The angio-architectural features of brain arteriovenous malformations: is it possible to predict the probability of rupture?

Abstract

Background

Hemorrhage is the most devastating complication of brain arteriovenous malformations (bAVMs), and to date, there is still concern about the needing for treatment in case of unruptured and asymptomatic bAVM. In fact, the morbidity and mortality of treatments may exceed that of the AVM’s natural history. None of the classifications and scores for bAVM allows to predict the risk of bleeding. In this study, we aimed to identify the angio-architectural characteristics of brain AVMs associated with bleeding.

Methods

We retrospectively evaluated all consecutive patients diagnosed with cerebral AVMs, between January 2010 and December 2019 from our prospective bAVM database. Univariate and multivariate logistic regression analysis were used to evaluate relationships between angio-architectural features of ruptured and unruptured bAVMs.

Results

Of the 143 retrieved bAVMs, 65 were unruptured and 78 were ruptured. The univariate logistic regression analysis demonstrated statistically significant differences into angio-architectural features of unruptured and ruptured bAVMs. The multivariate logistic regression analysis fitted well (p =.113) with a good discrimination capacity (ROC = 0.83) of three independent angio-architectural features mainly related to bleeding in bAVMs: a smaller diameter of the nidus (p < .001), the absence of venous drainage alterations (p = .047), of the presence of prenidal aneurysms (p = .005).

Conclusions

In our study, several features resulted related to an increased probability of rupture for bAVMs, among which the more relevant were a small diameter of the nidus, the absence of venous drainage alterations, and the presence of prenidal aneurysms.

Introduction

Intracranial hemorrhage is the most devastating complication of brain arteriovenous malformations (bAVMs) and could lead to severe outcomes, from death to severe neurological deficits. Thus, the possibility to predict bAVMs prone to rupture is of great interest, to treat them promptly and avoid potentially fatal sequelae.

Although results of the ARUBA trial were criticized on various aspects, to date, it remains the highest available level of evidence regarding the management of patients with bAVMs.^1,2 Its results states that for ruptured bAVMs, treatment is typically indicated,^2–4 while for unruptured bAVMs the medical management alone is superior to medical management with treatment. To date, there is still concern about the needing for preventive treatment in patients with unruptured bAVMs, ¹ as the morbidity and mortality of treatments may exceed that of AVM’s natural history.^5,6

Previous articles have shown that the angioarchitecture of bAVMs evolves across time,^7–11 and bleeding mainly occurs when the hemodynamical balance is lost. ¹² Thus, there is the needing to identify features associated with higher probability of rupture, to prevent this potentially dramatical event.^12–15

Despite several anatomic, hemodynamic, and angio-architectural features have been identified as being more prevalent in ruptured bAVMs, no scores or classification have been developed to consider the probability of bleeding. ¹⁶ In fact, none of the Spetzler and Martin (SM), ¹⁷ the Spetzler-Ponce, ¹⁸ the Lawton’s supplementary system, ¹⁹ the Buffalo, ²⁰ and Toronto classifications and scores allow to predict the probability of bleeding for bAVMs. ²¹

In this study, we aimed to identify the angio-architectural characteristics of brain AVMs associated with bleeding.

Methods

In this retrospective study, we reviewed all consecutive patients diagnosed with cerebral AVMs, between January 2010 and December 2019 from our prospective bAVM database. The scientific and ethics committees of our institution approved the conduct of this research and we adhered to the Declaration of Helsinki.

Inclusion criteria were any patient with diagnosis of bAVM, and with at least one pre-operative diagnostic cerebral digital subtraction angiography (DSA) performed in a standardized fashion with biplane angiography, including injection of both internal carotid arteries (ICAs) and external carotid arteries, and of the vertebral arteries (VAs) bilaterally. Both ruptured and unruptured bAVM of any grade according to the Spetzler Martin classification were considered suitable for this study. No limits of age and sex were considered. Exclusion criteria were a bad quality angiography and previously treated bAVMs.

