); Total: 58.6% (293/500)

Table 1.

Studies describing anatomic variances of CoW during autopsies after sudden cardiac/cerebrovascular death.

As we discussed, the incompleteness of these classification systems brought Ayre et al. [2] to propose a comprehensive classification system through the distribution of 82 distinct anatomical variations into five groups, based on the type of abnormalities (hypoplastic or absent segment), which can provide a clear and complete vision about all the patterns we may detect in patients. In this classification, the absence of segments is included, both hypoplastic and absent segments, and the presence of accessory segments. This system allows us to clearly distinguish every potential variation, but it is only based on an anatomical examination without information about functional implications. In addition, to identify the specific variants which may be observed in patients and attribute them to the corresponding group, it is necessary to consult the illustrated table of Ayre’s classification. This approach may be impractical and time-consuming, especially in an emergent or urgent situation. Another disadvantage in using this classification as a reference for cerebral perfusion in cardiac surgery is the absence of vertebral artery abnormalities, not taken into consideration either by classification previously seen. However, it may be helpful and more convenient for the attribution to each group and subgroup of an ideal perfusion technique to warrant blood flow for each cerebral hemisphere, based on a primary empirical evaluation, to enhance the identification and categorization of CoW variants.

Coulier et al. [10] described the frequencies of variations of CoW, identifying 27 types of variations in 511 patients with Computed Tomographic Angiographic 3D Volume Rendering reconstruction of the cerebral arterial circle (CAC). They classify variants into 3 groups: complete CACs (22.5%) with seven robust segments (6.1%) or with hypoplastic segments, nearly complete CAC (30.72%) with only one segment absent and incomplete CAC. They found that the posterior arch (73.18%) is much more frequently incomplete than the anterior arch (18.4%), due to the lack of PComAs, unilaterally or bilaterally.

Above all, it is important to highlight variations characterized by some very precarious vascular supplies, such as complete isolation of the MCA (iMCA), found in 1.8% of patients and detected in seven different types of configurations. The MCA is isolated when it is not connected to the contralateral ICA through the anterior circulation for the absence of ipsilateral or contralateral A1 segment of ACA or AcomA and, at the same time, it is not connected to vertebral arteries through the posterior circulation, due to the absence of PcomA or ipsilateral P1 segment of PCA. The most common configuration is the absence of both communicating arteries, anterior and posterior, followed by the absence of ipsilateral A1 and PcomA. Banga et al. [11] found that iMCA is an independent predictor of immediate neurological events after CEA using cross-clamping technique without shunt protection. Fortunately, in 545 patients who underwent CEA, the iMCA was found in 6.3%, so it’s a rare CoW variation.

Other situations of anomalous CoW are complete separation of the 2 hemispheres (2.15%), complete separation of basilar and carotid territories (40.1%, frequently missing both PComAs), complete separation of the three main arterial axis of CAC (both ICAs and VB trunk, 8.6%, frequently missing all the communicating arteries).

As seen, the posterior arch is most involved in CoW variations, specifically for the PcomA-P1-P2 complex, involving a particular variant named “Fetal Posterior Cerebral Artery” (FPCA) in which the distal posterior carotid artery territory is dependent on the ipsilateral ICA. All the posterior arch variations can be divided into 5 groups:

• Agenesis of the PcomA (30.52% unilateral, 35% bilateral)

• Hypoplasia of the PcomA (25.24% unilateral, 19% bilateral)

• “Intermediate” type of FPCA, in which the caliber of P1 is quite similar to PcomA

• “Partial” type of FPCA, where the caliber of PcomA is greater than P1

• “Full” FPCA type, characterized by the absence of P1 and a robust PcomA.

Those variants may influence the extension of an ischemic stroke, hence the prognosis of patients, which can be worse or better than patients without variations. For instance, in patients with unilateral full FPCA type, a basilar artery occlusion may result in a smaller lesion, although an ICA obstruction may result in a wider ischemia compared to patients without CoW variation [3].

Rangus et al. [12] demonstrated that more than one-third of patients with CoW variations suffer ischemic stroke, especially in the posterior circulation rather than the anterior. A higher frequency of right-sided FPCA was registered among patients with CoW variants affected by stroke. All the patients affected by this condition were reclassified into having one territory lesion stroke pattern from an initial categorization as more than one territory pattern after performing MRI examinations. Initially, vertebrobasilar territory stroke was re-classified as ICA territory stroke, re-assigning the stroke from the posterior to the anterior territory. These patients are more prone to misclassification of stroke patterns, which could affect care, not only in the acute phase but may also compromise the choosing of the appropriate secondary prevention protocol.

It appears clear that some classifications do not include types of variations that may be fundamental in the preoperative evaluation, such as iMCA and the absence of vertebral arteries, probably due to the different aims and methods of each classification, as well as the different types of population involved and different threshold to define hypoplasia. For example, the Krabbe-Hartkamp classification is functional in understanding the role of CoW variations in patients with various degrees of carotid stenosis, and since it involved healthy subjects, it is focused on the diameter variations of segments. On the other hand, the Riggs and Rupp classification was obtained by patients with clinical signs of neurological dysfunction and it is centered on the hypoplasia of the segments without even a mention of segment absence, pursuing the aim to reconstruct the embryological derivation and subsequently the defects. It may underestimate the presence of CoW variants in healthy population.

Another difference is the type of method utilized to study CoW variations, such as imaging techniques (CTA, MRA) or anatomic dissection at autopsy. Each method has a different sensibility and specificity, so it may lead to miss variations that play an important role in cardiac and vascular surgery.

Despite the numerous attempts to create a complete and comprehensive classification system, it is indispensable to study a new categorization of CoW variants that allows easier and faster identification of the ideal perfusion technique for each variation when aortic arch surgery is performed.

Therefore, the important incidence of variations of CoW and the multitude types of configurations, despite the different implications of each configuration in neurological outcomes after vascular and cardiac surgery, support the use of Computer Tomography Angiography (CTA). This examination, adopted as a complementary high-quality evaluation for the preoperative assessment of non-emergent pathologies, such as chronic aortic dissection, reduces postoperative neurological complications.

3. Techniques of cerebral protection and neuromonitoring

Aortic arch surgery performed with the open distal anastomosis technique requires a blood-free operating field and it may be ensured through a period of circulatory arrest (CA). This strategy requires the interruption of all body circulation, including cerebral perfusion, and several studies demonstrated that it is indispensable to provide an adequate cerebral protection technique during CA, to reduce the risk of neurological postoperative complications.

After cardiac surgery, neurological injuries may occur. Roach classifies them into two categories: Type 1, which includes transient ischemic attack (TIA), fatal and non-fatal stroke, stupor, coma and hypoxic encephalopathy; Type 2, which includes epilepsy without focal deficit, memory loss, cognitive decline and delirium. Later, Ergin introduced a new classification to distinguish the characteristics of neurological damage on the basis of clinical expression, describing Transient Neurologic Deficit (TND) and Permanent Neurologic Dysfunction (PND), a useful system to compare the outcome of each cerebral protection strategy. TND is defined as a global ischemic injury, which is the result of inadequate or interrupted cerebral protection but is considered a benign self-limiting condition with complete resolution before the discharge. It manifests clinically as delirium, agitation, obtundation, postoperative confusion, or transient Parkinson’s without evidence in cerebral imaging such as CT. PND is represented by ischemic stroke, due to an embolic event and it is not correlated with the strategy used for brain protection. MR shows cerebral lesions in the first 12–18 hours after the event, while CT after 24–48 hours. Symptoms and clinical manifestation depend on the site and the extension of the ischemic insult.

