Neurologic Complications in Hereditary Hemorrhagic Telangiectasia with Pulmonary Arteriovenous Malformations: A Systematic Review

ABSTRACT: Background: Hereditary hemorrhagic telangiectasia (HHT) is an autosomal dominant multi-organ condition occurring with a 1 in 3800 prevalence in Alberta. This genetic disorder leads to vascular malformations in different organs including the lungs and brain, commonly affecting pulmonary vasculature leading to pulmonary arteriovenous malformations (PAVMs). PAVMs lead to right-to-left shunts, which may be associated with neurologic complications. We aimed to evaluate and summarize the reported neurologic manifestations of individuals with HHT with pre-existing PAVMs. Methods: We performed a qualitative systematic review to determine available literature on neurological complications among patients with PAVMs and HHT. Published studies included observational studies, case studies, prospective studies, and cohort studies including search terms HHT, PAVMs, and various neurologic complications using MEDLINE and EMBASE. Results: A total of 449 manuscripts were extracted including some duplicates of titles, abstracts, and text which were screened. Following this, 23 publications were identified for inclusion in the analysis. Most were case reports (n = 15). PAVMs were addressed in all these articles in association with various neurological conditions ranging from cerebral abscess, ischemic stroke, hemorrhagic stroke, embolic stroke, and migraines. Conclusion: Although HHT patients with PAVMs are at risk for a variety of neurological complications compared to those without PAVMs, the quality and volume of evidence characterizing this association is low. Individuals with PAVMs have a high prevalence of neurological manifestations such as cerebral abscess, transient ischemic attack, cerebral embolism, hemorrhage, and stroke. Mitigating stroke risk by implementing proper standardized screening techniques for PAVMs is invaluable in preventing increased mortality.


Introduction
Hereditary hemorrhagic telangiectasia (HHT), previously known as Rendu-Osler-Weber disease, is an autosomal dominant genetic disorder with a prevalence of 1:3800 in the province of Alberta. 1,2,3 HHT is characterized by vascular malformations which can affect different organs, including the lungs and brain leading to what is known as pulmonary arteriovenous malformations (PAVMs) and cerebral arteriovenous malformations (CAVMS), respectively. Both can lead to stroke. The estimated annual cost of stroke in Canada is $2.8 billion according to a 2012 article published in the Canadian Journal of Neurological Sciences, with other estimates indicating upwards of $3.6 billion. 4,5 A better understanding of risk factors for HHT patients who may develop stroke is therefore important. Further, there are other associated neurological manifestations in HHT patients with PAVMs that have been noted in the past such as cerebral abscess, migraine, and transient ischemic attack (TIA).
Inherent in this genetic disorder, patients with HHT often exhibit bleeding from both epistaxis and / or gastrointestinal tract along with liver involvement which can also commonly occur. 1,2,6 Six different genes have been identified, where pathogenic mutations can result in HHT, most of which involve the endoglin (ENG) and ACVRL genes. The majority of HHT patients have mutations in the endoglin and activin type-II-like receptor kinase 1 (ACVRL1) gene causing HHT type 1 and HHT type 2, respectively. 7 Molecular genetic testing is often sought for patients who are suspected of having HHT and can reveal the specific mutation (i.e. ENG, ALK1 mutation, SMAD4 mutation).
Patients diagnosed with HHT are categorized as having "definite," "possible," or "suspected" HHT based on the clinical Curaçao criteria. These categories coordinate with the following clinical findings: epistaxis, (ii) telangiectasia (lips, oral cavity, fingers, nose), (iii) visceral AVMs (including the lungs, brain, and liver), and (iv) first-degree family history of HHT. 2,8 Patients with three or four of the criteria present are said to have definite HHT, while having two of the criteria constitutes possible HHT, and one unlikely.

