Experimental Arteriovenous Malformations Modeling in Laboratory Sheep Versus Swine

In a recent article in the AJNR (1), Qian et al presented an interesting attempt at creating an arteriovenous malformation (AVM) model in sheep, a laboratory animal that had not been used previously for this purpose. The authors are to be congratulated on their commendable intentions and efforts to create an experimental AVM model aimed at providing an additional laboratory tool to study these difficult and challenging lesions. Several issues in their article deserve further comment because their presentation raises two significant concerns. I would suggest that the “AVM model” presented is neither “simplified” nor is it truly representative of a cerebral AVM.

The authors stated that the main purpose of their study was to create a “simplified” AVM model specifically, compared with a previously presented swine AVM model (1–6) that does not require in its construction the “highly sophisticated neurointerventional skills” for occluding three “intracranial” arterial branches. Five observations can be made about this statement. First, the three arterial branches concerned, the muscular branch of the ascending pharyngeal artery (not the ascending pharyngeal artery proper, as stated at one point by the authors), the occipital artery, and the external carotid artery, all ipsilateral to the neck fistula, are all extracranial neck arteries easily accessible via the endovascular route should this be required. They are not “intracranial” as alluded to by the authors in their abstract. Arteries above the skull base generally cannot be accessed endovascularly in the swine.

Second and most importantly, occlusion of these three branches had been abandoned long ago in the swine model, as documented in the literature (5, 6) but not referenced by Qian et al. It was found subsequent to the first studies using the swine model, in which occlusion of side branches was performed, that in fact it behaved hemodynamically (ie, with satisfactory and realistic angiographic shunting across the nidus) just as well without having to occlude the three side branches. Occlusion of these branches was omitted to simplify even further the construction of the swine AVM model, and therefore, for some time now requiring only the creation of a surgical carotid–jugular fistula compared with the original model first described in 1993. This change was introduced while accepting the theoretical likelihood that this also resulted in marginal reduction of the postoperative blood shunting through and channeling from both retia to the fistula (5, 6). Importantly, hemodynamic validation studies (5, 6) of the swine model without occlusion of the side branches indicated that its transnidal blood flow and pressure characteristics were somewhat representative of human AVMs. Therefore, the raison d’être for the study of Qian et al would appear to be erroneous, because with regard to model construction, one model cannot be deemed to be a simplified version of another if both construction techniques are identical. The current swine AVM is already simple by the criteria of Qian et al.

Third, the methods of occluding arterial branches used in the original swine model were not at all “complex” and do not require “highly sophisticated neurointerventional skills;” quite contrary to the repeated suggestions by the authors. This is misleading because they are usually regarded as very simple and standard vascular interventional maneuvers that can be learned easily, should it be necessary to employ them. Indeed, the actual methodology of occluding these branches was of secondary importance within the overall context of constructing the earlier version of the swine model. For that matter, any possible device or technique (the simplicity/complexity of which could suit one experience) could have served to achieve the objective of vascular occlusion. Even the injection of small amounts of embolic glue could have been used as a way to practice occluding normal arteries, especially the two smaller ones, prior to the subsequent nidus embolization.

Fourth, if the object of creating an animal AVM model is to gain experience in embolization techniques and vascular interventional skills, as has been the intent of Qian et al, then it could be argued that the valuable experience gained in occluding arterial branches in the neck during construction of the AVM model (and in sheep only one branch would have to be occluded [read further]) would actually add considerably to the skills of the trainee and might be a desirable additional exercise to any vascular interventional training course. This had been pointed out in an early article on the swine AVM model (3). The extra cost of having to occlude these side branch(es), whether three in swine or one in sheep, would go toward providing additional trainee experience, or the extra cost and resulting trainee experience can be omitted, as desired by the course organizers. It would be fair to say as well that the very conventional “expensive microcatheters and wires” used in occluding side branches would most likely be reused in the subsequent embolization of the AVM nidus, and therefore these do not represent the source of much extra cost as portrayed by Qian et al.