The deep hypothermic circulatory arrest (DHCA) was used in the 70s as a substitute for cardiopulmonary bypass (CPB) and, later, used for cerebral protection during arch surgery in aortic aneurism. The aim is to reduce metabolism and oxygen use for cerebral tissue. Deep or profound hypothermia is defined as a temperature below 20°C, which makes it possible to reduce the oxygen consumption of the human brain from 2.9 mL/g/min at 37°C to 0.2 mL/g/min at 20°C. A retrospective review [13] analyzed DHCA performed at a main temperature of 16.5°C versus DCHA with RCP at 20.5°C in patients who underwent aortic hemiarch reconstruction for elective aortic aneurysm or acute type A aortic dissection. This study demonstrated the operative mortality and postoperative stroke rate were comparable and there was no significant difference in long-term prognosis, in both elective and acute ascending aortic pathologies. CA time was shorter in the DCHA group (9 vs. 15 mins), according to the STS database study of type A dissection, which suggested that isolated DHCA less than 20 minutes might be safe during hemiarch replacement for aortic dissection. Unfortunately, it is difficult to predict the exact limit in terms of safe time to warrant cerebral protection and the temperature threshold to avoid deleterious systemic effects of deep hypothermia, although it is considered to be 18°C. A period >40 and >65 minutes of circulatory arrest is predictive of ischemic stroke and mortality respectively. Another disadvantage of DHCA is the prolonged CPB time, due to the longer period to cool down and to warm up the patient.

Englum et al. demonstrated the importance of combining deep hypothermia with a cerebral perfusion technique to reduce the incidence of postoperative stroke and neurological complications, without any differences in terms of operative mortality and neurological outcomes between retrograde and anterograde cerebral perfusion [14]. They also compared different grades of hypothermia and found no differences among various grades of hypothermia when a cerebral perfusion technique was included. The different grades of hypothermia experimented on in this study were defined and divided into three groups: high-moderate (24.1–28°C), low-moderate (20.1–24°C), and deep or profound (<20°C).

Ueda et al. described, for the first time, the use of retrograde cerebral perfusion (RCP) in aortic arch surgery. This technique consists of perfusing the brain through the superior vena cava with a retrograde blood flow. The main perfusion pressure is maintained between 20 and 25 mmHg but there is no possibility for direct monitoring. This type of perfusion may eliminate toxic metabolites and cerebral air and embolic debris from the arterial side, but the effectiveness remains uncertain. It may be disadvantageous for prolonged CPB time, as over 80 minutes, and if pressure is maintained over 30 mmHg because it may cause cerebral edema, worsening cerebral damage. Moreover, the veno-venous shunting of blood flow from the cerebral to the systemic venous circulation may occur, which could reduce the perfusion of brain capillary and, therefore, the efficacy and benefits of RCP in cerebral protection. However, meta-analysis demonstrates that RCP is associated with lower postoperative stroke and operative mortality compared with DHCA, with no difference in any outcome when comparing ASCP, even if DHCA is performed principally but not exclusively at a main temperature of 20°C. If executed at low-moderate temperature, RCP has no differences with ASCP in terms of neurological complications and postoperative stroke, so it allows for lowering the grade of hypothermia. Thus, this cerebral protection technique is least utilized when performing elephant trunk surgery, preferring bilateral or unilateral ASCP, probably due to the prolonged CPB time.

The anterograde selective cerebral perfusion (ASCP) remains the most used technique for cerebral protection. It was first used by De Bakey in 1957, and later, by Kazui in 1995 and Pacini in 2010, who demonstrated the efficacy in terms of postoperative morbidity and mortality. ASCP is based on cerebral perfusion directly through the epiaortic vessels and it may be unilateral or bilateral, in relation to the cannulation of one or both carotid arteries, even if the preferred cannulation site is the right axillary artery or both the right and left axillary artery in the case of bilateral ASCP to avoid manipulation of the supra-aortic vessels and the risk of atherosclerotic plaques dislocation. The main blood flow is regulated at 8–16 ml/Kg/min to maintain a radial blood pressure between 40 and 70 mmHg, with higher flows in warmer temperature strategies but below 16 mL/Kg/min in order to avoid an increased cerebral pressure. The core temperature is kept between 20 and 30°C, while perfusate temperature is kept between 20 and 26°C, and the left subclavian artery is clamped to avoid the subclavian steal syndrome. The use of this cerebral protection offers the advantage of prolonging the safe time of circulatory arrest, which is very useful when a more complex and longer surgery technique is performed. It also allows the use of bilateral perfusion if both anterior and posterior communicating arteries are absent, to reduce the risk of hypoperfusion of the left cerebral hemisphere. This procedure is not without its risks, such as embolic events and vascular injury of cannulated vessels.

Multiple factors contribute to the final neurological outcome after aortic arch surgery: temperature management, blood flow, pharmacologic protection, aortic cross-clamping, and duration of CA. The type of perfusion strategy adopted is the most deciding factor for the neurological outcome, as the cannulation site is for ASCP, providing the following options: the femoral artery, the right or left axillary artery, the carotid arteries, the innominate or subclavian artery and the ascending aorta in the true lumen. The right axillary artery is the preferred cannulation site, due to the lack of sandblast effect of turbulent flow, which may cause the disaggregation of atheromatous plaques and calcifications, with a reduced risk of cerebral embolization. Another advantage is the continuous backflow in all aortic arch vessels which minimizes the risk of cerebral air embolization. In contrast, the left axillary artery exploits posterior cerebral circle anterograde perfusion through the left vertebral artery. However, this may not ensure a complete cerebral perfusion in the case of CoW variants, as in the case of iMCA where there is no connection from the contralateral internal carotid artery and at the same time there is no connection from the vertebral arteries. In an emergent situation, when isolating the axillary artery is time-consuming, direct innominate artery (IA) cannulation may be used as an alternative, which is non-inferior to axillary artery with regard to neurological outcomes in elective aortic arch surgery, as demonstrated by a randomized trial comparing axillary and innominate artery as cannulation sites [15]. In addition, IA cannulation reduces procedural and CPB times, particularly in obese patients, and warrants lower flow resistance, decreasing blood loss in the operative field and/or kinking of the cannula. In case of shortness or dissection of IA, the right subclavian artery cannulation without infraclavicular incision may be used as a cannulation site. Left common carotid artery is a feasible and fast approach, used for an emergency option, mostly in obese patients, due to the high compliance of the vessel which allows high CPB flow. On the other hand, in case of small lumen, it will not support high CPB flows. Ascending aorta as a cannulation site is usually avoided due to the high risk of false lumen perfusion and rupture or dissection of the aortic wall while the femoral artery has long been the standard site for peripheral cannulation in non-elective cases permitting the cerebral blood flow through retrograde aortic perfusion, but a diseased abdominal and descending thoracic aorta increases the risk of cerebral emboli and malperfusion. The absence of obstructive clamps or cannulas within the field and direct aortic manipulation in femoral artery cannulation is advantageous for the surgical technique. However, the disadvantages of retrograde perfusion are the possible development of new re-entry tears and the perfusion of re-entry tears. In these circumstances, there is an important risk of epiaortic vessels and visceral malperfusion and, then, an increased risk of neurological damage. On those grounds, in the last 10 years, the cannulation of femoral artery gradually diminished.

