Observation versus intervention for Borden type I intracranial dural arteriovenous fistula: A pooled analysis of 469 patients

Abstract

Background

While it is thought that Borden Type I intracranial dural arteriovenous fistula (dAVF) have a benign clinical course, their management remains controversial.

Methods

A comparative meta-analysis was completed to evaluate the outcomes of intervention verses observation of Borden Type I intracranial dAVF. Outcome measures included: grade progression, worsening symptoms, death due to dAVF, permanent complications other than death, functional independence (mRS 0–2), and rate of death combined with permanent complication, were evaluated. Risk differences (RD) were determined using a random effects model.

Results

Three comparative studies combined with the authors’ institutional experience resulted in a total of 469 patients, with 279 patients who underwent intervention and 190 who were observed. There was no significant difference in dAVF grade progression between the intervention and observation arms, 1.8% vs. 0.7%, respectively (RD: 0.01, 95% CI: −0.02 to 0.04, P = 0.49), or in symptom progression occurring in 31/279 (11.1%) intervention patients and 11/190 (5.8%) observation patients (RD: 0.03, CI: −0.02 to 0.09, P = 0.28). There was also no significant difference in functional independence on follow up. However, there was a significantly higher risk of dAVF related death, permanent complication from either intervention or dAVF related ICH or stroke in the intervention group (11/279, 3.9%) compared to the observation group (0/190, 0%) (RD: 0.04, CI: 0.1 to 0.06, P = 0.007).

Conclusion

Intervention of Borden Type I dAVF results in a higher risk of death or permanent complication, which should be strongly considered when deciding on management of these lesions.

Introduction

Intracranial dural arteriovenous fistula (dAVF) constitute 10–15% of all intracranial arteriovenous malformations. ¹ Approximately 33% of dAVF lack cortical venous drainage (CVD) on angiography and are low-grade according to the accepted Borden (Type I) and Cognard (Type I/IIa) grading scales.^2,3 The natural history of these lesions is widely considered benign, with an annual risk of intracerebral hemorrhage (ICH) of 0%–1.5% and an eventual upconversion to a higher grade between approximately 0%–2%.^4–6 Notwithstanding the benign natural history of these lesions, contemporary series detailing management report an intervention and treatment rate between 57%–67%.^6,7 Management of Borden Type I dAVFs remains controversial and it is unclear if observation or intervention better facilitates superior clinical outcomes in patients harboring these lesions. The aim of this study was to conduct a systematic literature review and meta-analysis to directly compare the outcomes between conservative observation and intervention of any kind regarding the management of low grade, Borden Type I dAVF.

Methods

Institutional experience

All dAVF were retrospectively identified from our single tertiary care institution. The diagnostic angiograms of each dAVF within our institution were initially reviewed by two authors and graded based on the Borden classification. Disparities in grading were settled by a third senior author with all disagreements settled by consensus of all three reviewers. Once Borden Type I dAVFs were identified, all clinical information was collected on the patients, including: age, gender, clinical symptoms, treatment, procedure complications, and clinical outcomes and follow up. The results from our institutional experience were subsequently integrated into the pooled quantitative synthesis and treated as a single center retrospective analysis. This study was approved by the Research Subjects Review Board from our institution.

