Number and location of draining veins in pediatric arteriovenous malformations: association with hemorrhage

Abstract

Object.

The significance of draining vein anatomy is poorly defined in pediatric arteriovenous malformations (AVMs). In adult cohorts, the presence of fewer veins has been shown to lead to an increased rate of hemorrhage, but this phenomenon has not yet been studied in pediatric AVMs. This report analyzes the impact of draining vein anatomy on presentation and outcome in a large series of pediatric AVMs.

Methods.

Eighty-five pediatric patients with AVMs were treated at the Columbia University Medical Center between 1991 and 2012. Charts were retrospectively reviewed for patient characteristics, clinical course, neurological outcome, and AVM angioarchitectural features identified on the angiogram performed at presentation. Univariate analyses were performed using chi-square test and ANOVA when appropriate; multivariate analysis was performed using logistic regression.

Results.

Four patients were excluded due to incomplete records. Twenty-seven patients had 2 or 3 draining veins; 12 (44.4%) of these patients suffered from hemorrhage prior to surgery. Fifty-four patients had 1 draining vein; 39 (72.2%) of these 54 suffered from hemorrhage. Independent predictors of hemorrhage included the presence of a single draining vein (p = 0.04) and deep venous drainage (p = 0.02). Good outcome (modified Rankin Scale [mRS] score < 3) on discharge was found to be associated with higher admission Glasgow Coma Scale (GCS) scores (p = 0.0001, OR 0.638, 95% CI 0.40–0.93). Poor outcome (mRS score > 2) on discharge was found to be associated with deep venous drainage (p = 0.04, OR 4.68, 95% CI 1.1–19.98). A higher admission GCS score was associated with a lower discharge mRS score (p = 0.0003, OR 0.6, 95% CI 0.46–0.79), and the presence of a single draining vein was associated with a lower mRS score on long-term follow-up (p = 0.04, OR 0.18, 95% CI 0.032–0.99).

Conclusions.

The authors’ data suggest that the presence of a single draining vein or deep venous drainage plays a role in hemorrhage risk and ultimate outcome in pediatric AVMs. Small AVMs with a single or deep draining vein may have the highest risk of hemorrhage.

PEDIATRIC arteriovenous malformations (AVMs) are complex cerebrovascular lesions with a high propensity for devastating hemorrhage prior to surgical intervention. Past studies have shown that intracerebral hemorrhage is the cause for presentation in 46%–87% of cases.^{2,12,13,15,19,28}

Children presenting without symptoms represent a prognostic challenge for physicians, as most natural history data available (suggesting a 2%–4% annual rupture rate) were acquired from predominantly adult cohorts.^4,5,11,26 These data may not adequately reflect the pathophysiology of pediatric AVMs. A principle focus of recent studies has been to identify AVM characteristics that may predispose asymptomatic patients to subsequent hemorrhage. Studies examining AVMs in adults have demonstrated key angioarchitectural features associated with an increased risk of hemorrhage. Traditionally, more attention has been given to arterial features such as arterial aneurysms, artery location, and artery size. In recent years, the role of venous drainage patterns in predisposing to hemorrhage has received more attention. In adult cohorts, deep venous drainage, venous stenosis or ectasia, and a single draining vein have been associated with higher hemorrhage rates.^1,23 In 2008, Guglielmi succinctly explained these phenomena in the “electrical circuit” model of hemodynamics.¹⁰ Analogous to a circuit, decreased diameter of individual veins and of the total venous drainage diameter was felt to increase impedance of the AVM. Selective adjustment of pressure gradients within the AVM does not occur, and thus the increased pressure within the nidus may tax the system to the point of rupture. The association between venous characteristics and hemorrhage in pediatric AVMs is unclear.^2,7,19 In this study, we examine the number and location of draining veins in pediatric AVMs, as well as their associations to presentation and overall outcome. We hypothesize that patients with single draining veins and/or deep venous drainage will suffer from a greater risk of hemorrhagic presentation due to a greater impedance to flow and thus will ultimately have worse outcomes.