Data acquisition

We retrospectively assessed the clinical, demographic, and imaging data (brain CT or MRI and, cerebral conventional DSA) of all included patients. Age at presentation, sex, and bleeding were included, as well as brain AVM’s angio-architectural features based on DSA evaluation. DSA images were carefully and independently reviewed by two neuroradiologists (a senior one with more than 15 years of experience and a junior one with 7 years of experience), to include numbers and characteristics of the arterial feeders (terminal, en-passage, pials, and the diameter of the major one, measured 1 cm before the nidus) (Figures 1, 2, and 3), the number of the major arterial supplies, the presence/absence of extracranial arterial supply, the diameter and volume of the nidus, the type of the nidus (compact, diffuse, or plurifocal), the presence/absence and characteristics of aneurysms (prenidal, intranidal, mixed), the number and type of venous drainage (superficial, deep, or mixed), the presence/absence of venous drainage alterations (ectasia, varices, stenosis, thrombosis, flow reversal), and the Buffalo and Spetzler Martin grades. The volume of the nidus was expressed in mm³ and calculated with the formula of an ellipsoid with different-axes lengths: $\frac{4}{3}\pi \left(lenght\times width\times height\right)$ . The diameter of the major arterial feeder measured 1 cm before the nidus was done on our DICOM image viewing software (CARESTREAM Vue PACS, Carestream Health Inc., USA). We used a curved line tool to measure 1 cm prior to the nidus on the major arterial arterial feeder and at that point we measured the diameter of the arterial feeders itself (Figures 1 and 2). All nomenclature and definitions of terms for the angio-architectural features of bAVMs were assessed basing on the joint writing group for the reporting terminology for Brain Arteriovenous Malformation. ²²

Figure 1.

(A) bAVM with a compact nidus receiving arterial supply mainly by terminal feeders arising from the right middle cerebral artery. (B) Demonstration of the measurement of the diameter of the major arterial feeder at 1 cm from the nidus. (C) Multiple intranidal aneurysms (arrows). (D) Single ectasic venous collector draining into the right transverse sinus (superficial venous drainage). (E) 3D DSA reconstruction demonstrating the presence of a venous varix (arrow) at the origin of the venous collector.

Figure 2.

(A) bAVM of the posterior fossa with a compact nidus. (B) The nidus receives arterial supply by multiple arterial feeders «en-passage» (arrows). (C) Demonstration of the measurement of the diameter of the major arterial feeder at 1 cm from the nidus. (D) A prenidal aneurysm (arrow). (E) Presence of venous reflux into one of the cerebellar cortical veins draining the bAVM (arrow).

Figure 3.

(A–B) bAVM with a plurifocal and diffuse nidus, receiving arterial supply by both anterior and posterior circulations. (C) Pials arterial feeders (arrow) also supply the plurifocal nidus of bAVM. (D) Presence of an intranidal giant aneurysm (arrow). (E) Mixed (superficial and deep) venous drainage draining the bAVM through cortical venous system and deep veins (the internal cerebral veins and the basal veins).

Statistics

Categorical variables were expressed as absolute numbers (n) and percentages (%), while continuous variables were expressed as mean and standard deviation (SD) or median and interquartile range (IQR) as appropriate. Variables differences were analyzed in the groups of unruptured and ruptured AVMs, with radiological evaluation of bleeding. The differences between the two groups were assessed by Mann–Whitney U test for continuous variables and by Pearson chi-square test or Fisher’ exact test for categorical ones.

The association between each demographic and clinical characteristic and the rupture of bAVM was assessed by univariate logistic regression models. Then, variables associated to the bAVM’s rupture were included in a multivariate logistic regression analysis; the final model was the one that best fit the data and with the greatest discrimination ability based on Hosmer and Lemeshow test and the area under the receiver-operating characteristic (ROC) curve, respectively. All statistical analyses were performed using Stata SE version 14.2 (StataCorp LLC).