Eligibility criteria and study selection

A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guided systematic literature review was completed to evaluate the clinical outcomes comparing intervention versus observation for Borden Type I dural arteriovenous fistula. For the sake of this study, dAVF reported as Cognard type I and IIa were treated as equivalent to Borden Type I. The MEDLINE and Cochrane Library databases were searched up to February of 2022 using the following key phrases: “cranial dural arteriovenous fistula” OR “intracranial dural arteriovenous fistula,” combined with either “Type 1”, “Type I,” “type II,” “type IIa”, “low grade”, “conversion”, or “without cortical venous drainage.” All studies were cross-referenced, and duplicates were removed. Subsequently, each study was screened for initial review by reading the title or abstract; if there was still question for possible inclusion, the full paper was read and evaluated. Only studies of comparative design that directly compared the outcomes of intervention and observation within the same study cohort were included in the analysis. After the initial screening process, all of the potentially included studies were more rigorously assessed for eventual inclusion in the final quantitative synthesis. Any disagreements on possible inclusion within the analysis were settled by consensus among the authors. Similarly, eventual data was extracted from primary papers independently by two authors, with disagreements settled by a third author. Within the “intervention” group, this included any endovascular embolization, radiosurgery, open surgery, or any combination of these. Randomized control trials, observational studies, and case-series written in the English language were included, while case-reports were excluded. Spinal dAVF were also excluded. For inclusion within the final quantitative synthesis, each study had to describe clinical outcomes following intervention of a Borden Type I intracranial dAVF with direct comparison to those that were observed. Moreover, sufficient detail regarding the primary outcome measures had to be explicated with abstractable data. Further details regarding outcome measures are described below. Cases of carotid-cavernous fistulas were excluded from the presented analysis.

Outcome measures and additional reported variables

Six primary outcome measures were investigated in order to evaluate the clinical outcomes of intervention versus observation for Borden Type I intracranial dAVF. These outcomes included: grade progression, worsening symptoms, death directly due to dAVF, intracerebral hemorrhage with a permanent complication, and death or permanent complication from either a dAVF related ICH or from an intervention. Grade progression was assessed and defined as any Borden Type I that was found to upconvert or upgrade to Type II or III, which had to be explicitly stated by the study. Worsening symptoms were also evaluated and had to be clearly defined for both arms of the included study, and had to either be characterized as the occurrence of a non-hemorrhagic neurological deficit (NHND) or worsening of venous-hyperdynamic symptoms (VHS). Deaths due to the dAVF were only included within the analysis if the death was specifically stated in the study to be secondary to the treated or untreated dAVF. Since death is a permanent complication, additional permanent complications (e.g., neurologic deficit) from dAVF related ICH or stroke, were separately considered and analyzed between the intervention and observation arms. Furthermore, the combined risk of death due to dAVF or having a permanent complication (other than death) due to either dAVF associated ICH or stroke, or the intervention to treat the dAVF was evaluated. Lastly, the rate of functional independence on follow up, defined as a score of 0–2 on the modified Rankin Scale (mRS), was evaluated between the two groups.

Several additional variables were investigated in this study. The presenting symptoms of patients with dAVF and the modality of intervention were collated and reported. Additionally, the procedure/intervention complication rate and the rate of permanent complication following intervention was assessed. The rate of spontaneous obliteration of dAVF or spontaneous resolution of symptoms was also investigated within the observation arm. The mode of intervention used was also reported as a rate based on the pooled cases for which they were reported. Similarly, frequency of reported presenting symptoms were also collected, and were not stratified between intervened and observed patients but was reported as a composite total.

Pooled quantitative synthesis and statistical analysis

Pooled Risk Differences (RD) with corresponding 95% confidence intervals (CI) were calculated using the Mantel-Haenzel method with a random-effects model for each of the primary outcome measures. All statistical analysis was completed using Cochrane Review Manager (RevMan; computer program; version 5.4). Pooled composite rates were calculated for variables that did not have a corresponding comparison within the opposite arm of the study. For all additional statistical analyses, a P value of ≤0.05 was used to indicate statistical significance.

Evaluation of interstudy heterogeneity

An assessment of interstudy heterogeneity was also completed using Cochran’s Q statistic and described using the I² measure and P-value. An I² of >30%, >50%, and >75% were used as benchmarks for moderate, substantial, and considerable heterogeneity, respectively, as dictated by the Cochrane Handbook for Systematic Reviews. A P-value of >0.10 was used to indicate statistically non-significant interstudy heterogeneity.