Pulmonary Arteriovenous Malformations
A PAVM is an abnormal connection between a pulmonary artery and a pulmonary vein, where there is low resistance and high blood flow which bypasses the normal capillary bed. 7 Such malformed connections may lead to what is known as an intrapulmonary right-to-left shunt which can lead to other life-threatening complications. 7,9,10,11,12 The etiology of such is due to the lack of proper capillary network between the pulmonary arterial branch and pulmonary venous tributary, whereby blood bypasses the network which serves as functions of gas exchange and filtration. As such, PAVMs left untreated can lead to adverse conditions such as cerebral abscess, stroke, paradoxical emboli, and hypoxemia. 12 PAVMs are estimated to be present in 15-50% of patients with HHT, and approximately 80-90% of patients with PAVMs eventually present with HHT. 7,9,10 They are more commonly found in patients with HHT1 associated with the ENG mutation, though reasons for this propensity remain unknown. 1,11 Patients who do have PAVMs have a 53-70% likelihood of them being in the lower lobes of the lungs. 11 PAVMs can be classified as either simple (80-90%) or complex. 11 Simple PAVMs consist of a single feeding artery and a single draining vein, whereas the complex PAVMs may consist of multiple feeding arteries or draining veins. 11 Patients with PAVMs often experience no symptoms, whereby the PAVMs can go unnoticed until further complications arise. International guidelines for the diagnosis and management of HHT underscore the importance of screening for PAVMs while they may be in a treatable phase, prior to becoming life-threatening. 9,13 Complications from untreated PAVMs include associations with stroke, TIA, cerebral abscess, hemoptysis, and hemothorax. 9 Due to the lack of filtration that normally occurs in the capillary beds of connecting arterial and venous connections in the lungs, patients with PAVMs are at high risk of developing paradoxical emboli, or a passage of a clot from a vein to an artery, which can develop into embolic strokes or cerebral abscesses. 14

Diagnosis of PAVM
The preferred method of screening for PAVMs, with up to 98.6% sensitivity and relatively low risk, is transthoracic contrast echocardiography (TTCE), which is often referred to as "bubble echocardiography." The combination of TTCE and chest X-Ray (CXR) provides a 100% pick-up rate. 7,9,13,14,15 The process involves the injection of an agitated saline solution intravenously and observing for microbubbles (contrast) in the left cardiac cavityindicative of right-to-left pulmonary shunt. 14 The approximate size of the shunt can be semi-quantified based on the opacification of the left ventricle. 14 The values are (a) Grade 0: no bubbles; (b) Grade 1: occasional filling with less than 30 bubbles; (c) Grade 3: moderate filling with 30-100 bubbles; and (d) Grade 4: complete opacification with more than 100 bubbles. 14 International guidelines for the diagnosis and management of HHT recommend that TTCE be used as the initial screening test for PAVMs and that all patients with either possible or confirmed HHT be screened for PAVMs. 9,13 There are other methods of screening for PAVMs such as chest radiography, pulse oximetry, arterial blood gas analysis, and CT chest. 15

Treatments for PAVM
Surgery was the only method of treatment for PAVMs between the 1940s and late 1970s, where either ligation, excision, segmentectomy, lobectomy, or pneumonectomy were the standard procedures. 11 In 1977, Porstmann first reported using embolotherapy for treatment of a PAVM. 11 The procedure consists of the obstruction of the feeding artery by means of inserting steel coils, detachable balloons, polyvinyl alcohols among others. 11 Few studies have evaluated secondary neurologic complications in HHT patients with PAVMs, in line with the rarity of the disease. We therefore conducted a systematic review to evaluate and summarize neurologic complications in HHT patients who have PAVMs versus those without PAVMs and to consolidate the available literature on the subject.

Methods
A systematic review was conducted, and search terms and strategy were defined to include records pertaining to HHT and PAVMs as well as any subset of neurologic conditions. A systematic approach was outlined in order to consolidate literature from the following databases: MEDLINE (1946 to the present, which also includes Epub Ahead of Print, In-Process & Other Non-Indexed Citations, and Ovid MEDLINE(R) Daily) and EMBASE. The systematic review was then updated with the same search terms both in EMBASE and MEDLINE and included any additional manuscripts as of June 2021. Any discrepancies were resolved by reviewer discussion. The search strategy was reviewed with the assistance of a health sciences librarian. Manuscripts were then exported into RefWorks where Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methods were used to screen records for inclusion in this study ( Figure 1).
Inclusion Criteria: We included published observational studies, case studies, prospective studies, retrospective studies, and cohort studies that looked at HHT patients who have PAVMs further complicated with secondary neurologic complications.
Exclusion Criteria: We excluded studies that were not written in English, studies which only included abstracts, review articles, and conference or poster presentations. Studies in which patients had PAVMs but not HHT were also excluded.
Data Analysis: A qualitative analysis was performed on the included manuscripts and summarized in tabulated format (Tables 2 and 3). We rated quality of included studies and risk of bias using the GRADE framework. 16,17