Results

Patient population and angio-architectural baseline characteristics

A total of 143 patients (mean age 46.1 year (sd: 19 years)) were included in the analysis (Table 1). Of them, 65 were patients with unruptured bAVMs, while 78 were patients with ruptured bAVMs. One patient presented two separate bAVMs with different arterial supply, nidus, and venous drainage; only one of the bAVMs was ruptured due to the distribution of blood in the CT scan. Thus, as suggested by the joint writing group for the reporting terminology for Brain Arteriovenous Malformation, ²² we characterized separately each cerebral AVM.

Table 1.

Angio-architectural baseline characteristics compared between ruptured and unruptured bAVMs groups.

		Unruptured bAVM (N = 65)	Ruptured bAVM (N = 78)	p-value
Age at diagnosis [years]	median (IQR)	43 (30–57)	50 (35–65)	0.039
Diameter of nidus (LL horizontal) [mm]	median (IQR)	2.88 (2.15–4.3)	1.64 (1.13–2.59)	<0.001
Diameter of nidus (LL vertical) [mm]	median (IQR)	2.84 (1.96–3.8)	1.37 (1.02–2.23)	<0.001
Diameter of nidus (AP transverse) [mm]	median (IQR)	2.68 (2.11–3.69)	1.4 (0.93–1.8)	<0.001
Volume of nidus [mm3]	median (IQR)	12.21 (5.97–26.27)	1.42 (0.63–4.73)	<0.001
N of arterial axes	median (IQR)	2 (1–2)	1 (1–2)	0.106
N of arterial feeders	median (IQR)	5 (3–7)	4 (3–5)	0.018
N° of terminal arterial feeders	median (IQR)	2 (1–4)	2 (0–3)	0.106
N° of “en-passage” arterial feeders	median (IQR)	3 (1–4)	2 (1–3)	0.299
N° of pial arterial feeders	median (IQR)	0 (0–0)	0 (0–0)	0.432
N of draining veins	median (IQR)	2 (1–3)	1 (1–2)	0.005
Maximal diameter of nidus [mm]	median (IQR)	3.31 (2.53–4.53)	1.89 (1.37–2.75)	<0.001
Diameter of the major arterial feeder measured 1 cm before nidus [mm]	median (IQR)	2.08 (1.62–2.6)	1.4 (1–2.03)	<0.001
Aneurysm bAVM	median (IQR)	1 (0–2)	1 (0–2)	0.828
Sex				0.999
Males	n (%)	30 (46.15)	36 (46.15)
Females	n (%)	35 (53.85)	42 (53.85)
Type of nidus				0.518
Compact	n (%)	2 (3.08)	4 (5.13)
Diffuse	n° (%)	63 (96.92)	72 (92.31)
Pluri-focal	n (%)	0 (0)	2 (2.56)
N° of arterial axes				0.036
1	n (%)	32 (49.23)	50 (64.10)
2	n (%)	26 (40.00)	21 (26.92)
3	n (%)	6 (9.23)	3 (3.85)
4	n (%)	0 (0)	4 (5.13)
5	n (%)	1 (1.54)	0 (0)
External carotid artery supply				0.251
No	n (%)	65 (100)	75 (96.15)
Yes	n° (%)	0 (0)	3 (3.85)
Type of venous drainage				0.049
Superficial	n (%)	31 (47.69)	45 (57.69)
Deep	n (%)	16 (24.61)	24 (30.77)
Mixed	n (%)	18 (27.60)	9 (11.54)
Venous drainage alterations				<0.001
No	n (%)	14 (21.54)	40 (51.28)
Yes	n (%)	51 (78.46)	38 (48.72)
Venous ectasia				<0.001
No	n (%)	22 (33.85)	52 (66.67)
Yes	n (%)	43 (66.15)	26 (33.33)
Venous varix				0.105
No	n (%)	51 (78.46)	69 (88.46)
Yes	n (%)	14 (21.54)	9 (11.54)
Venous stenosis				0.857
No	n (%)	56 (86.15)	68 (87.18)
Yes	n (%)	9 (13.85)	10 (12.82)
Venous thrombosis				0.501
No	n (%)	65 (100)	76 (97.44)
Yes	n (%)	0 (0)	2 (2.56)
Venous flow reversal				0.999
No	n (%)	63 (96.92)	75 (96.15)
Yes	n (%)	2 (3.08)	3 (3.85)
Eloquent area				0.614
No	n (%)	39 (60)	50 (64.10)
Yes	n (%)	26 (40)	28 (35.90)
Spetzler Martin grade				0.015
1	n (%)	10 (15.38)	23 (30.67)
2	n (%)	20 (30.77)	29 (38.67)
3	n (%)	23 (35.38)	18 (24.00)
4	n (%)	11 (16.92)	3 (4.00)
5	n (%)	1 (1.54)	2 (2.67)
Buffalo grade				0.997
1	n (%)	8 (12.31)	9 (12.00)
2	n° (%)	14 (21.54)	16 (21.33)
3	n (%)	24 (36.92)	29 (38.67)
4	n (%)	19 (29.23)	21 (28.00)
Atypical feature of bAVM				0.473
No	n (%)	28 (43.08)	29 (37.18)
Yes	n (%)	37 (56.92)	49 (62.82)
Type of aneurysms				0.010
Intranidal	n (%)	15 (23.08)	13 (16.67)
Prenidal	n (%)	9 (13.85)	29 (37.18)
Mixed	n (%)	13 (20.00)	7 (8.97)
No aneurysm	n (%)	28 (43.08)	29 (37.18)