Results

Systematic literature review

Searching of the MEDLINE and Cochrane Library databases resulted in 423 studies that were reviewed for possible inclusion. After preliminary review, 415 of these studies were excluded based on evaluation of the title or abstract, and 8 were sought for retrieval for potential inclusion in the eventual quantitative synthesis. From these 8, one study was excluded because it did not sufficiently detail the primary outcomes measures in order to be included. Four studies were excluded because their data overlapped with other published studies. In all, 3 comparative studies were included that met the required aforementioned inclusion criteria (Figure 1). Of these 3 studies, 2 were multicenter retrospective cohort studies of prospectively maintained databases, and 1 was a single center retrospective analyses.

Figure 1.

PRISMA guided literature review flow diagram.

Institutional experience

From our single institution, 8 Borden Type 1 dAVFs were identified between 2016–2022. There was a male to female ratio of 1:1 with a mean age of 66 years old. Pulsatile tinnitus was the presenting symptom in 6/8 (75%) of cases, while 2/8 (25%) were discovered incidentally. Endovascular embolization was utilized as treatment in 7 of these patients and 1 patient elected for observation. From the 7 treated patients, 1 (14%) experienced progressive tinnitus despite intervention. No mortalities were found in either group, though, one patient in the intervention arm did experience a permanent cranial nerve VII palsy following intervention. A representative example of a low-grade dAVF from our institution is described in Figure 2. Further details of our experience can be seen in Table 1 and is included in the pooled analysis.

Figure 2.

Low-grade borden type I dural arteriovenous fistula. (a) 77M presented with progressive tinnitus and MRA illustrating arterialization of the right sigmoid sinus. (b) lateral subtracted right external carotid artery angiogram illustrating Borden I, Cognard I fistula centered on the transverse-sigmoid sinus junction. (c) unsubtracted lateral x-ray illustrating onyx cast within pedicles of the middle meningeal and occipital arteries (d) one-year follow-up lateral ECA DSA showing regression of hypertrophied occipital artery, no residual fistula or early venous drainage.

Table 1.

Characteristics of included studies within the meta-analysis.

Study design	Samaniego 2021		Satomi 2002		Nishi 2022		Authors’ Institution
	Multicenter, Prospective database		Single Center, Retrospective		Multicenter, Prospective database		Single Center, Retrospective
Length of Follow-Up	Mean: 37.7 months (SD: 54.3)		Median: 27.9 (Range 1 mo −17.5 yrs)		36 months		Mean: 17.9 months (SD: 20.5)
Variable, n/total (%)	Intervention	Observation	Intervention	Observation	Intervention	Observation	Intervention	Observation
Patients (n)	218	103	44	68	10	18	7	1
Grade Progression	5/218 (2.3)	0/103 (0)	1/44 (2.3)	2/68 (2.9)	0/10 (0)	0/18 (0)	0/7 (0)	0/1 (0)
Worsening Symptoms (NHND, VHS)	28/218 (13)	6/103 (5.8)	1/44 (2.3)	1/68 (1.5)	1/10 (10)	4/18 (22)	1/7 (14)	0/1 (0)
Death due to dAVF	3/218 (1.4)	0/103	1/44 (2.3)	0/68 (0)	0/10 (0)	0/18 (0)	0/7 (0)	0/1 (0)
mRS 0–2 at Follow Up	204/218 (94)	96/103 (93)	—	—	9/10 (90)	18/18 (100)	7/7 (100)	1/1 (100)
Procedure Complications	35/218 (16)	—	6/44 (1.4)	—	0/10 (0)	—	1/7 (14)	—
Procedure with Permanent Complication	5/218 (2.3)	—	1/44 (2.3)	—	0/10	—	1/7 (14)	—
ICH with Permanent Complication (other than death)	0/218 (0)	0/103 (0)	0/44 (0)	0/68 (0)	0/10 (0)	0/18 (0)	0/7 (0)	0/1 (0)
Death or permanent Complication from procedure, ICH	8/218 (3.7)	0/103 (0)	2/44 (4.5)	0/68 (0)	0/10 (0)	0/18 (0)	1/7 (14)	0/1 (0)
Spontaneous Resolution	—	22/103 (21)	—	—	—	2/18 (11.1)	—	—