In most centers, unilateral ASCP is performed with the axillary artery as a cannulation site, while bilateral ASCP may be performed using the right axillary artery and the left common carotid artery or the so-called “Kazui technique” [16] based on the direct cannulation of both the right and left carotid arteries.

Consensus regarding optimal cerebral protection strategy in aortic arch surgery is lacking, so the ideal method of cerebral perfusion remains undefined and is still a matter of debate. A systematic review and meta-analysis [17] assess the difference between cerebral protection strategies in terms of postoperative mortality and disabling stroke, by comparing different types of aortic arch surgery and pathologies: degenerative aortic aneurysm, acute type A aortic dissection, and post-dissection chronic aneurysm. The group of patients with unilateral ASCP had the lowest rate of operative mortality and disabling stroke. This was observed despite the unilateral ASCP group having the highest proportion of acute aortic dissections, especially relative to the bilateral group. The longest CPB, aortic cross-clamp, hypothermic circulatory arrest, and selective cerebral perfusion times were noted in the bilateral ASCP group, whereas bilateral ASCP was mainly used for total arch replacement. Even if it seems advantageous to prefer unilateral ASCP in case of a short period of circulatory arrest, not all patients may be suitable for this cerebral perfusion technique because of their cerebral vascular anatomy and physiology.

Even if some studies demonstrate the importance of collateral vessels like ophthalmic arteries, leptomeningeal vessels, and external carotid arteries which may permit the use of unilateral ASCP in patients with CoW variations, it is not considered protective when the duration of circulatory arrest is more than 50 minutes. For these reasons, De Paulis et al. recommend using bilateral ASCP if the duration is estimated to exceed 30–40 minutes [18]. Patel et al. support the use of bilateral ASCP, especially when the duration of circulatory arrest time is estimated to be greater than 30 minutes, due to the complexity of the procedure, as a total arch replacement, since 6–17% of adult population has an incomplete CoW, even if unilateral ASCP is a simple and reproducible technique [19].

In relation to CoW variations, the current expert consensus document of EACTS and ESVS recommends preoperative imaging studies to assess the patency and the morphology of CoW where treatment involves the aortic arch. Furthermore, it may not be feasible in an acute setting, as acute type A aortic dissection (ATAAD), which has a mortality rate of 50% in the first 48 hours after the onset of symptoms if surgery is not timely. Overall, the mortality rate of non-surgical treated patients is 58% while among surgically treated patients is 26%, with an incidence of postoperative neurological deficit of about 10–30%.

A systematic review and meta-analysis [20] show the difference between ASCP and RCP, also comparing bilateral with unilateral ASCP in patients affected by ATAAD who underwent emergent surgery. TND risk did not differ between RCP and ASCP, even if ACP increases the risk of embolic events and neurologic deficit while RCP theoretically decreases the risk of emboli formation but cannot provide physiological brain perfusion with a greater risk of hypoperfusion. However, the pooled risk ratio for TND demonstrates that unilateral ASCP is superior to bilateral in preventing neurological deficit, even if the ICU stay is longer in the unilateral group than in the bilateral one. No significant differences were shown between RCP and ASCP, either bilateral or unilateral, in terms of mortality and PND. Hence, circulatory arrest times were longer in the ASCP group than the RCP group, according to other studies where ASCP is preferred over other brain perfusion strategies, mainly bilateral, for longer periods of circulatory arrest. Furthermore, the core temperature was elevated in the ASCP group, resulting in reduced durations for cooling and rewarming, even if no significant difference in the CPB duration was seen.

In conclusion, for acute pathologies and emergency aortic arch surgery, bilateral ASCP is suggested for longer circulatory arrest durations; unilateral ASCP is safe for shorter durations but is a superior strategy for preventing TND and may be encouraged in case of augmented neurological deficit risk as poor preoperative mental status, diabetes, and peripheral arterial disease if it is allowed by surgeon expertise and resources; RCP is the less complicated method and can be performed in more centers and by most surgeons but requires a greater grade of hypothermia.

ASCP is the preferred strategy both in acute type A aortic dissection and in elective aortic arch surgery because the physiological blood flow pattern direct to the brain is associated with improved neurological protection with a lower risk for embolization into the right-sided cerebral vessels.

For chronic pathologies and non-emergent aortic arch surgery, in addition to circulatory arrest duration and co-morbidities, preoperative imaging of CoW may contribute to the selection of the most appropriate cerebral perfusion strategy, because the presence of any CoW variation could be decisional. Elective surgery allows to improve the preoperative study of patients with chronic pathologies, such as aortic arch aneurysm or type B aortic dissection, who may have a CoW variant, because of the important prevalence in the neurologically healthy population (62,22%) and who may benefit from a specific cerebral protection technique rather than another one, with the aim to reduce TND and postoperative stroke. Even if neurological events are less frequent in elective surgery (3.4%) than the emergent one (12%), the efficacy of cerebral protection strategy is still the most important determinant of favorable postoperative neurological outcomes both in elective and emergent aortic arch surgery. It is important to underline that neurological complications are associated with increased mortality, longer hospitalization, healthcare resources utilization, and impaired quality of life.

Therefore, it is appropriate to perform CTA in non-acute pathologies and non-emergent surgery involving aortic arch to identify patients who may require a bilateral ASCP, in addition to intraoperative monitoring.

As concerns neuromonitoring, it may guide the decision of the cerebral brain protection strategy, because bilateral monitoring assures the achievement of a sufficient amount of arterial blood and, then, an adequate supply of oxygen in both cerebral hemispheres. According to many studies, a longer duration of DHCA is associated with neurocognitive impairment, but TND is consistent with ASCP as well. The question of an effective distribution and ideal quantity of cerebral blood flow, particularly in the left hemisphere in presence of CoW variants, is one of the main issues about unilateral ASCP.

Intraoperative neurological monitoring may prevent negative consequences of a suboptimal cerebral perfusion. Different techniques are available for neuromonitoring [21], such as near-infrared spectroscopy (NIRS), which is a non-invasive method used to determine the regional oxygen saturation of hemoglobin (rSO2) regarding blood in the brain, based on the different absorption capacity of unsaturated and saturated hemoglobin in the near-infrared spectrum. This method employs disposable sensors placed on each side of the patient’s forehead which allows the detection of rSO2 of the cortical brain tissue (3–4 cm of depth) perfused by anterior and middle cerebral arteries, but it only detects rSO2 reduction in the frontal lobes. It permits to continuously monitor changes in the local brain oxygen imbalance to recognize early hypoperfusion through prolonged rSO2 desaturation. The cut-off to define a high risk of neurological injury varies among available NIRS devices, and may be defined as a drop of ≥10% from the baseline with the ForeSight method or ≥20% with the INVOS system, due to the different grades of depth detection, 20 mm below the skin for INVOS and 25 mm for ForeSight [22]. The disadvantage is the lack of monitoring of the parietal and occipital lobes which are part of the posterior arterial territory, useful in CoW variants with absence of anastomosis between the anterior and the posterior circle. Moreover, it does not detect the perfusion of deeper brain regions, as required in elderly patients in which cortical atrophy gets the cortex away from the skull, and the signal may be influenced by ambient light and skin or muscle blood flow. Hemoglobin, total bilirubin, end-tidal carbon dioxide concentration, blood pressure, and inspired oxygen fraction are known factors influencing rSO2 measurement. In emergent surgery, when CTA is not performed to detect CoW variants, an impaired rSO2 may be caused by interrupted CoW if the cannula is not displaced or the arterial line is not obstructed. In this case, a switch from unilateral ASCP to bilateral should be considered. NIRS can be helpful in detecting the opposite situation: persistent hyperemia caused by hyperperfusion and hyperoxia induces vasogenic edema, the major determinant of cerebral hyperperfusion syndrome characterized by migraine symptoms, delirium, focal neurologic deficit, seizures, and coma.