The female proportion in both unruptured and ruptured bAVM was the same (53.85%). Ruptured bAVMs were more frequently detected in older patients (median age 50, IQR 35–65) compared to unruptured bAVMs in which patients were younger (median age 43, IQR 30–57) (p = .039). Between the two groups, statistical significant differences were present in the number of the arterial feeders (p = 0.018), in the diameter of the major arterial feeder measured 1 cm before the nidus (p < 0.001), in the number of major arterial supply (p = 0.036), in the 3 diameters (p < 0.001) and in the volume of the nidus (p < 0.001), in the number of draining veins (p = 0.005) and in the type of venous drainage (p = 0.049), as well as the presence of alterations of the venous drainage (p < 0.001), particularly venous ectasia (p < 0.001). Moreover, also the Spetzler and Martin grade (p = 0.015) and the type of aneurysms (p = 0.010) were significantly different among ruptured and unruptured bAVMs.

Factors associated with ruptured bAVMs

The univariate logistic regression analysis (Table 2) demonstrates that for every one unit increase of the diameter of the major arterial feeders measured 1 cm before the nidus (p < 0.001, OR = 0.42, 95% CI 0.26–0.67), of the maximum diameter of the nidus of the nidus (p = 0.002, OR = 0.71, 95% CI 0.58–0.89), of the volume of the nidus (p = 0.031, OR = 0.98, 95% CI 0.97–1.00) and of the number of the draining veins (p = 0.005, OR = 0.64, 95% CI 0.46–0.87) the odds of bAVMs rupture decrease.

Table 2.

Univariate logistic regression analysis.