Methods

A retrospective review of pediatric patients (age ≤ 21 years) who presented to the Columbia University Medical Center/Morgan Stanley Children’s Hospital of New York between 1991 and 2012 for the treatment of an intracranial AVM was performed. Seventy-seven cases involving patients presenting from 1991 to early 2010 were previously reported, but this occurred prior to the rise in literature suggesting an importance of draining veins in pediatric AVMs.^7,8,24 Patient records and radiographic imaging studies were examined; all patients underwent cerebral angiography prior to intervention. Patients with previous surgical interventions at outside hospitals and those who did not receive treatment were excluded. Angiographic studies were reviewed by an attending neurosurgeon and an interventional neuroradiologist to assess the architecture of each lesion, including the number of veins draining directly from the lesion and the location (deep vs superficial) of the drainage (Fig. 1). Cannulation of draining veins during angiography was not performed as it is not routine in pediatric patients.

FIG. 1.

Catheter angiograms demonstrating an AVM with 2 draining veins (left) compared with a second AVM with a single, deep draining vein (right).

Patients were stratified into 2 groups: those whose AVMs had a single draining vein and those whose AVMs had 2 or more draining veins based on preoperative catheter angiography. Admission characteristics collected during the review included demographic data, AVM features, presence of concurrent aneurysms, presence of hemorrhage, deep versus superficial drainage pattern, Spetzler-Martin grade, and admission Glasgow Coma Scale (GCS) score. The variable of interest was the presence of hemorrhage on admission.

Stratifying by single versus multiple draining veins, univariate analyses were conducted to determine associations between the number of veins and the presenting clinical characteristics. Additional univariate analyses were conducted to determine the association of presenting characteristics with hemorrhagic presentation. Based on the above analyses, a subsequent univariate assessment of the relationship between the number of veins, size of the nidus, and hemorrhagic presentation was performed. Outcome was assessed for univariate associations with the venous characteristics and presenting characteristics using the modified Rankin Scale (mRS). The Fisher’s exact test, chi-square test, t-test, Wilcoxon signed-rank test, and Mann-Whitney test were used as appropriate. Parametric tests were preferred to nonparametric in cases where normalcy was approximated on a histogram and q-q plot. All variables with a univariate association of p < 0.2 were entered in a logistic regression model to predict AVM hemorrhage, mRS score > 2 at discharge, and mRS score > 2 on long-term follow-up. All statistical analyses were performed using the Statistical Package for the Social Sciences Version 18 (IBM/SPSS).

Results

Patient Population

A total of 85 patients underwent interventions for AVMs between 1991 and 2012. Sixty-three of these patients were previously reported on in our study of AVM-associated aneurysms in the pediatric population;² 4 patients who were included in the previous study (which reported on a total of 77 cases) were excluded due to lack of sufficient records to complete the analysis (1 patient) or an inability to determine the number of draining veins from available imaging (3 patients). The mean age of included patients was 13.3 ± 5.4 years. Forty-three patients (53%) were male and 38 (47%) were female. Ten patients (12%) had infratentorial lesions and 71 (88%) had supratentorial lesions. Thirty-four patients (42%) had right-sided lesions and 47 (58%) had left-sided lesions. The drainage was exclusively deep in 21 patients (26%), exclusively superficial in 47 (58%), and mixed in 13 (16%). In terms of AVM size, 58 patients (72%) had AVMs less than 3 cm in greatest diameter, and the remaining 23 (28%) had AVMs with a maximum diameter of 3–6 cm. No patients had an AVM larger than 6 cm. Fifty-six AVMs (69%) involved eloquent areas. The mean Spetzler-Martin grade of the AVMs was 2.38 (median 2). Twenty-three patients (28%) had concurrent aneurysms. Fifty-one patients (63%) presented with hemorrhage. The mean admission GCS score was 13.7 and the mean mRS score at discharge was 1.14. The mean mRS score on follow-up was 0.69, and the mean length of follow-up was 36 months (Table 1). Forty-five patients underwent embolization and microsurgery, 17 patients underwent Gamma Knife radiosurgery, 13 patients underwent microsurgery alone, 4 patients underwent embolization and radiosurgery, 1 patient underwent embolization alone, and 1 patient underwent radiosurgery and microsurgery.