Choi et al. [23] observed that, during CEA, a significant predictor for shunt insertion is the fractional reduction of rSO2 from the preclamping reference point recorded with NIRS system. A fractional decrease of 25.8% considered as cut-off value was not influenced by the patency of the CoW. In case of not patent anterior circulation of CoW, in patients with shunt placement, the maximum fractional decrease of rSO2 and the degree of stenosis in their contralateral carotid artery are significantly higher. Anyway, the drop in the cerebral oxygen saturation registered with NIRS system during the period of uASCP, may be the result of hypoperfusion of left cerebral hemisphere if selected CoW variations are present. Switching to bASCP may lead to restoration of perfusion in the left cerebral hemisphere with a significant increase in cerebral oxygen saturation. Despite its high false positive rate, cerebral oximetry may be properly sensitive to predict cerebral ischemia during CEA as suggested by these findings. While the positive predictive value was lower, cerebral oximetry could be an advantageous intraoperative device during aortic arch surgery owing to its high negative predictive value.

The latest evolution in terms of optical techniques are Diffuse Correlation Spectroscopy (DCS) and Frequency-Domain NIRS (FDNIRS). These techniques are based on the evaluation of the optical properties of tissues correlated to blood flow and hemoglobin concentration, respectively. Further to cerebral oxygen saturation, they also offer quantitative measures of cerebral blood volume and perfusion, useful for the calculation of a quantitative index of cerebral oxygen consumption (iCBF). However, the role of these techniques in cardiac anesthesia is still to be defined.

Transcranial Doppler (TCD) is a non-invasive method useful for the evaluation of blood flow in large intracranial vessels (i.e., middle cerebral artery) and the detection of micro- and macro-emboli. In particular, for patients with an incomplete CoW, when an inadequate cross-filling is detected and there is no perfusion in one of the middle cerebral arteries (i.e., due to iMCA), a shift from uASCP to bilateral could be quickly performed. When the decrease of middle cerebral artery velocity is ≥50% from the baseline during cerebral perfusion, it may guide the decision to change cerebral protection strategy. The TCD, differently from NIRS, allows real-time detection of cerebral blood flow, possibly optimizing the perfusion strategy earlier. Unfortunately, TCD is not applicable in approximately 20% of patients because of poor acoustic transcranial window [24]. It is not a recommended technique to detect CoW variants in an intraoperatory setting because a percentage of patients, estimated to be 15%, do not tolerate carotid compression for the assessment of CoW function. Even under the carotid compression test, the exact hemodynamic condition of cerebral perfusion is not accurately simulated, so the effective perfusion pressure during brain perfusion could be underestimated. Patients with severe carotid atherosclerosis who undergo aortic arch repair, estimated to be 3% of all patients, should avoid carotid compression. Thus, hemodynamics between preoperative and intraoperative evaluation with TCD is very different, because of the effects of non-pulsatile blood flow, lower mean pressures, hypothermia, and anesthesia which are not considered during the initial functional evaluation of CoW. Lastly, TCD is largely reliant on the operator and has not been commonly employed in cardiac surgery. These findings encourage the preoperative assessment of CoW with CTA whenever possible and the limitation of TCD as an integrative approach for NIRS in intraoperative monitoring in selected cases.

Electroencephalogram (EEG) has been widely used as an intraoperative monitoring tool to detect cerebral ischemia. Variations in EEG, such as diffuse or focal attenuation of the high and the middle frequencies, asymmetric changes in frequency and amplitude of the recording, and persistent increase of slow waves, are considered signs of cerebral ischemia. However, this method is susceptible to influences by anesthetic effects, leading to potential false negatives, especially in the case of micro-embolism. It cannot detect subcortical lobe ischemia or silent brain stroke caused by micro-embolisms. In this regard, the Bispectral index (BIS) is a processed EEG parameter introduced to indicate the depth of anesthesia and cerebral ischemia. A certain decrease in BIS may be indicative of cerebral ischemia, but the exact threshold has not been yet identified and it has a limited positive predictive value, so false positives may occur.

Somatosensory evoked potential (SSEP) is the electrical activity response measured at the skin’s surface following controlled peripheral nerve stimulation and may indicate cerebral ischemia if it changes ≥50% from the baseline. Unlike EEG and NIRS, SSEP analyzes not only the cortex but also the deepest structure of the brain, but it demands expertise and may not properly evaluate all aspects of cerebral perfusion. Thus, EEG and SSEP measurements are affected by hypnotic agents and electrosurgical devices, including electrocautery.

As demonstrated by various studies, the intraoperative neuromonitoring technique cannot replace CTA as preoperative technique for the evaluation of CoW variants and they are characterized by different limitations and positive or negative predictive values. In conclusion, neuromonitoring is useful to shift cerebral perfusion technique for emergent surgery, when CTA is not performed and is reasonable to suspect CoW variants, and in elective surgery to evaluate technical intraoperative problems with CPB or cerebral protection strategy, such as cannula malposition or arterial line obstructions.

4. Systematization of CoW variants and cerebral perfusion techniques

As discussed, a new and specific classification is needed when performing non-emergent and elective aortic arch surgery in patients affected by CoW variants, in order to choose the best approach to cerebral perfusion.

The normal anatomy of CoW protects patients from ischemia because it warrants the perfusion of all cerebral areas, even when an internal carotid artery is clamped and even when the contralateral internal carotid artery is occluded, despite the brain being perfused only via the basilar artery. However, during cardiac and vascular surgery, in presence of anomalies of the CoW, the perfusion of some cerebral areas can become insufficient and produce clinical neurological deficits. Besides the variable presence of connections in leptomeningeal and extracranial networks, collateral flow in the brain (which includes the brain stem, the cerebellum, and both cerebral hemispheres) is primarily dependent on the CoW. For these reasons, unilateral ASCP is not always suitable and in selected cases, it is indispensable to perform a bilateral ASCP, while RCP is always suitable but is CPB-time limited and it is impossible to determine whether the brain is really perfused with oxygenated blood or it is shunted with cerebral oedema as a major complication. In addition, DHCA, if performed for 20 minutes is considered safe, but this period is too short for complex arch reconstruction. A recent European survey [18] showed that approximately two-thirds of centers prefer bASCP, whereas one-third of them deliver the cerebral perfusate in an unilateral manner; DHCA and RCP are rarely used in Europe.

The aim is to develop a systematic classification of CoW variants which are likely to have an increased neurological deficit risk if aortic arch surgery is performed with unilateral ASCP. The preoperative prediction of bilateral ASCP reduces cerebral hypoperfusion and, consequently, TND and neurological complications.

Evidence are summarized from studies in patients undergoing carotid artery surgery (Tables 2 and 3) and non-emergent arch surgery (Tables 4 and 5).