Lastly, because there is no ascending pharyngeal artery in the sheep, and the external carotid artery ipsilateral to the fistula would have to be preserved to conduct blood down to the fistula, this leaves the ipsilateral occipital artery as the only branch in question (eg, swine model) as to whether it should be occluded or not in the sheep model presented. In this regard, a significant concern of the author's work is their incorrect interpretation of the angiograms they present. Contrary to their statement in the results section, the occipital artery on the fistula side is visualized very clearly as it fills in a retrograde direction from the contralateral occipital artery because of the sump effect of the nearby carotid-jugular fistula. This has significant detrimental implications in terms of flow diversion higher up through the so-called “nidus” (see below). Occlusion of this occipital artery on the fistula side would appear to be essential in the sheep, regardless of its cost, to increase flow diversion and enhance the poor flow through the so-called “AVM” as seen in the postsurgical angiograms presented.

The very act of experimental model construction necessitates the observation of clear fundamental rules and links in a chain of events related to the scientific method (7). First among these rules is that model selection and construction must be predicated on the best available known description of the real world entity to be modeled, and in this case, of cerebral AVMs. Following this closely is the mandatory requirement of preliminary model validation by testing its accuracy; this is achieved by matching the behavior of the model with known scientific observations about this entity. The sheep AVM model presented fails on both counts in relation to both its morphologic characteristics and its angiographic hemodynamic behavior.

In terms of basic morphologic characteristics, the sheep model bears little resemblance to cerebral AVMs in two ways. First, the overall shape/contour of the so-called “nidus” is very unlike that of cerebral AVMs. The sheep “nidus” is composed of one rete mirabile, ie a distinct vascular network connected to the contralateral rete (another distinct structure) by a few midline interretial vessels. This connection may even be a single vessel, as seen in a plastic cast presented by Daniel et al (8). Such a distinctly discontinuous appearance is almost never seen in conventional nidi of cerebral AVMs, unless some other event has occurred to induce this odd appearance, such as a hemorrhagic episode with an unusual subsequent partial thrombosis or a grossly unsatisfactory partial radiosurgical obliteration of the middle portion of a large nidus. Qian et al have justified this appearance in their “model” by stating that the sheep “actually has a double AVM nidus” more characteristic of multiple AVMs (mAVMs). This is unsatisfactory on two counts: 1) mAVMs constitute less than 2% of all AVMs (9), and therefore a model of this specific entity is of little use because it is nonrepresentative of the great majority of cerebral AVMs; and 2) the great majority of mAVMs are distant from each other and are fed and drained by separate arteries and veins. To my knowledge, the presence of two communicating mAVMs might have existed (because they shared a common draining vein) in one patient reported in the literature by Voigt et al (9, 10). This extremely rare configuration is what is being modeled in the sheep by Qian et al. As well, there are negative practical implications of this “double nidus” in the sheep model, because the goal of any attempted simulation of transarterial (3, 4) or transvenous (2) embolization of the nidus should be to target the entire nidus. That is, it should be necessary to demonstrate with whatever embolic agent being used, that it should at least be capable of reaching the rete contralateral to the site of injection. With such a degree of physical separation of the retia in the “double nidus” of the sheep, this is unlikely. Finally with regard to morphologic characteristics, could this sheep model be that of a single cerebral AVM with multiple true “anatomic” or “structural” nidus compartments (“sectorization” of the nidus, as coined by Pertuiset et al [11]) represented by the two barely connected retia mirabilia? The answer should be negative, because in these uncommon cerebral AVMs true “hemodynamic” nidus compartmentalization is much more readily recognized in AVMs, each compartment is fed and drained by its own artery and vein, respectively, and therefore, it is quite unlike what is found in the sheep “AVM model.” None of the above issues are relevant to the nidus of the swine AVM model because of its extensive midline interretial connections that join bilateral retia mirabilia into an overall single vascular network representative of the nidi of most cerebral AVMs.