Results
We secured a total of 449 records from MEDLINE and EMBASE, which were screened based on title and abstract. Search strategy including detailed search terms was saved and included in Table 1. Of the total n = 449 records, n = 117 were identified as duplicate articles from MEDLINE and EMBASE and were subsequently excluded. The first round of screening involved title and abstract review, where n = 281 were excluded due to not meeting the inclusion criteria. After text review, a total of n = 23 articles were included in the study, and results and major findings were summarized in Tables 2 and 3.
Of the 23 articles included, 15 were case studies, and 8 were non-case studies. PAVMs were addressed in all these articles in association with various neurological conditions ranging from cerebral abscesses, ischemic stroke, TIA, migraine, CAVM and seizure among others.
In summary, there were a total of n = 15 case studies with a total of n = 20 patients who presented with HHT, PAVM, and some type of neurologic complication. A total of 20 patients resulted from two of the case studies reporting multiple patient cases each, which met our inclusion criteria (Table 3). Ischemic stroke (n = 13) was by far the most common neurologic complication. One patient was described to have a stroke, with no detail as to what subtype. Patients also presented with TIA (n = 2), cerebral abscess (n = 2), with migraine (n = 1), CAVM (n = 2), seizure (n = 2), epilepsy (n = 1), nocturnal headache (n = 1), and Todd's Palsy (n = 1). Study quality was rated as very low for case series, and therefore, risk of bias was high.
There were a total of n = 8 non-case studies ( Table 2). Of these, they were cross-sectional studies (n = 3), prospective cohort studies (n = 1), and retrospective cohort studies (n = 4). These eight studies are summarized below by category of methodology. Study quality was rated as low for cross-sectional studies and retrospective cohort studies, and therefore, risk of bias was high. Study quality was rated as moderate for the prospective cohort study with moderate risk of bias.

Angriman et al., 2014 18
This cross-sectional study was based on data from 108 patients between 2010 and 2012 accessed through the Institutional records of HHT in Argentina to evaluate for complications related to PAVMs. A primary finding of this study emphasized the disparity between the prevalence of embolic complications (ischemic stroke can be embolic or thrombotic) and between patients with significant PAVMs in comparison to the control group, without significant PAVM. A significant PAVM was defined as those with contrast echocardiography of grade 2 or greater, a PAVM feeding artery larger than 3 mm, or the patient having had treatment for PAVM in the past. Individuals diagnosed with significant PAVMs indicated a higher risk of developing embolic complications (34.3%) when evaluated with the control group, 9.4% (p = 0.006), which was attributed to the potential influence of sex, age, anemia, and age of onset of the PAVM. The most common embolic complication was stroke. Additionally, the study highlighted the increased risk of secondary embolic complications due to PAVMs. Of the 35 patients with significant PAVMs, 9 developed further stroke/TIA, 2 had brain abscesses, and 12 had embolic complications.
This study demonstrated that patients with significant PAVMs are at a higher risk of developing embolic complications. Further, this study also noted that patients with PAVMs were significantly younger (p = 0.01) with more pulmonary symptoms (60% vs. 11%, p < 0.001) and embolic complications as noted earlier.

Moussouttas et al., 2000 19
This cross-sectional study aimed to evaluate the risk of neurological manifestations of 75 patients with PAVMs. Prevalence of brain abscess and seizure was twice as high in patience with multiple PAVMs as compared to those with only one PAVM, though these differences were not statistically significant. Most neurological 11. cerebrovascular accident/ 12. (brain ischaemic attack or CAVM or cerebral arteriovenous malformation or brain arteriovenous malformation or brain infarct* or brain attack or cerebrovascular accident* or brain accident*  -15 presented with brain abscess, 26 presented with ischemic stroke -Larger feeding artery significantly associated with more ischemic strokes -Largest feeding artery diameter of PAVM significantly associated with frequency of ischemic stroke, and the # of PAVMs with the risk of brain abscess 19 Moussouttas   Infarct in the left middle cerebellar artery territory (ischemic stroke), and small chronic cerebellar infarcts, epilepsy