Meta-Analysis and pooled quantitative synthesis

Our institutional experience combined with 3 comparative studies from the systematic literature review yielded 495 total patients who possessed a Borden Type I intracranial dAVF.^4,6,7 From these, 469 patients had primary outcome measures that were sufficiently detailed for inclusion in the final pooled quantitative synthesis, which included 279 patients who underwent intervention, and 190 who were observed (Table 1). The most common reported presenting symptoms was pulsatile tinnitus in 65% of patients, followed by orbital or ocular symptoms in 32% of patients, and headache in 27% of cases (Table 2). Regarding the mode of intervention, endovascular embolization alone was reported most frequently occurring in 237/290 (81.7%) of cases. Embolization combined with radiosurgery was the next commonest type of intervention in 33/290 (11.4%) cases. This was followed by radiosurgery alone (11/290, 3.8%), embolization plus surgery (5/290, 1.7%), and open surgery alone (4/290, 1.4%) (Table 2). Lastly, in those studies that reported instances of spontaneous resolution of symptoms, this was found to occur in 24/121 (19.8%) of cases that were observed.

Table 2.

Presenting symptoms and mode of intervention for patients included within the meta-analysis.

Variable	n/total (%)
Presenting Symptoms
Asymptomatic	73/495 (15)
Pulsatile Tinnitus	99/153 (65)
Headache	42/153 (27)
Orbital/Ocular Symptoms	49/153 (32)
Seizure/Cognitive Dysfunction	2/153 (1.3)
Congestive Heart Failure	2/153 (1.3)
NHND	18/495 (3.6)
Intervention Modality
Endovascular embolization	237/290 (81.7)
Open Surgery	4/290 (1.4)
Radiosurgery	11/290 (3.8)
Embolization + Surgery	5/290 (1.7)
Embolization + Radiosurgery	33/290 (11.4)

Regarding the primary outcome measures, there was no significant difference in dAVF grade progression between the intervention and observation arms, 1.8% vs. 0.7%, respectively (RD: 0.01, 95% CI: −0.02 to 0.04, P = 0.49). There was also no significant difference in death directly attributable to a dAVF, with 4/279 (1.4%) in the intervention group and 0/190 (0%) in the observation group (RD: 0.01, CI: −0.01 to 0.03, P = 0.15). Additionally, there was no significant difference between patients who were intervened upon compared to those who were observed, regarding symptom progression occurring in 31/279 (11.1%) intervention patients and in 11/190 (5.8%) observation patients (RD: 0.03, CI: −0.02 to 0.09, P = 0.28). An intervention related complication was reported in 42/279 (15.1%) of cases, with permanent complications reported in 7/279 (2.5%) of patients. Excluding dAVF related hemorrhage that resulted in death, there were no permanent complications including neurologic sequalae due to intracerebral hemorrhage, within the intervention and observation groups. However, when combining the rates of dAVF related death, permanent complication from either intervention or a dAVF related ICH or stroke, there was a significantly higher risk of occurrence in the intervention group (11/279, 3.9%) compared to the observation group (0/190, 0%) (RD: 0.04, CI: 0.1 to 0.06, P = 0.007). Lastly, there was no significant difference regarding the reported rates of functional independence (mRS 0–2) between the intervention arm (220/235, 93.6%) and the observation group (115/122, 94.3%) (RD: 0.0, CI: −0.06 to 0.05, P = 0.87). Full details on the pooled quantitative analysis can be seen in Table 3.

Table 3.

Meta-analysis of pooled outcome variables from the included studies. RD: Risk Difference.