Study (year)	Patients	Age	Clinical presentation / neurological comorbidities	Intervention	CoW preoperative assessment (method)	Functionally complete CoW (%)	Intraoperative cerebral protection strategy	Intraoperative cerebral protection monitoring	In-hospital mortality	Stroke (minor, transient, major, global brain ischemia)
Banga et al. [11]	545	Patients without Immediate Neurologic Events (INE): 68 ± 8.2); Patients with INE: 71.6 ± 8.9	38% with previous neurological symptoms	CEA	CT angiography	60.3%	Selective shunt	N/A	0.2%	3.7%; (8 patients with permanent stroke, 12 patients with transient stroke)
Choi et al. [23]	310 (273 shunt, 37 no shunt)	Group shunting (GS), 66.6 ± 10.4; Group no shunting (GNS), 67.3 ± 8	N/A	CEA	MR angiography	N/A	Selective shunt	Clinical evaluation in awake patient	0	1.0% (dysarthria, disorientation, and headaches at territories of the middle cerebral artery and anterior communicating artery)
Gibello et al. [25]	503	73.34 ± 7.43	33.7% with previous neurological symptoms	CEA	CT angiography	N/A	Pruitt-Inahara selective shunt	N/A	N/A	Previous neurological symptoms, diabetes and female sex are correlated with a lower tolerance to carotid cross-clamping.; Chronic obstructive pulmonary disease is correlated to higher tolerance to carotid cross-clamping.
Montisci et al. [26]	71	Group “Complete CoW” (35 patients), 71.8 ± 7.2).; Group “one agenesia/obstruction in the CoW” (21 patients), 71.6 ± 9.6.; Group “2+ agenesias / obstructions in the CoW” (15 patients), 73.1 ± 5.2	29.5% preoperative neurological symptoms	CEA	MR angiography	78.9%	Pruitt-Inahara shunt (11 patients)	Clinical evaluation in awake patient	0	0
Pennekamp et al. [27]	431	69 ± 9	Neurological symptoms: 88% (TIA 65%, stroke 23%)	CEA	CT angiography (33%); MR angiography (67%)	83.8% of patients studied with MR angiography	Selective shunt	TCD + EEG	0.7%	3% stroke
Soares et al. [28]	98; (54 with complete CoW, 44 incomplete CoW)	72.8 ± 3.0	Previous neurological symptoms (53.1%)	CEA	CT angiography, MR angiography	55%	Selective shunt	EEG	2%	3.1%
Zhou et al. [29]	482	61.2 ± 10.4	Hypertension in subjects with incomplete CoW was significantly greater vs. complete CoW (72.2% vs. 55.7%; P = 0.011)	/	MR angiography	12.7%	/	/	/	/

Table 2.

Studies describing completeness of CoW in patients undergoing carotid endarterectomy: Details and methods.

Study (year)	CoW variants, details and outcomes	Comments
Banga et al. [11]	Complete CoW: 1 INE; Isolated anomaly in anterior semicircle (72 patients): 1 INE; Isolated anomaly in posterior semicircle (196 patients): 2 INEs; No differences in INEs between complete CoW and single/isolated anomaly (P = 0.57); Abnormalities in both semicircles (215 patients): 16 INEs. Significant differences compared to other groups in terms of INEs (P = 0.001). Among these patients, 8 out of 34 patients with iMCA (isolated middle cerebral artery) had INEs (23.5%).; iMCA is a strong predictor of INEs (OR 11.1, 95%CI 3.6–35.9, P < 0.01)	An iMCA (isolated middle cerebral artery) carries more than a 10-fold higher risk of INEs after CEA with cross-clamping without shunt protection. In these patients, routine shunting is recommended to prevent INEs.
Choi et al. [23]	Absence of the anterior circulation in CoW is the most important factor for shunt insertion, assuming no stenosis in the contralateral carotid artery during CEA.; rSO2 can be applied in predicting shunt insertion, irrespective of the patency of the circle of Willis, although the positive predictive value was low. The cut-off value was 25.8%.	When the anterior circulation was non-patent, the probability of shunt insertion increased as the contralateral carotid arteries were more severely stenosed.
Gibello et al. [25]	The most common anomaly is the ipsilateral or contralateral hypoplasia/occlusion of the posterior communicating artery.; Significant correlation between the number of CoW anatomic anomalies and neurological tolerance to cross-clamping: two or more hypoplasic or occluded segments of the circle are positively correlated to worse cerebral compensation.	A proposed score (PTOT) is the summary of patency of CoW and the supraortic trunk.; PTOT is a strong predictor of cerebral tolerance to carotid cross-clamping.
Montisci et al. [26]	Strong association between intolerance to carotid cross-clamping and two or more agenesiae / obstructions in the CoW (OR 51, 95% CI 5–480, P < 0.001).; No significant difference in the risk of developing carotid cross-clamp intolerance between the presence of one agenesia/obstruction in the CoW versus complete CoW (OR 1.7, 95% CI 0.1–28.7, P = 0.71).; Probability of having two agenesiae / obstructions in the CoW in patients with carotid cross-clamp intolerance: 82%; Probability of having zero or one agenesia / obstruction in patients without carotid cross-clamp intolerance: 90%; Probability of a patient having carotid cross-clamp intolerance when the CoW has two agenesiae / obstructions: 60%; Probability of a patient not having carotid cross-clamp intolerance when the CoW has zero or one agenesia / obstruction: 96%	Two or more agenesia / obstructions of the CoW were associated with a high risk of intolerance to carotid cross-clamp.
Pennekamp et al. [27]	Posteriorly, the most observed abnormality was a not-visible P-coms (47% and 45% for the ipsilateral and contralateral P-com respectively); 1. Circle of Willis morphology with MR (284 patients); 1A) Not visible anterior segment (any) (48 patients); among them, shunt used in 14 patients (P < 0.01); 1B) Not visible posterior segment ipsilateral (any) (136 patients); among them, shunt used in 34 patients (P < 0.01); 2. CoW morphology with MRA / CTA (431 patients); 2A) Absent anterior segment (any) (77 patients); among them, shunt used in 21 patients (P < 0.01); 2A1) Absent A-com (49 patients); among them, shunt used in 12 patients (P = 0.05); 2A2) Ipsilateral A1 segment hypoplastic (61 patients); among them, shunt used in 1 patient (P < 0.01); 2B) Absent ipsilateral post segment (241 patients); among them, shunt used in 49 patients (P = 0.01); 2B1) P-com not-visible (201 patients); among them, shunt used in 38 patients (P = 0.04); 2B2) P-com < P1 (121 patients); among them, shunt used in 9 patients (P < 0.01); 2C) Absent contralateral post segment (194 patients); among them, shunt used in 25 patients (P = 0.75)	In the MRA group, any not-visible anterior (OR 4.96, 95% CI 1.95–12.58) or ipsilateral posterior segment of the CoW (OR 5.08, 95% CI 2.10–12.32) was independently related to the need of shunt use during CEA; In the CT group none of the factors were independently related to the need for shunt use during CEA. Therefore, only predictors describing morphology of CoW derived from MRA findings were included in this model.; Imaging with MR seems superior compared to TC in terms of image quality and clinical / prognostic implications.
Soares et al. [28]	During CEA 4 patients required selective shunt (2 with complete CoW, 2 with incomplete CoW; P = 0.15).; The incidence of postoperative stroke was 3.1% (asymptomatic 2%, symptomatic 1.1%), with no differences among CoW complete and incomplete groups (3.7 vs. 2.3%, P = 0.68).	The completeness of the CoW might not influence the outcomes of carotid interventions.; Anatomy of the CoW is not a strong enough predictive factor for the prognosis of cross-clamping intolerance during CEA
Zhou et al. [29]	The prevalence of incomplete CoW was significantly higher in Posterior-COW than that of anterior CoW (79.7% [384/482] versus 40.7% [196/482], P < 0.001).; The prevalence of an incomplete A-COW was significantly higher in patients with presence of carotid intraplaque hemorrhage than those with a complete CoW (52.7 versus 38.5%; P = 0.022) especially with the absence of ACoA (69.2%; P = 0.001); Incomplete A-COW significantly associated with the presence of intra-plaque hemorrhage before (OR, 1.781; 95% CI, 1.083–2.931; P = 0.023) and after adjusted for clinical risk factors including age, sex, body mass index, smoking, diabetes mellitus, pulse pressure, and hyperlipidemia (OR, 1.945; 95% CI, 1.139–3.321; P = 0.015)	An incomplete CoW is independently associated with intraplaque hemorrhage of carotid atherosclerotic plaques.