ACVRL1
-Case of undiagnosed pulmonary AVM leading to recurrent paradoxical emboli to the brain -Case underscores the need to consider PAVM in the differential diagnosis for cryptogenic stroke -CT scans should be done 6-12 months after embolization then every 3 years thereafter No information 32 Felix case study highlights the importance of identifying pts with PAVMs before they lead to neurologic complications such as stroke iron deficiency is a factor associated with ischemic stroke in patients with HHT No information 35 Kawaguchi, 1990 1 52 (m) -Multiple PAVMs -Multiple CAVMs, brain abscess No information -At age 49, pt underwent drainage of brain abscess, at which time a CAVM was detected near the brain abscess -Development of CAVMs and PAVMs "may" have been unrelated neurological complications secondary to PACMs include 1) hypoxic encephalopathy due to a reduction in the blood oxygen concentration caused by the pulmonary arteriovenous shunt; 2) cerebral thrombosis due to secondary polycythemia; 3) paradoxical embolus; and 4) brain abscess

SMAD4
-Pt presenting with PAVM (asymptomatic) with feeding artery of 2.4 mm. One month later pt. had infarction as detected through MRI -Antiplatelet therapy was started afterwards and no stroke recurrence or complications were observed during 7 years of follow-up

ENG
-Pt presented with stroke. Paradoxical embolism thought to be the cause of stroke -After embolization (coil) of PAVM, no complications occurred in the following two years -Long-term follow-up of pts after embolization is critical -Importance of treating PAVMs with transcatheter embolotherapy on a case-by-case basisregardless of feeding artery diameter No information 41

Velthuis et al., 2013 20
This cross-sectional study compared cerebral complications between patients with HHT and differing pulmonary shunt grading associated with PAVMs. The findings suggest that there was a higher prevalence of cerebral complications (either ischemic stroke, TIA, or brain abscess) among patients with a pulmonary right-to-left shunt (8.3%), as compared to patients without a pulmonary right-to-left shunt (1.4%; p < 0.001). Further to this, the percentage of patients with cerebral manifestations differed significantly depending on the grade of the pulmonary shunt (pulmonary shunt grade 0, 1, 2, and 3 corresponding to 1.4%, 0.4%, 6.5%, and 20.9%, respectively). This finding differs from Shovlin, 2014 wherein shunt severity was not found to have an impact on neurologic complications. Findings from this study indicate that shunt size needs to be taken into consideration when determining potential development of neurological complications.

Boother et al., 2017 21
This prospective cohort study analyzed the data from a total of 445 patients, all of whom had CT-confirmed PAVMs. They evaluated whether previously recommended measures (including judicious dental hygiene, antibiotic prophylaxis prior to dental and surgical procedures, and embolization of asymptomatic PAVMs) had led to reduced morbidity from cerebral abscess. Cerebral abscess poses a serious risk for patients as it is often associated with right-to-left shunting, where infected material can enter the brain. Previous literature analyzed in this study suggested that cerebral abscess affects 7-9% of PAVM patients, due to lack of proper filtration through the normal diameter (7-10 μm). It was noted that 8.3% of these patients experienced a cerebral abscess (confirmed through neurosurgical drainage). Additionally, of these patients 37% reported having untreated infections in their mouth (dental infection) when assigned to their associated groups based on presentation data where case notes were documented to capture patient-specific variables. This provides insight into the preventative measure, of dental hygiene, and its significance as an effective precautionary method of preventing cerebral abscesses. Males with PAVMs were found to have an increased risk of developing cerebral abscess, estimated as 2.63-fold higher, indicating a greater than double assessment of PAVM diagnosis in males compared to females. Overall, the role of PAVMs on cerebral abscess had not yet been defined; however, the importance of screening is stressed for patients who have experienced cerebral abscess. Further research needs to be evaluated to analyze a broader range of preventative measures to suggest more effective mechanisms to reduce morbidity from PAVMs.