Outcome	Intervention (n = 279)	Observation (n = 190)	RD [95% CI]	P Value	I^2
Grade Progression	5/279 (1.8%)	2/190 (1.1%)	0.01 [−0.02 to 0.04]	0.49	9%
Worsening Symptoms	31/279 (11.1%)	11/190 (5.8%)	0.03 [−0.02 to 0.09]	0.28	24%
Death due to dAVF	4/279 (1.4%)	0/190 (0%)	0.01 [−0.01 to 0.03]	0.15	0%
Procedure Complications	42/279 (15.1%)	—	—	—	—
Procedure with Permanent Complication	7/279 (2.5%)	—	—	—	—
ICH with Permanent Complication (other than death)	0/279 (0%)	0/190 (0%)	—	—	—
Death or Permanent Complication from procedure or ICH	11/279 (3.9%)	0/190 (0%)	0.04 [0.01 to 0.06]	0.007	0%
mRS 0–2 on Last Follow Up	220/235 (93.6%)	115/122 (94.3%)	0.00 [−0.06 to 0.05]	0.87	0%
Spontaneous Resolution	—	24/121 (19.8%)	—	—	—

Evaluation of interstudy heterogeneity

Low interstudy heterogeneity was found within each of the primary outcomes measures in the pooled analyses. Within the death due to dAVF, functional independence, and death or permanent complication from procedure or ICH analyses, the I² value was 0% indicating low interstudy heterogeneity between the included studies. The grade progression analysis had an I² of 8%, and the worsening symptoms analysis had an I² of 24%, both which were also indicative of low heterogeneity between the included studies (Table 3).

Discussion

Here, pooled analysis of 469 patients with Borden Type I intracranial dAVF revealed that patients who were intervened upon rather than observed had a significantly higher risk either dying from their dAVF or experiencing a permanent neurological complication from the dAVF or the performed intervention. Furthermore, there was no difference in grade progression between the two groups, and nearly 1 in 5 patients in the observation arm experienced spontaneous resolution of their symptoms.

High grade dural arteriovenous fistula, which are defined by the presence of cortical venous reflux, are perhaps the most dangerous intracranial vascular abnormality. A study by van Dijk et al. reported that patients with Borden Type II and III dAVF experienced an annual neurologic event rate of 15% and an annual mortality rate of 10% in patients who were observed or underwent non-obliterative therapy. ⁸ The annual rate of new neurological events in patients with high grade fistulae with symptomatic intracranial hemorrhage (ICH) or non-hemorrhagic neurological deficit (NHND) is 19.5% whereas it is 1.4% in those with asymptomatic high-grade fistulae. ⁹ An additional study of 75 dAVFs with cortical venous drainage found an annual risk of hemorrhage of 13% before any intervention, which was reduced to 4.7% following partial treatment. ¹⁰ High grade fistula carry a considerably higher rate of neurological morbidity than other large categories of intracranial vascular pathology, including unruptured aneurysms ¹¹ and arteriovenous malformations. ¹²

In contrast to high grade lesions, low-grade dAVF have a benign natural history such that risk of hemorrhage or non-hemorrhagic neurological deficit cannot be used to justify treatment. Gross et al. found that during a follow up period of 409 lesion-years, there were no hemorrhages from a Borden Type I dAVF, and that 13% of type I lesions in their cohort demonstrated spontaneous obliteration of angiographic follow up. ¹³ Similarly, in a retrospective of analysis of 23 patients, there was a 0% annual rate of ICH, NHNDs, or death due to Borden Type I dAVF during a mean follow up of 6 years. ¹⁴ In a retrospective analysis of 68 patients managed conservatively with a total follow-up time of 348 patient-years, 67 (98.5%) patients with low grade dAVF had stable and tolerable disease. ⁴