TABLE 1:

Demographic, presentation, and radiographic variables^*

Variable	1 Vein (n = 54)	>1 Vein (n = 27)	p Value	Total Population (n = 81)
age (yrs), mean	12.8	14.1	0.3	13.3
male sex	26	17	0.2	43
infratentorial AVM	6	4	0.1	10
right-sided AVM	24	10	0.5	34
drainage			0.0008
deep	15	6		21
superficial	37	10		47
both	2	11		14
AVM size			0.6
<3 cm	40	18		58
3–6 cm	14	9		23
>6 cm	0	0		0
eloquence	34	22	0.09	56
Spetzler-Martin grade	2.2	2.74	0.06	2.38
1	11	1		12
2	24	10		34
3	16	11		27
4	3	5		8
5	0	0		0
ICH on presentation	39	12	0.02	51
aneurysm	14	9	0.5	23
admission GCS scoret†	13.7	13.5	0.7	13.7
>12	45	21		66
9–12	4	3		7
<9	5	3		8
mRS score at discharge	1.11	1.19	0.1	1.14
0	17	8		25
1	25	8		33
2	5	8		13
3	4	1		5
4	2	1		3
5	1	0		1
6	0	0		0

^*Values are for Student t-test, Mann-Whitney test, Fisher’s exact test, or chi-square test as appropriate. Boldface type indicates statistical significance. ICH = intracerebral hemorrhage.

^†Data missing for 1 patient.

Trends in the Presenting Characteristics of Patients With Single Versus Multiple Draining Veins

The mean age of patients with 1 draining vein was 12.8 ± 5.6 years compared with 14.1 ± 5.3 years for other patients (p = 0.3). Twenty-six patients (48%) with 1 draining vein were male, compared with 17 (63%) of the patients with multiple draining veins (p = 0.21). Six (11%) patients with 1 draining vein had infratentorial AVMs, compared with 4 (15%) in the other group (p = 0.1). Twenty-four patients (44%) with 1 draining vein had a right-sided lesion, compared with 10 (37%) in the other group (p = 0.5). Among patients with 1 draining vein, the drainage was deep in 15 cases, superficial in 37, and both superficial and deep in 2 cases in which the single draining vein was seen to bifurcate distal to the nidus. Among patients with more than 1 draining vein, the drainage was deep in 6 cases, superficial in 10, and both deep and superficial in 11 (p = 0.0008). The mean Spetzler-Martin grade was 2.2 for patients with 1 draining vein versus 2.74 in other patients (p = 0.06). Thirty-nine patients (72%) with 1 draining vein had a hemorrhagic presentation, compared with 12 (44%) of the patients with more than 1 draining vein (p = 0.02). Fourteen patients (26%) with 1 draining vein had a concurrent arterial aneurysm, compared with 9 (33%) of those with multiple draining veins (p = 0.5). The mean admission GCS score was 13.7 in patients with a single draining vein and 13.5 in those with multiple draining veins (p = 0.68); the mean discharge mRS scores were 1.11 and 1.19, respectively (p = 0.1) (Table 1). At 36-month follow-up, the mean mRS score was 0.48 in patients with 1 draining vein and 1.12 in patients with more than 1 draining vein (p = 0.07). However, the mean duration of follow-up was 42.6 months in patients with a single draining vein and 21.6 months in patients with more than 1 draining vein (p = 0.01).

Univariate and Multivariate Predictors of Hemorrhage

On univariate analysis, factors associated with hemorrhage with a p < 0.2 were age (p = 0.08), admission GCS score (p = 0.005), presence of a concurrent arterial aneurysm (p = 0.06), presence of a single draining vein (p = 0.02), presence of exclusively deep venous drainage (p = 0.08), and small AVM size (p = 0.094). On multivariate analysis, the presence of a single draining vein (p = 0.04, OR 5.16, 95% CI 1.05–25.5) and exclusively deep venous drainage (p = 0.02, OR 8.22, 95% CI 1.32–51.1) were found to be predictive of hemorrhage, but age (p = 0.3), admission GCS score (p = 0.8), presence of a concurrent arterial aneurysm (p = 0.5), and large AVM size (p = 0.1) were not (Table 2).