Etievant et al., 2018 22
This retrospective cohort study in HHT populations determined the associations of PAVMs with cerebral abscess and ischemic strokes. Of the 170 patients included in the study, 8.8% presented with cerebral abscess. Impressively, these patients with cerebral abscess were exclusively in the group who had non-treated PAVMs, highlighting the importance of proper screening and treatment. Further, 16.5% of the total patients had ischemic stroke, which is marginally high compared to expected. Most of these stroke patients were also non-treated for PAVMs and further had significantly larger feeding artery (4.9 mm versus 3.2 mm; p = 0.009). Significant correlations were found between the prevalence of neurological complications and CT imaging of PAVMs. This study further subclassified the types of PAVMs and found that there was an increased risk of cerebral abscess in those with multiple, diffuse, or disseminated PAVM. Along with cerebral abscesses, the study highlighted the statistically significant relationship between diameter of the largest feeding artery of the PAVM and frequency of ischemic stroke. The authors urge clinical practice to treat PAVMs by embolotherapy at the soonest notification of cerebral complications in order to decrease further risk. Causality amongst these fascinating correlations remains speculative or unaddressed.

Post et al., 2005 23
A total of 538 HHT patients were included in this retrospective cohort study, and the findings suggest that migraine is more common in HHT patients with PAVM, as well as more common in HHT patients as compared to the general population. Migraines occur in 10-12% of the population, and studies show a higher prevalence in HHT patients. The current study indicated that migraine was present in 88 of the 538 (16.4%) patients who had HHT. Migraine prevalence was higher in HHT patients who had PAVM (21.2%) as opposed to those without PAVM (13%) and interestingly, significantly more prevalent in women with PAVM than men (65.4% vs 34.6%; p = 0.009). Further to migraines, the percentage of patients developing other neurological conditions with PAVMs as opposed to without PAVMs respectively was as follows: CAVMs (13% vs 4.2%), brain abscess (19% vs 0%), TIA (18% vs 4%), and brain infarctions (35% vs 4%) all with p < 0.00. The study further attempted at describing possible reasons for the linkage between migraines and PAVMs. One such speculation considers that due to the possibility that HHT and some subtypes of migraines are autosomal dominant disorders, a genetic reason may be involved in providing an explanation. However, a more plausible explanation is the possibility of "trigger" substances passing through the right-to-left shunt and inducing cerebral instabilities which can lead to migraines. This study postulates that PAVMs play an important role in the pathogenesis of migraines, and further work is needed to develop this understanding.

Shovlin et al., 2008 24
In this retrospective cohort study, 323 individuals with PAVMs were included where 74 (33.8%) with PAVMs had a brain abscess or stroke. Interestingly, univariate analysis found no significant relationship between brain abscess and the markers of PAVM severity. A large proportion of those with brain abscess and PAVMs had events which are known to have associations with bacteremia in the weeks prior to the cerebral abscess (three with scale and polish procedures, and four with occlusive braces and fillings). This study also found no significant relationship between PAVM parameters and ischemic stroke incidents.

Shovlin et al., 2014 25
In another retrospective cohort study with 497 individuals, they were able to identify that the strongest risk factor for stroke in patients with HHT and PAVMs was iron deficiency rather than the severity of the right-to-left shunt. No statistically significant association was found between PAVM and stroke when looked at in isolation. This is an important finding that the authors hypothesize is related to a model which suggests that iron deficiency is associated with platelet aggregation. 26 This needs to be further studied to better understand the physiological underpinnings.