Table 3.

Studies describing completeness of CoW in patients undergoing carotid endarterectomy: Outcomes and clinical implications.

Study (year)	Patients	Age	Clinical presentation / comorbidities	Intervention	CoW preoperative assessment (method)	Functionally complete CoW (%)	Cannulation strategy and hypotermic arrest	Intraoperative cerebral protection monitoring	In-hospital mortality	Stroke (minor, transient, major, global brain ischemia)
Mariscalco et al. [30]	76; (61 m, 15 f)	65.4 ± 16.8	atherosclerotic descending thoracic aortic aneurysms (35, ruptured 3), chronic type B dissecting aneurysm or complicated acute type B dissection (16), penetrating ulcers (14, ruptured 2), traumatic pseudoaneurysms (11, chronic 3).	TEVAR	CT angiography	70	not applicable	—	12	0 hemorrhagic stroke, 7 ischemic (embolic) stroke,; 1 low-flow ischemic stroke
Smith et al. [31]	391	62 ± 12 years, 61% male (n. 237)	—	aortic arch replacement	CT angiography	27% complete Cow (107)	temperature of 25°; two 15-French perfusion cannulas into the ostia of the innominate and left common carotid arteries	—	—	—
Urbanski et al. [32]	99; (67 m, 32 f)	65 years; (range 25–84 years)	—	aortic arch replacement	CT angiography	—	the left (93) or right (1) carotid artery was prepared in 94 patients.; moderate; hypothermia with rectal temperature of 30°C.; circulatory arrest and unilateral cerebral perfusion time: 18 ± 10 min (range 7–70).	trans-cranial doppler	1	—

Table 4.

Studies describing completeness of CoW in patients undergoing cardiac / arch surgery: Details and methods.

Study (year)	CoW variants, details, and outcomes	Comments
Mariscalco et al. [30]	CVA Willis anomal vs. CVA normal Willis; 37.5% vs. 4.4%, p < 0.01).	incompleteness of the Willis circle (OR 17.19, 95% CI 2.10 to 140.66; p = 0.008) and presence of coronary artery disease (OR 6.86, 95 CI% 1.18 to 40.05; p = 0.032) predicted neurological events
Smith et al. [31]	Moderately safe group, Total moderately safe group 70 (18); Left fetal variant 12 (3); ACoA dysfunction 9 (2); Bilateral fetal variant 30 (8); ACoA and right PCoA dysfunction 7 (2); ACoA dysfunction and right fetal variant 6 (2); Right PCoA dysfunction and left fetal variant 6 (2); Unsafe group, Total unsafe group 40 (10); ACoA and left PCoA dysfunction 6 (2); ACoA dysfunction and left fetal variant 0 (0); ACoA dysfunction and bilateral fetal variant 5 (1); ACoA and bilateral PCoA dysfunction 15 (4); ACoA and right PCoA dysfunction and left fetal variant 1 (<1); ACoA and left PCoA dysfunction and right fetal variant 1 (<1); Right VA dysfunction 12 (3)	ischemic brain damage during right u-ASCP (28%) is greater than during bi-ASCP (<1%); 62% functional CoW safe for uSCP; 10% PCoAr disfunction (41); 13% PCoAl disfunction (50); 17% bilateral PCoA disfunction (65)
Urbanski et al. [32]	59 variants, 3 groups, 18 patients; group I: abnormality anterior or posterior part; group II: hypoplastic or missing posterior communicating arteries on both sides; group III: abnormalities within the anterior and posterior communicating arteries	—

Table 5.

Studies describing completeness of CoW in patients undergoing cardiac / arch surgery: Outcomes and clinical implications.

Urbanski et al. [32] divided CoW variations into three groups: in the first group, there is only one hypoplastic or absent segment; both sides of the second group had abnormalities in the posterior pre-communicating (2a) or communicating arteries (2b); in the third group, the abnormalities are in both anterior and posterior segments of the CoW on one or even on both sides. This last group is the only one that might benefit from bASCP and, in some cases (missing posterior communicating segments on the left and anterior communicating segments on the right, as iMCA, or above two anomalous positions within the CoW) even a bASCP could not warrant a sufficient cerebral perfusion. In this study, 99 patients affected by different CoW variants underwent elective aortic arch surgery, with hemiarch, subtotal, or total arch replacement. uASCP was performed through the left carotid artery as a cannulation site and, when required, the right carotid artery was used as adjunctive perfusion cerebral artery for bASCP. Unexpectedly, during the carotid artery occlusion test with TCD, no significant decrease of flow velocity in the MCA was observed and for patients in whom the use of TCD has not been possible, no signs of cerebral ischemia were observed at EEG and SSEP. Even the measurement of blood pressure in the radial arteries, which reflects the blood pressure in subclavian and, then, in vertebral arteries, demonstrated that there aren’t differences in cerebral blood supply during uASCP in patients with complete and incomplete CoW. Urbanski supports that incompleteness of CoW does not correlate with insufficient cerebral protection during uASCP and, for this reason, does not recommend CTA as a standard procedure before elective aortic arch surgery, only functional carotid occlusion test and intraoperative neuro-vascular monitoring are considered essential. These conclusions are probably influenced by very short duration of ASCP, optimal pharmacological cerebral protection, and the etiopathology of degenerative or atherosclerotic aneurysm. Since atherosclerosis is a generalized disease, in this particular patient’s cohort, there is a possibility for pathological expansion of secondary collaterals (as leptomeningeal and ophthalmic arteries). Moreover, the old-generation CT scan used in this study might misclassified segments with a diameter over 1 mm as aplastic or hypoplastic, because of the low spatial resolution of this equipment.

Mariscalco et al. [30] demonstrated that patients affected by atherosclerotic descending thoracic aortic aneurysm, chronic type B dissecting aneurysm or complicated acute type B dissection, penetrating ulcers and traumatic pseudoaneurysm who underwent thoracic endovascular aortic repair (TEVAR) intervention, were prone to cerebrovascular accidents after the procedure if they had incompleteness of CoW. This study, then, proves that anomalies of CoW are independent cerebrovascular accident predictors after TEVAR procedure.