TABLE 2:

Multivariate analysis of characteristics predicting hemorrhage^*

Characteristic	Univariate	Multivariate
	p Value	p Value	OR (95% CI)
age	0.08	0.3	NA
admission GCS score	0.005	0.8	NA
concurrent aneurysm	0.006	0.5	NA
male sex	0.3
right-sided AVM	0.9
infratentorial AVM	0.6
1 draining vein	0.02	0.04	5.16 (1.05–25.5)
deep drainage	0.08	0.02	8.22 (1.32–51.1)
eloquent	0.9
large AVM size	0.09	0.12	NA

^*Boldface type indicates variables with a p value < 0.2 on univariate analysis (i.e., those included in the multivariate analysis) and statistical significance (p < 0.05 on multivariate analysis). NA = not applicable.

Univariate and Multivariate Predictors of Outcome

On univariate analysis, factors associated with a discharge mRS score greater than 2 (p < 0.2) were admission GCS score (p = 0.02), presence of a concurrent arterial aneurysm (p = 0.1), infratentorial location (p = 0.1), presence of a single draining vein (p = 0.1), presence of exclusively deep venous drainage (p = 0.08), and eloquent location (p = 0.1). On multivariate analysis, a higher admission GCS score was predictive of a lower discharge mRS score (p = 0.001, OR 0.638, 95% CI 0.40–0.93), whereas the presence of exclusively deep venous drainage was found to be predictive of a higher mRS score (p = 0.04, OR 4.68, 95% CI 1.1–19.98). The presence of a concurrent arterial aneurysm (p = 0.2), infratentorial location (p = 0.3), the presence of a single draining vein (p = 0.1), and eloquent location (p = 0.2) were not found to be associated with a discharge mRS score greater than 2 (Table 3).

TABLE 3:

Multivariate analysis of characteristics predicting outcome^*

Characteristic	Discharge mRS Score >2			Recent mRS Score >2
	Univariate	Multivariate		Univariate	Multivariate
	p Value	p Value	OR (95% CI)	p Value	p Value	OR (95% CI)
age	0.9			0.6
admission GCS score	0.02	0.0001	0.638 (0.40–0.93)	0.003	0.0003	0.6 (0.46–0.79)
concurrent aneurysm	0.1	0.19	NA	0.004	0.02	10 (1.35–74.23)
male sex	0.5			0.4
right-sided AVM	0.7			0.8
infratentorial AVM	0.1	0.3	NA	0.1	0.2	NA
1 draining vein	0.1	0.1	NA	0.1	0.04	0.18 (0.032–0.99)
deep drainage	0.08	0.04	4.68 (1.1–19.98)	0.1	0.3	NA
eloquent	0.1	0.2	NA	0.5
large AVM size	0.6			0.3

^*Boldface type indicates variables with a p value < 0.2 on univariate analysis (i.e., those included in the multivariate analysis) and statistical significance (p < 0.05 on multivariate analysis).

On univariate analysis, factors associated with a follow-up mRS score greater than 2 (p < 0.2) were admission GCS score (p = 0.003), presence of a concurrent arterial aneurysm (p = 0.004), infratentorial location (p = 0.1), presence of a single draining vein (p = 0.1), and presence of exclusively deep venous drainage (p = 0.1). On multivariate analysis, a higher admission GCS score was associated with a lower follow-up mRS score (p = 0.0003, OR 0.6, 95% CI 0.46–0.79), as was the presence of a single draining vein (p = 0.04, OR 0.18, 95% CI 0.032–0.99). The presence of concurrent aneurysms (p = 0.2, OR 10, 95% CI 1.35–74.23) was found to be associated with a higher follow-up mRS score. Infratentorial location (p = 0.2) and the presence of exclusively deep venous drainage (p = 0.3) were not found to be independently associated with follow-up mRS score. We found no difference in outcome between the use of microsurgery (mean mRS score 0.64) or radiosurgery (mean mRS score of 0.81) on long-term follow-up (p = 0.90).

Relationship Between Size, Hemorrhage, and Draining Veins

Of the 51 hemorrhagic presentations, 40 (78%) involved patients with small AVMs and 11 (22%) involved patients with large AVMs (p = 0.08; Fig. 2). When veins were stratified by hemorrhagic presentation and small AVM size, 32 of the hemorrhagic cases were found to occur in small AVMs with a single draining vein compared with 8 small AVMs with more than 1 draining vein. The remaining hemorrhagic presentations occurred in 7 large AVMs with 1 draining vein and 4 large AVMs with more than 1 draining vein (p = 0.05; Fig. 3).

