Pathophysiology of carotid-cavernous fistulas in vascular Ehlers-Danlos syndrome: a retrospective cohort and comprehensive review

Vascular Ehlers-Danlos syndrome (vEDS, OMIM #130050) is a rare inherited connective tissue disorder with a prevalence estimated at 1/150000. It is an autosomal dominant transmitted disorder caused by diverse mutations in the COL3A1 gene which encodes the pro-alpha1 chain of type III procollagen [1‐3]. Vascular EDS is characterized by connective tissue fragility leading to life-threatening complications in seemingly healthy young adults. Easy bruising, thin and translucent skin, acrogeria and characteristic facial features may be a hint for clinicians. Intestinal and uterine fragility are classical features of vEDS, but generalized arterial fragility dominates the clinical picture with spontaneous dissections, aneurysms and arterial ruptures [2].

It is therefore unsurprising that vEDS has been associated with the development of spontaneous direct carotid-cavernous fistula (sCCF). Carotid-cavernous fistula (CCF) is an abnormal communication between the high-pressure carotid arterial system and the low-pressure cavernous venous system (CS). In direct CCF, internal carotid artery (ICA) wall disruption allows high-pressure blood to move into the CS, resulting in short-circuiting of the ICA arterial blood into the venous system of the CS. The most common symptoms are swelling of the eye, chemosis, pain, ophthalmoplegia and bruit [4]. Direct sCCF is rare in the general population as most direct fistulas are posttraumatic but seems to be more common in vEDS patients and especially in females.

We report in this article a series of 13 molecularly proven vEDS patients diagnosed with direct sCCF along with a review on literature regarding direct sCCF occurring in vEDS patients. The possible mechanisms of occurrence in this subset of patients are discussed.

Classification of CCF is based on aetiology (traumatic versus spontaneous), blood flow velocity (high versus low) and anatomy (direct versus dural, or ICA versus ECA). According to the Barrow classification, there are 4 types of CCF [55]. Type A or direct CCF is a high-flow direct shunt due to a posttraumatic or ruptured aneurysm which results in short-circuiting of the ICA arterial blood into the venous system of the CS. Type B, C and D are low-flow lesions and grouped under the common definition of dural or indirect CCF. In the general population, direct CCF represents 75% of all CCF and is mostly posttraumatic (0.2–0.3% of craniofacial trauma) [56] or associated with surgical complications, while spontaneous dural (indirect) low-flow CCF are mostly found in postmenopausal hypertensive females [57]. Direct sCCF only accounts for 10% of all spontaneous CCF [58]. The intracavernous ICA is prone to spontaneous shunting due to the amount and direction of ligaments in relation to the adventitia, the conformation and mobility of the siphon (horizontal segment), and the inextensibility of the dura [59, 60]. Direct sCCF can be found in patients with intracavernous ICA aneurysms, but only 2 to 9% of intracavernous ICA aneurysms are complicated with a direct CCF [61]. The rupture of a persistent trigeminal artery-associated aneurysm has also been associated with the development of CCF [62, 63]. Though direct sCCF seems to be rare in healthy individuals, it appears to be a significant complication in patients with collagenopathies such as vEDS [40, 46, 64]. Spontaneous CCF does not seem to associate with other rare vascular diseases such as Marfan disease or Loeys-Dietz syndrome [65].

Vascular EDS is a rare heritable disease mainly characterized by vascular manifestations such as rupture, dissection or aneurysm affecting large or medium-sized arteries. Cervicocerebral arteries can be affected with disorders such as intracranial aneurysm, carotid artery dissection or direct CCF [64‐66]. In our tertiary referral centre, subjects with direct CCF represent 9.8% of vEDS patients and all fistulas were spontaneous. Posttraumatic direct CCF seems to be rare in vEDS patients and occurs mostly in males [7, 11, 14].

The median age of our patients at direct sCCF diagnosis, 33.0 years, IQR [26.0–39.5], is quite similar to previous data reported in vEDS population [46, 54, 64], and considerably lower than patients free from vEDS who experience CCF [55]. These results are also concordant with the median age (31.0 years, IQR [24.0–39.0]) of the patients identified in the literature review. In our series, it is remarkable that 9 patients (69.2%) experienced a direct sCCF as their first vascular event, prompting further investigations for vEDS molecular diagnosis in 4 patients only. Notably, direct sCCF has recently been included in the 2017 classification for vEDS as a major diagnosis criterion [2]. As direct sCCF in young patients seems to be pathognomonic of vEDS, genetic testing of the COL3A1 gene should be mandatory in such patients.

The type of pathogenic variant identified within the COL3A1 gene encoding the pro-alpha1 chain of type III procollagen in vEDS patients who experienced sCCF is another interesting finding. The natural course of vEDS depends on the type of COL3A1 variant, and there are more severe clinical and phenotypical presentations in Glycine missense and splice-site variants than in variants leading to haploinsufficiency [67]. Our 13 patients presented either with heterozygous Glycine missense or splice-site variants of exons encoding a triple helix sequence. The literature review identified 9 patients with available pathogenic variants within the COL3A1 gene, and all were unsurprisingly either Glycine missense [5, 8, 18, 51, 53] or splice-site variants [8, 29, 45, 51]. Therefore, it seems reasonable to hypothesize there might be a strong association between these variants and the occurrence of sCCF, while haploinsufficiency does not associate with fistulas. However, the total number of pathogenic variants identified in our patients or reported in the literature (22/97, 22.7%) is too small to confirm this genotype-phenotype correlation. As the genetic test is now mandatory to confirm clinical diagnosis, the presence of pathogenic COL3A1 variants should be systematically reported in future articles related to sCCF.