Smith et al. [31] assumed that a Cow variant may be considered functionally safe as long as the collateral flow when performing uASCP did not have to reach a vessel beyond the left MCA. An example is the isolated variation of PcomA because the blood flow of the left MCA and the left A1 is possible through the AcomA, while the basilar and the right vertebral arteries can perfuse the left PCA. In this case, uASCP is considered safe. When the cerebral areas were inaccessible to the CoW-dependent collateral blood flow during right uASCP, the functional variation is defined as unsafe, such as right vertebral artery dysfunction, combined with either bilateral PcomA dysfunction or combined with fetal type variant. Finally, right uASCP was defined as moderately safe if right-to-left posterior-to-anterior collateral perfusion had to arrive at a vessel outside the left MCA. On this basis, they propose a clinical traffic light algorithm (red light, amber, and green light) categorizing different functional anatomical variants as unsafe, moderately safe, and safe for uASCP. They also consider vertebral arteries hypoplasia or absence, differently from other studies. This classification is based not on the morphology but on the functional anatomy of CoW as is to be expected during uASCP, lie on preoperative transcranial color-coded duplex (TCCD) examination of bilateral common, internal and external carotid artery, and both vertebral arteries, performing carotid compression test whenever possible. As emphasized by Etz and Borger [24], a significant number of patients are not eligible for the proposed TCD assessment of CoW functionality, and its clinical significance for the majority of patients remains unclear. Therefore, a preoperative CTA-based classification is needed.

Papantchev et al. [9] studied 500 CoW with two different methods: 250 were obtained from cadavers via routine dissection, while the others were visualized with CTA imaging technique. Otherwise, other classifications that are not considered relevant whether the variation is on the right or left side of the circle, the left–right lateralization in uASCP is of crucial importance, since in uASCP with the cannulation of the right axillary artery or brachiocephalic trunk, the brain receives blood only via the right common carotid artery and right vertebral artery (VA). Therefore, the protective effect of uASCP is entirely based on the CoW which maintains adequate perfusion of the contralateral left cerebral hemisphere through A1s, P1s, AcomA, and PcomAs segments. Variations of the circle exist in at least 50% of the people, so it is a not inconsiderable factor to take into account in the selection of the adequate cerebral perfusion strategy. Moreover, the classifications seen before do not consider vertebral arteries variations because they are not part of the CoW, but variations in those arteries could have crucial importance during uASCP. For these reasons, Papantchev defines five segments of CoW as critical for uASCP: right A1, AcomA, left PcomA, left P1, and right VA. A diameter of 1 mm was used as the threshold for hypoplasia, in order to be consistent with other major morphological studies of the CoW reported in the literature. Nine circle configurations were identified and they were subdivided into seven groups, according to the number of major cerebral vessels (ACAs, MCAs, PCAs) at risk of hypoperfusion during uASCP:

• Type IA includes all circles with hypo- or aplasia of left PcomA. With this variation, the territories supplied by the left MCA are at risk of hypoperfusion during uASCP. This is the most frequent variant (35.6%), according to data reported in literature where PcomAs is the most variable part of the circle;

• Type IB includes all circles with AcomA variations. The territories supplied by the left ACA are at risk of hypoperfusion;

• Type IIA includes all circles with hypo- or aplasia of both left PcomA and AcomA. The left cerebral territories supplied by ACA and MCA are at risk of hypoperfusion;

• Type IIB includes all circles with left P1 or right VA variations, which leads to hypoperfusion risk of brain territories perfused by both left MCA and left PCA;

• Type IIIA includes all circles hypo- or aplasia of right A1, with high hypoperfusion risk of brain territories supplied by left MCA, right and left ACA

• Type IIIB includes all circles hypo- or aplasia of both AcomA and right VA. If this circle type is present, the entire left hemisphere is at risk of hypoperfusion during uASCP. This is the rarest variation (0.2%);

• Type IV includes all circles with hypo- or aplasia of both right A1 and right VA or both right A1 and left P1. This is the most severe variation because it leads to the risk of hypoperfusion of territories perfused by right ACA, left ACA, left MCA, and left PCA.

Hence, if the CoW variation leads to the risk of hypoperfusion of only one major cerebral vessel, it is classified in group I and so on. All types were present in 58.6% of all examined CoW, according to the high frequency of CoW variants in healthy people found by other studies. Papantchev’s classification is simple and straightforward, dividing a large number of complex CoW variations into only seven types, which is a small number if compared to other classifications. It is focused on high-risk variations that may benefit from bASCP when the preoperative neurological examination with CTA method is possible.

The study of Papantchev may be integrated with vascular surgery studies, comparing patients with CoW variants who needed shunting during CEA, to be able to predict, in the preoperative assessment, when the use of bilateral ASCP is recommended to prevent neurological sequelae.

Patients who underwent CEA intervention, affected by CoW variants, are estimated to be 50%, the same frequency encountered in cardiac surgery. When performing CEA during vascular surgery, the carotid cross-clamping (CC) intolerance and the demand for shunting in patients affected by some specific types of CoW variants, can be helpful in creating a complete classification of high-risk variants because CC is based on the same principle of unilateral ASCP, where cerebral blood flow is provided by only one carotid artery. The identification of high-risk variants simplifies the choice between unilateral and bilateral ASCP in elective aortic arch surgery, because patients with some specific CoW variants may require a bilateral ASCP instead of unilateral to ensure adequate cerebral perfusion in both brain hemispheres, even when CPB time is short.

Bagan et al. [33] found that configurations type number 5, 14, and 15 according to Lazorthes’ classification are likely to have CC intolerance. Type 5 is characterized by the absence of AcomA and PcomA ipsilateral, while in type 14 A1 segment contralateral and PcomA ipsilateral are absent and in type 15 AcomA and P1 segment contralateral are absent. The common peculiarity of these configurations is the absence of 2 segments and, except for type 15, the isolation of the middle cerebral artery from the contralateral carotid artery and vertebral arteries. Later, Montisci et al. [26] demonstrated that the number of absent segments in the CoW is significantly associated with intolerance to CC, independent of the CoW configurations. When two or more segments are absent, a cerebral perfusion deficit is highly probable, whereas other factors such as ipsilateral, contralateral carotid stenosis, and preoperative symptoms are not statistically associated with CC intolerance. The absence of only one segment does not appear to be correlated with CC intolerance if not associated with contralateral carotid occlusion or low blood pressure. The location of the absent segments can influence the extension of hypoperfused cerebral area and consequent clinical signs. In those patients with the absence of two or more segments without CC intolerance, it is likely that a very small area or a neurologically irrelevant area was hypoperfused. Also, TCD investigation showed correlation between undetectable CoW segments and decreased flow velocity in MCA with ischemic EEG abnormalities. Even if the patients of this report were studied using magnetic resonance angiography (MRA), it is recommended to prefer CTA for the preoperative assessment of CoW because CTA is not flow-dependent and allows for image reformation along any plane. Gibello et al. [25] found a strong association between the insertion of shunt during CEA and the number of obstructed or hypoplastic CoW vessels. The results revealed that there is no one specific CoW configuration that is more susceptible to carotid CC intolerance, but the increased number of hypoplastic or occluded CoW segments leads to a higher likelihood of neurological deficits following carotid CC. Hence, it is reasonable to adopt bilateral ASCP as a preferred cerebral perfusion strategy in patients with two or more absent segments of CoW to prevent hypoperfusion of cerebral areas.