FIG. 2.

Comparison of the numbers of hemorrhagic presentations in patients with small and large AVMs.

FIG. 3.

Pie charts depicting a relative preponderance of single draining veins over multiple draining veins among patients with small AVMs presenting with hemorrhage (B) compared with patients with small AVMs presenting without hemorrhage (A). In contrast, a similar proportion of single draining veins in patients with large AVMs presenting with (D) or without (C) hemorrhage was noted. ICH = intracerebral hemorrhage.

Discussion

In this report, we have shown that: 1) the presence of a single draining vein is associated with hemorrhagic presentation in pediatric patients with AVMs; 2) the presence of exclusively deep venous drainage is associated with hemorrhagic presentation in pediatric patients with AVMs; and 3) single draining veins and deep venous drainage are independently associated with outcome.

Several prior studies have investigated venous characteristics in pediatric AVMs (Table 4). Maher and Scott reported that approximately half of their patients presented with a linear AVM with a single deep draining vein. These patients were less likely to suffer from postoperative complications, but no quantitative assessment of hemorrhagic risk, deep venous drainage, or outcome was presented.¹⁹ It has been postulated that single draining veins are the baseline status of early-stage AVMs, with subsequent veins developing later on in development. Guglielmi’s electrical circuit theory of AVM hemodynamics can be applied to this postulate to provide a possible explanation.¹⁰ The elevated intranidal stress due to the increased impedance to blood drainage may predispose the AVM to hemorrhage, but may also stimulate the development of collateral arteriovenous circulation in the form of perinidal vessels as a mechanism of compensation.

TABLE 4:

Pediatric and adult literature evaluating AVM venous architecture

Authors & Year	Journal	Study Population	Findings
Lv et al., 2011	World Neurosurg	adult	Single draining veins & combined superficial & deep drainage are associated w/ hemorrhage.
Stefani et al., 2002	Stroke	adult	Lower number of draining veins, deep drainage, & venous ectasia are associated w/ hemorrhage.
Miyasaka et al., 1992	J Neurosurg	adult	Single draining veins, impeded venous drainage, & deep venous drainage are associated w/ hemorrhage.
Turjman et al., 1995	Neurosurgery	adult	Deep venous drainage is associated w/ hemorrhage.
Duong et al., 1998	Stroke	adult	Periventricular drainage, deep drainage, & low vein number are associated w/ hemorrhage.
Miyasaka et al., 2000	Acta Neurochir	adult	Single draining veins & elevated drainage venous pressure are associated w/ hemorrhage.
Gross & Du, 2012	Neurosurgery	adult	Exclusive deep venous drainage is associated w/ hemorrhage.
Niu et al., 2012	Neurosci Bull	adult	Low vein number is associated w/ hemorrhage.
Stapf et al., 2006	Neurology	adult	Exclusive deep venous drainage is associated w/ hemorrhage.
Mansmann et al., 2000	Neurosurgery	adult	Venous stenosis is associated w/ increasing risk of hemorrhage.
Shankar et al., 2012	Can J Neurol Sci	adult	Stenosis >50% in veins & presence of long pial drainage course are associated w/ seizures.
Li et al., 2012	World Neurosurg	adult	Presence of >3 draining veins is associated w/ nonhemorrhagic neurological deficits.
Marks et al., 1990	Radiology	adult	Deep venous drainage is associated w/ hemorrhage, but morphology is not.
Langer et al., 1998	Neurosurgery	adult	Deep venous drainage is associated w/ hemorrhage, but morphology is not.
Kandai et al., 2010	Malays J Med Sci	adult	Deep venous drainage is associated w/ hemorrhage.
Ellis et al., 2013	J Neurointervent Surg	adult	Deep drainage is associated w/ hemorrhage, but single draining veins & stenosis are not.
Maher & Scott, 2009	J Neurosurg Pediatr	adult	Linear vein-based AVMs are common in children.
Niazi et al., 2010	Neurosurg Clin N Am	pediatric	Solitary draining veins are most common in children.
Anderson et al., 2012	J Neurosurg Pediatr	pediatric	Deep venous drainage is not an independent predictor of hemorrhage, but exclusively deep drainage was not investigated.
Ellis et al., 2010	J Neurosurg Pediatr	pediatric	Most hemorrhagic AVMs had superficial drainage (no statistical analysis).

