Anatomical and clinical factors associated with multiple-stage embolization of brain arteriovenous malformation as a stand alone treatment

Abstract

Introduction: Embolization is one of the modalities used to treat brain arteriovenous malformations (bAVMs), which often require a multiple-stage approach. Factors influencing the need for a multiple-stage embolization are poorly defined and not very well understood. The aim of this study was to establish those factors. Patients and methods: We analysed 117 consecutive patients undergoing embolization due to bAVM with intention to treat. Patients undergoing embolization prior to surgery and with history of previous interventions or stereotactic radiosurgery were excluded. We used univariate and multivariate logistic regression analysis to establish possible predictors of the need for a multiple-stage embolization. Results: In total, 50 patients (42.73%) required more than one embolization procedure, with a mean of 1.67 sessions per patient to achieve complete obliteration. A history of hemorrhage was less frequent among patients requiring staged embolization (29.55% vs. 52.38%; p = 0.019). In this group, arterial supply more commonly originated from the posterior cerebral artery (53.19% vs. 32.31%; p = 0.027) and the anterior choroidal artery (10.64% vs. 0.00%; p = 0.007). Lesions located in the parietal lobe were also more frequently associated with the need for multiple-stage treatment (18.37% vs. 6.25%; p = 0.046). In multivariable logistic regression analysis, nidus size < 3 cm was independently associated with a lower likelihood of requiring multistage embolization (OR 0.06; 95% CI 0.02–0.23; p < 0.001), whereas parietal lobe location remained an independent predictor of staged treatment (OR 8.34; 95% CI 1.34–51.84; p = 0.023). Conclusions: Multistage embolization was mainly determined by anatomical factors, particularly lesion size and parietal location, rather than clinical presentation. These findings support an individualized, anatomy-based approach to endovascular treatment of brain arteriovenous malformations.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-54611-z.

Introduction

Brain arteriovenous malformations (bAVMs) are complex vascular lesions characterized by direct arteriovenous shunting without an intervening capillary network, resulting in high-flow and low-resistance circulation. Owing to their low prevalence, estimated at approximately 1.4 per 100,000 individuals, optimal management strategies remain challenging and heterogeneous¹. Currently, three main treatment modalities are available: microsurgical resection, endovascular embolization, and radiosurgery². Among them, endovascular embolization has gained an important role, offering a minimally invasive approach with the potential for immediate hemodynamic effect, particularly in selected or high-risk cases³.

Nevertheless, embolization as a stand-alone treatment is appropriate only in selected patients, as emphasized in current consensus recommendations⁴. Reported rates of complete obliteration following embolization vary widely, ranging from 5% to 70%, reflecting substantial heterogeneity in lesion characteristics and treatment strategies. Superficially located arteriovenous malformations are more likely to be cured with fewer procedures, whereas more complex lesions often require staged interventions⁵. Indeed, the average number of embolization sessions needed to achieve complete occlusion ranges from 1.28 to 2, although up to 80% of patients may be successfully treated in a single session^4–6.

Despite ongoing advances in endovascular techniques, the risk of complications remains non-negligible. Reported complication rates range from as low as 0.8% to as high as 24.1% in pooled analyses⁷. Consequently, careful patient selection and procedural planning are essential. While several studies have investigated predictors of successful embolization or spontaneous occlusion, data specifically addressing factors associated with the need for multiple-stage embolization remain limited. Prior work by Sultan et al. demonstrated that spontaneous occlusion after partial embolization may be associated with smaller nidus size, superficial venous drainage, and limited arterial supply, particularly in cases with delayed venous outflow⁸. However, determinants guiding the necessity for staged embolization rather than single-session treatment remain insufficiently defined.

Multiple-stage embolization has been proposed to reduce procedural risk by limiting embolic load, preserving venous outflow, and allowing hemodynamic adaptation. Prior studies suggest improved safety in complex AVMs, particularly those with deep feeders, large nidus size, or eloquent location⁹.

Therefore, the aim of the present study was to identify clinical and anatomical factors associated with the requirement for multiple-stage endovascular embolization in patients with brain arteriovenous malformations treated with curative intent.

Patients and methods

We collected retrospective data for all consecutive patients who were hospitalized between 2018 and 2024 in a tertiary neurosurgical centre and underwent endovascular treatment of bAVM with an intention to cure. Patients were eligible for inclusion in the study if they met all of the following criteria: (1) bAVM confirmed with computed tomography angiography or magnetic resonance angiography, (2) endovascular embolization with curative intent as the primary treatment modality, (3) embolization performed using Onyx (ethylene–vinyl alcohol copolymer) as the embolic agent, (4) endovascular treatment performed via a transarterial approach, (5) availability of complete clinical, radiological, and procedural data allowing assessment of AVM characteristics and treatment course with full follow-up of 6 months after last treatment. Patients that were excluded from the study when one or more of the following criteria was met: (1) embolization performed during the acute phase of hemorrhage, defined as within 24 h from symptom onset, (2) previous stereotactic radiosurgery for treatment of the arteriovenous malformation, (3) preoperative embolization performed as part of a planned surgical resection strategy rather than as definitive therapy, (4) history of prior intracranial interventions, (5) presence of other intracranial pathologies that could influence treatment outcomes, (6) incomplete clinical, radiological, or procedural data precluding reliable analysis.