In vEDS patients, spontaneous rupture of a tortuous ICA emerging from the petrous bone into the CS can result in a direct fistula with [8, 15, 19] or without [6, 10, 19, 26, 39, 49] a previous known intracranial ICA aneurysm. Another mechanism could be a spontaneous dissection of the ICA with pseudoaneurysm formation [68]. Spontaneous ICA dissection leading to direct sCCF is described in 5 of our patients (cases 1, 6, 7, 11 and 12) and in the literature review [33, 46, 52]. Intracranial arteries have a histological constitution that reportedly differs from diameter-matched extracranial arteries. They have a decreased amount of elastic fibers in the media and adventitia layers, along with thinner collagen fibers and no external elastic lamina [69]. It is well known that vEDS leads to vascular complications due to structural weakness of the artery wall secondary to type III collagen defects. Therefore, arterial ruptures can also occur on previously non-aneurysmal arteries as we found in 3 patients (cases 5, 9 and 13). Notably, direct sCCF in vEDS patients do not systematically result from aneurysm rupture or dissection of the cavernous ICA and cannot be exclusively described as a complication of an aneurysm or dissection of the cavernous ICA. An explanation for such spontaneous ruptures in seemingly healthy arteries can be found in Boutouyrie et al. [70] who demonstrated an increased carotid circumferential wall stress and a higher carotid distensibility in vEDS patients compared with healthy individuals. Moreover, vEDS patients are known to have lower IMT [70] and lower diastolic-systolic arterial stiffening than healthy control individuals [71]. Impairment of type III collagen generates a lower carotid stiffness and an abnormally low IMT, leading to higher wall stress and risk of arterial rupture. As direct sCCF is probably due to a sudden increase of the ICA intraluminal pressure [72], the previous findings explain why vEDS are at high risk for direct sCCF.

The male to female ratios in our study (0.08) and literature review (0.15) stress the fact that female vEDS patients seem to be more at risk than males for direct sCCF. In comparison, there was a 0.79 male to female ratio in our non-sCCF vEDS patients. In addition to potential sex differences in IMT, sex differences in ICA anatomy could account for a higher prevalence of fistulas in females. Though gathering evidence in vEDS patients is challenging, inferring results from the general population is a first step in understanding the female predominance for sCCF in vEDS. Differences between males and females in various segments of the vascular system have been reported [73‐75], as well as in intracranial arterial bifurcations and more specifically of the ICA [76, 77], suggesting the involvement of a physiological or an anatomical factor. First of all, sex differences in ICA can be found in the diameter of the parent ICA and its largest and smallest branches, with a significantly larger diameter in male vessels than in female vessels [77, 78]. Secondly, the study of ICA maximum wall shear stress shows a 50% higher stress in the female bifurcation compared with the male bifurcation [77]. As females have smaller ICA, their vessel walls are exposed to higher blood flow velocity, and this shear stress is reflected through a higher pressure drop in the female bifurcation compared with the male bifurcation. Vessel morphology and more specifically vessel curvature is also an interesting factor to take into account to determine hemodynamic patterns. Healthy females reportedly have higher mean curvature of their intracranial ICA than males (P 0.04) with a general tendency to a more curved ICA [79]. Only 20.5% of males have their peak curvature located within the CS, in comparison with as much as 45.1% of females. Therefore, we hypothesize that anatomical differences account for an irregular blood-flow pattern—which in turn accounts for the alteration of the arterial wall—due to higher shear forces in female intracranial ICA, and lead to more ICA ruptures and fistulas in female vEDS patients. Finally, evidence regarding lower IMT in healthy females compared to males can be found in the literature [80‐82]: sex is significantly correlated with carotid artery IMT (P 0.001), as females have significantly lower IMT than age matched males [82]. However, due to the rarity of vEDS, there is no available evidence in the literature regarding IMT differences between male and female vEDS patients. Moreover, as carotid IMT is measured on the extracranial part of the carotid artery 1 cm beneath the bifurcation, the IMT profile of proximal ICA might not be translatable into structural fragility of the cavernous ICA. Nonetheless, assuming that IMT difference between healthy male and female individuals might also be found in vEDS males and females could account for the overwhelming prevalence of direct sCCF in female vEDS patients.

Another appealing and complementary hypothesis for the frequent occurrence of direct sCCF in vEDS females could be the decrease of wall thickness due to sex hormones, and in particular the role of oestrogens on arterial wall remodelling. An arteriovenous fistula model in rats reveals a significant decrease in wall thickness of the proximal and distal aorta in castrated male rats receiving oestrogens [83]. These results are consistent with the findings of a previous study: neointima formation is reduced by 60% in rats treated daily by oestrogens before and after balloon injury of carotid artery [84]. The role of oestrogens in spontaneous rupture of carotid-cavernous aneurysm has also been suspected during pregnancy in healthy females free from pregnancy-induced hypertension [85, 86]. We could not find evidence of oestrogen taking in oral-contraceptives for example in our female vEDS patients who experienced direct sCCF, but it is noteworthy that the median age of pregnancy in these patients is quite similar to the median age of occurrence of their sCCF. These results lead us to think that collagen fragility, IMT and anatomy differences are not the only factors predisposing to the rupture of arteries in vEDS females.

Finally, direct sCCF can be the first symptomatic event in a previously non-diagnosed vEDS individual: in one third of our patients, the occurrence of a fistula led to the molecular diagnosis of vEDS. As sCCF seems to be highly suggestive and possibly pathognomonic of vEDS, type III collagen deficiency should be suspected in young patients and especially in females presenting with a direct sCCF.