Anticipating upconversion to a higher-grade fistula could be used to justify intervention on Borden I lesions, however conversion to a higher grade is a rare phenomenon. Shah et al. reported an annual rate of upconversion of 1% of managed Borden Type I dAVF over a mean follow up of 5.6 years, with a 0% upconversion rate in untreated patients. ¹⁴ Satomi et al. described 68 patients with low grade lesions who were observed with a mean follow up of 2.4 years, during which 67 (98.5%) of dAVFs were stable and did not progress to a higher grade. ⁴ In a 3-year multicenter prospective study of 28 patients with low grade dAVF, no patients experienced conversion to a higher type of Borden classification, both within the treated and non-treated patients. ⁷ Though scarcely reported, upconversion appears to be more frequent among treated low-grade fistula as opposed to those that are observed. Shah et al. reported a 15.6% rate of upconversion in treated patients, versus 0% in untreated low grade dAVF. ¹⁴ Likewise, in the recent Consortium for Dural Arteriovenous Fistula Outcomes Research (CONDOR) database with a mean follow up of 38 months, analysis of Borden Type I dAVF, 0% of the 103 conservatively managed lesions progressed to a higher grade; however, 5/218 (2.3%) of the treated dAVFs upconverted to a Borden type II or III. ⁶

Another rationale for treatment of low-grade fistulae is possible palliative symptomatic relief. Samaniego et al. of CONDOR suggested treatment may yield superior symptomatic control when angiographic obliteration is achieved ⁶ ; though, propensity-matched analysis of the same cohort revealed no difference in clinical outcomes between observed and treated patients when accounting for incomplete treatments. ¹⁵ Here on meta-analysis, there is no significant difference between worsening symptoms, or a higher risk of upconversion between the two arms. However, patients who were treated appear to have a higher risk of either dying or experiencing permanent sequala from their dAVF or the treatment they receive. This should be seriously considered when deciding on intervention in these patients and should be weighed with the goals of palliative intervention. Future research should focus on the nuances that are involved in deciding on intervention aside from angiographic grade alone.

In light of the above, there may be a possible worrisome chronological trend toward progressively higher percentages of low-grade fistulae being treated. Published in 2002, a study from the University of Toronto retrospectively reviewed 117 patients with “benign” dAVF, in which 37% of patients underwent palliative endovascular embolization. ⁴ A study in 2012 detailing management of 23 patients reported that 57% of their patients were endovascularly treated. ¹⁴ Further, in the recently published results by CONDOR database, of the 342 patients with Borden Type I dAVFs from 12 institutions, 67% of these lesions were treated. ⁶ Our institutional experience, in which 7/8 (88%) Borden I patients were treated endovascularly, keeps with this trend (Figure 2). It may be that technological innovation in catheters, liquid embolics, and noninvasive imaging have rendered a solution in search of a problem.

There are several limitations of this study that need to be considered. Firstly, there are only three included studies combined with our institution, all of which are retrospective in design. Additionally, the definition of a Borden Type I dAVF and the clinical outcomes associated with it are inherently based on an angiographic classification, and consequently the data here is not stratified based on the clinical severity of symptoms. It is conceivable that a considerable majority of patients that were treated had more severe symptoms than those that were observed. In fact, in the study by Satomi et al., “intolerable bruit or ophthalmological sequalae” were indications for palliative embolization, even though their lesions were “low grade.” ⁴ Also, the presented analysis did not evaluate dAVF location as a function in regard to the outcome variables, mainly since this was not consistently reported within the literature in regard to clinical outcomes. Similarly, while the majority of cases were treated with endovascular embolization alone (82%), the modality of intervention was not considered in regard to the primary clinical outcome variables, and it is possible that different intervention modalities facilitate different clinical outcomes. Likewise, variations in embolization technique and embolic agent were not able to be evaluated and could function as a theoretical confounding variable, which should also be explored in future studies. There was also not a standard endpoint between the included studies, which limits evaluation of follow-up. Lastly, because the data was not clearly enough defined within the included studies, this analysis did not stratify results based on the degree of dAVF obliteration (e.g., completely versus incompletely treated) and likely functions as a variable regarding the primary outcomes. These above limitations should be considered when interpreting the findings of this study.

On pooled analysis, intervention of Borden Type I dAVF is associated with a higher risk of death or permanent complication when compared to conservative management. While this analysis does not account for the subjective experience of the patient’s symptoms, these findings should be strongly considered when deciding if intervention should be utilized. Future studies should emphasize and explore the decisional nuances that occur when determining on intervention versus observation for Borden Type I dural arteriovenous fistula.