The high prevalence of single draining veins and small AVMs in our study is supported by previous studies of pediatric cohorts.^3,15,19 Small AVM size has been reported as a risk factor for hemorrhagic presentation in past studies.^16,29 From the perspective of the electrical circuit model, a greater number of perinidal vessels would provide a decrease in hemodynamic stress of the AVM secondary to reduction in vascular resistance provided by parallel flow. Perinidal vessels are a significant factor in AVM size; thus, it may be the area of parallel flow, rather than AVM diameter, which is the true risk factor. Our data demonstrated that small AVMs with a single draining vein (p = 0.05) had the highest rate of hemorrhage, whereas small AVM size alone trended toward significance on univariate and multivariate analysis. Interestingly, patients with small AVMs who did not present with hemorrhage were primarily those with multiple draining veins. Patients with large AVMs appeared less affected by the number of draining veins, although there was a non–statistically significant trend toward a single draining vein being more common with hemorrhage. This suggests that, in patients with few perinidal arteriovenous vessels, the presence of the additional impedance of a single draining vein is more likely to be associated with hemorrhage. These patients may warrant more aggressive treatment than patients with larger AVMs and multiple veins.

Ellis et al. demonstrated an association for hemorrhagic presentation in patients with deep venous drainage, small AVM size, or infratentorial location.⁸ We similarly found an association between deep venous drainage and a trend toward small AVM size. Of note, exclusively deep venous drainage was independent of the association between the presence of a single draining vein and hemorrhage in our cohort. However, in our past report, the presence of any contribution of deep venous drainage was not associated with hemorrhage, although a nonsignificant trend was seen.² Deep draining veins typically have a smaller diameter than superficial draining veins, as well as a high frequency of stenosis. These characteristics may suggest that the predisposition to hemorrhage frequently seen in patients with exclusively deep venous drainage is a manifestation of increased impedence in a similar matter to single draining veins. Infratentorial location was not found to be associated with hemorrhage in our cohort, although only 10 patients with infratentorial AVMs were included.

The relationship between venous architecture and outcome appears to be complex. In this study, exclusively deep venous drainage was found to be associated with a worse outcome (higher mRS score) on discharge but not at 36-month follow-up. Regarding the number of draining veins, contrary to expectations, a single draining vein was found to reduce the risk of poor outcome at 36 months’ follow-up. One possible explanation for this is that there is a greater disruption of venous drainage from the normal brain adjacent to the AVM when multiple veins, rather than a single draining veins, are resected during microsurgery or thrombosed after radiosurgery.³² However, given the tendency for patients with an AVM with a single draining vein to have longer follow-up times (mean 42.6 months vs 21.6 months for patients with multiple draining veins), this conclusion must be cautiously stated given the potential for slow, long-term improvements in function unrelated to AVM angioarchitecture. Similar to prior studies, arterial aneurysm presence and low admission GCS score were also found to be associated with poor long-term outcome. These data are in agreement with with our center’s conviction that aggressive intervention for patients with arterial aneurysms may be warranted due to the increased risk of hemorrhage and poor outcome.²

The primary limitation of our study is that it is a single-institution retrospective review with relatively short-term follow-up for a pediatric study. Data other than angiographic features were based on chart review, with its inherent limitations of accuracy and interpretation. Children receiving treatment elsewhere or who were evaluated but not surgically treated at our institution were excluded due to inconsistent follow-up (including imaging) with the neurosurgery department at our institution. This limits our findings to surgical cohorts. Given the associations between hemorrhage and venous characteristics, the associations with outcome and venous architecture may be spurious. Furthermore, we did not incorporate venous diameter or stenosis in our analysis.

Conclusions

The results of this study suggest that the presence of a single draining vein or exclusively deep venous drainage increases the risk of hemorrhage in children with AVMs, similar to adults. However, patients with a single draining vein may ultimately have a better long-term outcome than those with multiple draining veins. Further studies on the natural history of pediatric AVMs are needed and will require collaboration between multiple centers.

Abbreviations used in this paper:

AVM

arteriovenous malformation

GCS

Glasgow Coma Scale

mRS

modified Rankin Scale