	Odds ratio	95% Conf. Interval		p-value
Age	1.02	1.00	1.04	0.036
Diameter of nidus (LL horizontal) [mm]	0.67	0.54	0.85	0.001
Diameter of nidus (LL vertical) [mm]	0.60	0.46	0.79	<0.001
Diameter of nidus (AP transverse) [mm]	0.48	0.35	0.66	<0.001
Volume of nidus [mm3]	0.98	0.97	1.00	0.031
N° of arterial axes	0.79	0.52	1.21	0.274
N° of arterial feeders	0.88	0.78	1.00	0.059
N° of terminal arterial feeders	0.86	0.73	1.02	0.079
N° of “en-passage” arterial feeders	0.93	0.77	1.11	0.415
N° of pial arterial feeders	0.99	0.53	1.87	0.977
N° of draining veins	0.64	0.46	0.87	0.005
Maximal diameter of nidus [mm]	0.71	0.58	0.89	0.002
Diameter of the major arterial feeder measured 1 cm before nidus [mm]	0.42	0.26	0.67	<0.001
Aneurysm bAVM	1.02	0.79	1.32	0.878
Female sex	1.00	0.52	1.94	0.999
Type of nidus
Compact	1.00 (base)
Diffuse	0.57	0.10	3.23	0.526
Pluri-focal	1.00
N° of arterial axes
1	1.00 (base)
2	0.52	0.25	1.07	0.075
3	0.32	0.07	1.37	0.125
4	1.00 (empty)
5	1.00 (empty)
External carotid artery supply	1.00 (empty)
Type of venous drainage
Superficial	1.00
Deep	1.03	0.47	2.26	0.934
Mixed	0.34	0.14	0.87	0.023
Venous drainage alterations	0.26	0.12	0.55	<0.001
Venous ectasia	0.26	0.13	0.51	<0.001
Venous varix	0.48	0.19	1.18	0.110
Venous stenosis	0.92	0.35	2.41	0.857
Venous thrombosis	1.00 (empty)
Venous flow reversal	1.26	0.20	7.78	0.803
Eloquent area	0.84	0.43	1.66	0.614
Spetzler Martin grade
1	1.00 (base)
2	0.63	0.25	1.61	0.334
3	0.34	0.13	0.89	0.029
4	0.12	0.03	0.52	0.005
5	0.87	0.07	10.73	0.913
Buffalo grade
1	1.00 (base)
2	1.02	0.31	3.35	0.979
3	1.07	0.36	3.21	0.898
4	0.98	0.32	3.06	0.976
Atypical feature of bAVM	1.28	0.65	2.50	0.474
Type of aneurysms
Intranidal	0.84	0.34	2.07	0.700
Prenidal	3.11	1.25	7.73	0.015
Mixed	0.52	0.18	1.49	0.224
No aneurysm	1.00 (base)

Mixed venous drainage (p = 0.023, OR = 0.34, 95% CI 0.14–0.87), as well as the presence of alterations in the venous drainage (p < 0.001, OR = 0.26, 95% CI 0.12–0 0.55), as venous ectasia (p < 0.001, OR = 0.26, 95% CI 0.13–0.51) were found to be protective factors for the probability of bAVMs rupture. Instead, the presence of a prenidal aneurysm was found to be a factor strongly associated with the probability of bAVMs rupture (p = 0.015, OR = 3.11, 95% CI 1.25–7.73).

Predictive model

Multivariate logistic regression analysis (Table 3) suggested that there were four independent factors related to bAVMs rupture suggested that there were four independent factors related to bAVMs rupture: the major diameter of the nidus, the presence of venous drainage alterations, and the presence and type of aneurysms. This final model was the best predictive model, that is, the one that fitted well (Hosmer & Lemeshow test, p = 0.113) and had a good discrimination capacity (ROC = 0.83), allowing to detect three independent angio-architectural features mainly related to bleeding in bAVMs: a smaller diameter of the nidus (p < 0.001), the absence of venous drainage alterations (p = 0.047), and the presence of prenidal aneurysms (p = 0.005).

Table 3.

Multivariate logistic regression analysis.