Banga et al. [11] studied the impact of iMCA on immediate neurological events, such as transient ischemic attack and stroke, diagnosed immediately after CEA. They divided CoW into two semicircles: the anterior, formed by the contralateral anterior cerebral artery (A1), anterior communicating artery (AcomA) and ipsilateral A1; the posterior, composed by the ipsilateral carotid artery (P1) and posterior communicating ipsilateral artery (PcomA). Among the patients with normal CoW, only one had a stroke and among patients with one hypoplastic or absent segment in the anterior or the posterior semicircle, three had immediate neurological events. Statistical data confirms that there is no difference in terms of neurological deficits between patients with complete CoW and patients with only one semicircle affected. The difference becomes significant when both semicircles (the anterior and the ipsilateral posterior) have at least one hypoplasia or absent segment at the same time. This evaluation allows us to consider hypoplasia as absent when both semicircles are involved in anomalous CoW, confirming the hypothesis of augmented neurological risk when two or more segments are absent or, in addition, hypoplastic. Among the patients with at least one absent or hypoplastic segment in both semicircles, those with iMCA carried more than a 10-fold higher risk of immediate neurological events after CEA with CC without shunt protection. This means that, on the side of the clamped ICA, the MCA flow is not supported by contralateral carotid nor from the vertebral artery, due to the absence of communication between the two semicircles.

Thus, iMCA may be considered a particularly high-risk variant in the group of two absent segments who are susceptible to bilateral ASCP. Therefore, all CoW variants characterized by at least two absent segments, in addition to Papantchev’s types IIA, IIIB, and IV, may benefit from bASCP.

Other vascular surgery studies [23] in patients who underwent CEA procedure demonstrated that the absence of anterior circulation was the primary factor in the insertion of a shunt, given that there was no stenosis in the contralateral carotid artery. In this case, the evaluation of contralateral carotid stenosis became important because it increases about 27% the probability of shunt insertion. Previous studies documented that the collateral flow rates in AcomA rose with the increases of the degree of stenosis of ICA and this is particularly true when AcomA is absent because an augmented blood flow and pressure is registered in patients with ICA stenosis, probably due to a physiological increased blood flow through leptomeningeal collaterals if a patient has poor CoW integrity. This is also the reason why an incomplete CoW is a risk factor for intraplaque hemorrhage and atherosclerotic plaque rupture [29]. Hence, the second risk group may be constituted of patients with CoW variants involving absent segments in the anterior circle who may need an evaluation of contralateral carotid stenosis with carotid ultrasound, in addition to the first risk group formed by the absence of at least two segments and the iMCA subgroup. The cut-off of contralateral carotid stenosis which allows patients with anterior segments absence to benefit the bilateral ASCP instead of unilateral, is estimated to be above 50% (Table 6).

Papantchev’s group	Variation	Incidence	Proposed strategy of cerebral perfusion
Group I; Group I-A; Group I-B	Left PComA; AComA	30–35%; 2–5%	Monolateral (RS/RIA); Monolateral (RS/RIA)
Group II; Group II-A; Group II-B; iMCA	Left PComA + AComA; Left P1 or right vertebral artery; Isolated middle cerebral artery	3–8%; 7–10%; 2–5%	Bilateral (RS/RIA + LCC); Monolateral (RS/RIA); Bilateral (RS/RIA + LCC)
Group III; Group III-A; Group III-B	Right A1; AComA + right vertebral artery	4–8%; 1–2%	Trilateral (RS/RIA + LCC + LS); Bilateral (RS/RIA + LCC)
Group IV	Right A1 + right vertebral artery / left P1	1–2%	Trilateral (RS/RIA + LCC + LS)

Table 6.

Proposed management strategy of antegrade cerebral perfusion during non-emergent arch surgery based on CoW variants. In group I, there is only one location of an anatomical variation (hypoplasia, aplasia) within the CoW. In this group, an adequate collateral supply of the entire CoW could generally have been expected, even if it had been supplied through only one artery (i.e., the right carotid artery). In group II, the abnormalities existed within the posterior pre-communicating (group II-A) or communicating arteries (group II-B) on both sides. In cases belonging to group II-A, a sufficient supply of the posterior cerebral region could have been expected, regardless of whether the CoW had been supplied from the left side (left carotid artery) or from the right side (right carotid and right vertebral arteries). However, in cases belonging to group II-B, a perfusion only from the right side would offer a sufficient collateral flow. In group III, the pathologies were found in both anterior and posterior components of the CoW on one or even on both sides so that an insufficient cerebral supply could have occurred during the perfusion of the left side and in some cases (e.g., missing anterior communicating components on the right and posterior communicating components on the left or in cases with more than two abnormal locations within the CoW) even during perfusion from the anterior part of the right side; therefore a 3-way perfusion (right innominate artery + left carotid + left subclavian) would have been advised. In group IV, functional incompleteness of CoW is total and a 3-way perfusion appears mandatory.

PComA: posterior communicating artery; AComA: anterior communicating artery; RS: right subclavian / axillary artery; RIA: right innominate artery; LCC: left common carotid artery; LS: left subclavian artery.

5. Conclusions

Neurological injuries, stroke, and paraplegia are the most devastating complications of aortic arch surgery. The non-emergent surgery and chronic diseases of aortic arch permit to notably reduce neurological deficits by the given opportunity to study in a preoperative assessment with CTA the patients’ CoW which may be affected by various abnormalities, such as hypoplasia or absence of one or more segments.

A simple systematic classification may provide decisional criteria to prefer bASCP over uASCP in selected CoW variations to easily identify, among 82 variants, the ones which have a high risk of hypoperfusion during uASCP. Papantchev’s classification may be integrated by carotid cross-clamping studies during CEA, which allows to identify other two high-risk groups: the first one, which includes CoW variants with the absence of at least two segments (which includes types IIA, IIIB, and IV), characterized by a subgroup identified as iMCA; the second group, which includes at least one absent segment in the anterior circle and the contralateral carotid stenosis over 50%.

The intraoperative monitoring of cerebral blood perfusion is essential but cannot replace preoperative CTA because the evaluation of CoW incompleteness is decisional in preferring a determinate cerebral perfusion strategy, due to an augmented risk of stroke and TND if aortic arch surgery is performed with uASCP. bASCP is recommended over uASCP in high-risk groups to warrant an adequate cerebral perfusion and the reduction of postoperative neurological complications. The use of uASCP when the CoW proves to have an adequate and sufficient collateral capacity, and bASCP only when necessary and in emergent situations, may minimize surgical manipulation of supraortic vessels, and subsequently decrease the rate of embolic stroke.

Acknowledgments

Authors would like to acknowledge the efforts of Marco Piastra for their help in English language.

Conflict of interest

The authors declare no conflict of interest.

Appendices

ACAs	anterior cerebral arteries
AcomA	anterior communicating artery
ASCP	anterograde selective cerebral perfusion
bASCP	bilateral anterograde selective cerebral perfusion
CA	circulatory arrest
CC	carotid cross-clamping
CEA	carotid endarterectomy
CoW	circle of Willis
CPB	cardio-pulmonary bypass
CTA	computed tomography angiography
DHCA	deep hypothermic circulatory arrest
EEG	electroencephalogram
ICAs	internal carotid arteries
iMCA	isolated middle cerebral artery
MCAs	middle cerebral arteries
MRA	magnetic resonance angiography
NIRS	near-infrared spectroscopy
PCAs	posterior cerebral arteries
PcomAs	posterior communicating arteries
PND	permanent neurological deficit
RCP	retrograde cerebral perfusion
SSEP	somatosensory evoked potential
TCD	transcranial Doppler
TEVAR	thoracic endovascular aortic repair
TND	transient neurological deficit
VAs	vertebral arteries