A second minor shortcoming of the sheep “AVM model” applies to the feeding arteries. The great majority of cerebral AVMs are fed by terminal or direct feeders. Instead, the sheep “AVM” is fed only by two or more “en passage” type of feeders. Conversely, the most important feeder to the swine AVM nidus is an easily accessible terminal type of feeder represented by the ascending pharyngeal artery. In the swine model, the presence of the ramus anastomoticus and the arteria anastomotica, acting as additional en passage feeders, completes the appearance of appropriately simulated feeders of both types. All three swine feeders can be microcatheterized during realistic interventional/embolization simulations (3). Thus, in terms of strict adherence to the rules of modeling, the feeding arterial system to the swine AVM model is also much more representative of cerebral AVMs than that found in sheep.

In relation to the angiographic appearance of the sheep “AVM model” and the hemodynamic inferences that can be made, one striking observation is noted: the arterial feeders, nidus, and draining veins are only faintly opacified precisely down to the level of the origin of the occipital artery ipsilateral to the fistula. Indeed, on the lateral image provided, there is no discernible contrast medium in the right external carotid artery (the simulated draining vein) above its junction with a markedly opacified right occipital artery. These observations indicate an overall poor flow through that part of the vascular circuit, which represents the “AVM model.” In comparison, both occipital arteries are very well opacified, indicating that most of the flow diversion from the contralateral side of the neck to the fistula occurs at the level of the occipital arteries, and not at the level of both retia. Hemodynamically speaking, this is a point of critical importance, because cerebral AVMs represent a short-circuit for blood flow between arteries and veins. Instead, the neck anastomoses between both occipital arteries in the sheep would appear to offer a pathway of less resistance than that of the “AVM.” This hemodynamic inadequacy of the sheep “AVM” is not surprising when the shape of the “nidus” is taken into account. The almost complete separation of bilateral retia and the presence of only one/few vessels between them means that the resistance to flow from one rete to the other would be considerably higher than that found in a single-entity nidus, as seen in the swine AVM model. This principle may be better appreciated by analogy with elementary laws of hydrology or electricity. Indirect support for this comes also through knowledge that in the normally fast-flowing swine AVM model, if the resistance is increased in any part of the AVM circuit (eg, by partially embolizing the nidus, or by inducing vasospasm in the main feeding artery or draining vein), then flow diversion to the neck fistula occurs through pathways parallel with both retia at the level of other arteries in the neck. This typically occurs through occipital artery to occipital artery anastomoses, giving the exact angiographic appearance of the sheep “AVM model.” Once again, none of these issues relate to the swine AVM model in which the low-resistance single-entity nidus permits preferential fast shunting blood flow and, hence, a realistic angiographic appearance and a representative transnidal drop in flow and pressure parameters.

In view of the morphologic and hemodynamic shortcomings described, the sheep would seem to be an inappropriate laboratory animal for the purpose of AVM model construction. Owing to its obviously simple construction technique and its underlying structure and hemodynamics, the carotid-jugular fistula type AVM model in swine is considered a very suitable experimental simulator of cerebral AVMs, especially with regard to interventional neurovascular research, ie, the performance and study of experimental embolotherapy and the development of new embolization techniques. If the swine is unavailable in a particular laboratory, caution should be exercised in referring to an alternative morphologically and hemodynamically unacceptable construct as an “AVM model.” The authors should be encouraged to intensify their efforts, through strict attention to detail and adherence to the principles of experimental modeling (7), in seeking out new and improved in vivo AVM modeling techniques other than in sheep. For example, one proposal for an appropriate alternative laboratory animal to use for AVM model construction might be the goat. This animal is only 15% more expensive than either swine or sheep. Vascular anatomy of the goat head and neck is practically identical to that of the sheep, with the most important exception that both retia are connected by extensive midline interretial vessels similar to the swine and unlike the sheep.