	Odds ratio	95% Conf. Interval		p-value
Major diameter of the nidus	0.54	0.39	0.75	<0.001
Venous drainage alterations	0.40	0.16	0.99	0.047
Aneurysms type
Intranidal	1.16	0.42	3.21	0.611
Prenidal	4.69	1.58	13.89	0.005
Mixed	1.21	0.364	4.09	0.756
No aneurysm	1 (base)
Constant	5.64	2.29	13.88	< 0.001

Discussion

It should be first mentioned that the analysis of the differences of unruptured and ruptured bAVMs groups showed statistically significant differences into groups’ ages and lots of angio-architectural features (Table 1). In fact, in our series, ruptured bAVMs were more frequently detected in older patients, supporting the hypothesis that bAVMs are not static in their shape and in hemodynamic, but evolve during lifetime.^{7–10, 12} Among the angio-architectural features of bAVMs between ruptured and unruptured groups, we found a statistically significant differences into several features, both into the arterial, nidus, and venous portions. These statistically significant differences among ruptured and unruptured bAVMs support the hypothesis that angio-architectural features should be considered during treatment decision, both for newly discovered bAVMs and during follow-up.^12–15

As reported in the results, the logistic regression analysis demonstrates that for every one unit increase of the diameter of the arterial feeder measured 1 cm before the nidus, of the maximum diameter and of the volume of the nidus, and of the number of the draining veins, the odds of bAVMs to being rupture decrease (Table 2). This could be explained by the fact that all those factors together are related to a hemodynamical stability into bAVMs. In fact, if the arterial feeders, the size/volume of the nidus and the number of draining veins increase, the system can support the increase amount of blood and flow enter in the bAVM. These results are in accordance with a study published by Sahlein et al., assessing that bAVMs with higher numbers of draining veins would be protected from hemorrhage, compared to those with single draining vein, because in the first group there is a lower outflow impedance. ¹⁵ We also found other venous features relates to the reduction of bAVM probability of rupture, as the presence of mixed venous drainage, alterations in the venous drainage, particularly venous ectasia (Table 2). These results can be explained by the fact that a greater venous drainage system of bAVMs can accommodate a greater amount of blood and flow, until a hemodynamical balance exists. However, in our study the only factor related to a greater probability of bleeding for AVMs was the presence of a prenidal aneurysm (Table 2). This result can be supported by the intrinsic weakness of a prenidal flow-related aneurysm, which lacks the arterial layers in a high-pressure system. ²³

Finally, the logistic multivariate regression analysis demonstrated four independent factors related to AVMs rupture (Table 3). In our model, both a larger diameter of the nidus and the presence of venous drainage alterations were protective factors (respectively, OR = 0.54 and OR = 0.40), whereas the presence of prenidal aneurysms was a factor increasing the likelihood of bAVMs bleeding (OR = 4.69). These results corroborate the fact that some of the angio-architectural features of brain AVMs are associated with higher probability of rupture and there is the needing to develop a scoring system to estimate bleeding risk, as it has been done for intracranial aneurysms.

To date, none of the different scores proposed for bAVMs assess the risk of rupture and thus, the needing for treatment. Therefore, it is difficult to decide towards the needing for treatment in patients with brain AVMs, as far as the risk-benefit balance between invasive treatment and follow-up relies mainly on single physicians’ experience. Further prospective studies are needed to develop a scoring system on angio-architectural features for brain AVMs to help physician in the decision towards treatment or conservative management for brain AVMs, and also to distinguish patients at low risk from patients at high risk of bleeding.

Limitations

Because of its retrospective nature, our study has several limitations. First, the lack of hemodynamical evaluation of the angio-architectural features, which is of a certainty relevance. Another limitation of this study is the small sample size, although consistent with previously published series. Further prospective studies with grater sample size are needed to confirm our results and eventually find new angio-architectural features, to detect the risk of bleeding of bAVMs, and eventually create a scoring system

Conclusions

Our study evaluated the angio-architectural features of brain AVMs to assess the probability of bleeding, to underly the importance of identifying features associated with higher risk of rupture that would require early and aggressive treatments. In our study, several angio-architectural features were related to an increased probability of rupture for bAVMs, among which the more relevant resulted in a small diameter of the nidus, the absence of venous drainage alterations and the presence of prenidal aneurysms.

ORCID iDs

Arianna Rustici https://orcid.org/0000-0002-3753-6189

Matteo Zoli https://orcid.org/0000-0001-8838-5408

Ciro Princiotta https://orcid.org/0000-0001-7018-5280