Discussion
The aim of our systematic review was to consider HHT patients and identify their risk of developing neurologic complications in those who have PAVMs as compared to those without PAVMs. This is the first comprehensive review of this topic in an HHT population to our knowledge. The findings of this review suggest that HHT patients with PAVMs have a high prevalence of neurological manifestations such as cerebral abscess, TIAs, cerebral embolism, hemorrhage, and stroke. Strokes were highlighted most often in literature analyzed within this review; however, a positive correlation between PAVMs and the risk of cerebral abscess was also indicated, with one study indicating males at a higher risk. 21 HHT patients with PAVMs are at risk of life-threatening complications and require proper screening and diagnosis. Review of the available literature indicates a clear need for HHT patients to be screened for PAVMs to prevent this risk. 27,28,29 Risk of cerebral complications which arise because of rightto-left shunting in patients with PAVMs remains. Interestingly, there was conflicting data regarding the risk of neurological complications based on the PAVM right-to-left shunt grade. Shovlin et al were able to demonstrate the risk factor for stroke in patients with HHT, and PAVM was more strongly associated with iron deficiency rather than the severity of the right-to-left shunt, which was in direct opposition to the findings of Velthuis et al who demonstrated a higher percentage of neurologic complications as the shunt grade increased from 0 to 3. 20,25 This was also demonstrated by Etievant et al, who showed a significant relationship between diameter of feeding artery and frequency of ischemic stroke. 22 Moussouttas et al demonstrated that brain abscess and seizure were two times as high among those with multiple PAVMs as compared to those with only one PAVM. 19 Shovlin et al hypothesize that iron deficiency among patients with PAVMs can lead to platelet aggregation thus increasing the risk for stroke and indeed demonstrated this. The lack of studies on this topic given the prevalence of disease indicates there is further research needed to better understand the physiologic underpinnings of risk of cerebral events, and further work is needed to better understand the role of iron among patients with HHT and PAVMs.
Neurologic manifestations can arise when bacteria or other larger fragments pass through a PAVM and into the brain causing a cerebral abscess. The pathophysiology of PAVMs is a cause of concern for the patient in many regards. The irregular arterial to venous connection of a PAVM may lead to right-to-left shunting leading to inadequate oxygenation of the surrounding tissue as well as improper filtration. Patients with PAVMs must be careful especially when receiving dental work, to reduce the risk of infections. Boother et al demonstrated 37% of those with cerebral abscess had untreated infections in their mouth. 21 Patients with PAVMs should be especially careful in such situations, and healthcare workers should be aware of these risks. Boother et al also described the bacterial isolates indicating that most cerebral abscess cultures were microaerophilic and anaerobic bacteria, and culture negative. 21 Etievant et al studied this relationship between PAVMs and cerebral abscess and found that the number of PAVMs strongly positively correlated with risk of cerebral abscess, and that the larger the feeding artery was, the more strongly related with risk of ischemic stroke. 22 Small sample size was noted in many reports due to HHT being a rare disease. One study included a large number of patients (n = 445) and found similar results with strong associations between PAVMs and cerebral abscesses. 21 Of these, 10 cases included PAVMs which were diagnosed because of a cerebral abscess. The study further identified that low oxygen saturation and intravenous iron may be associated risk factors involved in cerebral abscess. Through multiple regression analysis, association of cerebral abscess was found to be insignificant with number of PAVMs. The prospective cohort studies which are likely to provide the most accurate measures of prevalence quote a variable cerebral abscess risk from 8.3% to 33.8%. 21,24 There is need for better understanding of the pathophysiological mechanism for cause of cerebral abscess, especially in HHT patients who have PAVMs.
It is clear through this review that PAVMs and migraines are not reported commonly. Studies have found an increase in migraine frequency in cardiac right-to-left shunts; however, Post et al described this phenomenon in pulmonary right-to-left shunts. They were able to find a modest (21.2%) increase in the prevalence of migraine in PAVM patients than in non-PAVM patients (13.3%), p = 0.02. This is not necessarily unexpected, as a rightto-left shunt is often associated with undesired particles passing through. The study describes a more plausible possibility of the passing of a "trigger" substance through the right-to-left shunt, which may induce cerebral vascular instability. 23 Though this is plausible, there is need for further research for any definitive answers. Further, there are potential further research opportunities considering the correlation between the specific genetics an HHT patient with a PAVM has in relation to risk of migraine.
A recent study looked at the links between strokes and HHT in Alberta and found there to be a significantly higher risk than compared to the general global population (450 per 100,000 versus 260 per 100,000, respectively). 3 There may very well be an increased prevalence of stroke among those with PAVMs in this same HHT population. Our work underscores the need for appropriate screening for PAVMs within HHT Centres. The risk of cerebral events associated with PAVMs demands that this be done consistently to reduce such incidents. According to the most recent international guidelines for the diagnosis and management of HHT patients, there is strong emphasis on flagging pregnant women with PAVMs or CAVMs as being high risk for hemorrhagic and neurologic complications. 13,45 This also applies to those patients who have not yet been screened.
The strength of this narrative review is that it addresses an important gap in the literature in regard to the neurological manifestations of HHT patients with pulmonary AVMs and identifies areas for future work. The review is limited by the low volume of studies and the relatively low quality of evidence available. There is overall need for further research considering the prevalence of neurologic complications among the various types of HHT and genotype-phenotype correlations within this population. 1,11