Patient data, including demographics and vascular risk factors, history of bleeding, and history of epilepsy, were assessed based on medical records. Factors such as presence of coexisting aneurysms, origin of the feeding arteries, the anatomical location of the bAVM, nidus size, venous drainage, draining vein stenosis or draining vein ectasia, as well as the pattern of venous drainage, were determined based on imaging data. The Spetzler-Martine grade was calculated on baseline, using angiography performed before the first embolization. All images were assessed by two neuroradiologists. In case of discrepancy, a consensus decision was made. Intra-observer reliability was not calculated.

Embolization technique

All AVM embolization procedures were performed using a standardized technique and constituted the primary treatment approach. Under general anesthesia, the vascular access was obtained via transfemoral route, using the Seldinger technique. After advancing the microcatheter and positioning its tip in close proximity to the AVM nidus, Onyx 18 was administered. Onyx is a liquid embolic agent composed of an ethylene–vinyl alcohol copolymer (EVOH) dissolved in dimethyl sulfoxide (DMSO), with tantalum powder added to provide radiopacity. Following embolization, a control digital subtraction angiography was performed to evaluate the extent of nidus occlusion. Hemostasis at the vascular access site was achieved using a compression dressing. The need for multiple-stage embolization was operationally defined as the requirement for more than one endovascular session to achieve complete angiographic obliteration of the arteriovenous malformation. In our institutional practice, staged embolization was planned when the nidus diameter exceeded approximately 2.5 cm, reflecting concerns regarding procedural safety, embolic load, and hemodynamic tolerance during a single-session intervention. All AVMs were embolised using the arterial route. Complete obliteration was determined on the basis of control digital subtraction angiography at the final embolization stage.

Feeding arteries were additionally categorized into deep arterial supply and cortical/superficial arterial supply. The deep supply was defined as feeders originating from deep perforator or deep-seated intracranial arteries (e.g., anterior choroidal artery and/or posterior cerebral artery branches supplying deep structures). Extracranial arterial supply was not observed in the study cohort and therefore was not analyzed as a separate category.

Minor extracranial arterial feeders were occasionally observed; however, no AVM demonstrated a dominant extracranial arterial supply, and therefore extracranial supply was not analyzed as a separate category.

Ethics

The study was performed in accordance with the Declaration of Helsinki and the Good Clinical Practice (GCP). Each participant gave consent to participate. The study protocol was approved by Jagiellonian University Bioethical Committee. Approval ID number: 1072.6120.84.2024.

Statistical analysis

The database management and statistical analysis were performed with RStudio version 8.5 for Windows (RStudio, Inc, USA).

To describe the group, the elements of descriptive statistics such as percentage value, mean value and standard deviations were used. We used the Shapiro-Wilk test to assess normality. Non-normally distributed variables were normalized using linear or logarithmic scaling, as appropriate. We used the t-test for continuous variables and the χ2 or Fisher’s exact test for proportional variables. For multivariate analyses we used penalized logistic regression with the Least Absolute Shrinkage Selection Operator (LASSO) as a penalty function. All significance tests are two tailed and the p-value of < 0.05 has been considered significant. Continuous variables are presented as mean (standard deviation). For statistical analyses, AVM size was categorized using predefined thresholds (< 3 cm, 3–6 cm, > 6 cm), consistent with commonly used classifications and the Spetzler–Martin grading system.

The final multivariable logistic regression model was specified using a priori selected covariates, based on prior literature and clinical plausibility, regardless of univariate statistical significance. To avoid overfitting, the number of covariates was limited in accordance with the number of observed events.

Results

We analysed data of 117 consecutive patients (43.22% female, mean age 39.18 ± 15.15 years). Thirty-four patients were excluded from the analysis (Fig. 1). In total, 50 patients (42.73%) required more than one embolization procedure. An average of 1.67 procedures per patient were needed to achieve complete obliteration. In univariate analysis, we found a number of factors associated with the need for multiple-stage embolization. Patients necessitating staged embolization tended to be younger (32.70 ± 15.01 vs. 42.55 ± 16.51; p = 0.052). History of bleeding (subarachnoid, intracerebral and intraventricular haemorrhage) was less prevalent in those requiring staged embolization (29.55% vs. 52.38%; p = 0.019), while we also observed an insignificant trend towards the history of intraventricular bleeding being less prevalent in the studied group (13.95% vs. 28.33%; p = 0.084). When it comes to the source of feeders, patients requiring multiple-stage embolization significantly more often had feeders from the posterior cerebral artery (53.19% vs. 32.31%; p = 0.027) and the anterior choroidal artery (10.64% vs. 0.00%; p = 0.007). We also observed an insignificant trend towards more common feeders from the middle cerebral artery (68.09 vs. 49.23; p = 0.082). Patients with malformations located in the parietal lobe (18.37% vs. 6.25%; p = 0.046) and in the temporo-occipital area (8.16% vs. 1.56; p = 0.091) were more likely to need multi-staged embolization. Higher Spetzler-Martin grade (2.81 ± 1.12 vs. 2.23 ± 1.23; p = 0.012) was also associated with the need for a multiple-stage embolization. Moreover, patients requiring multiple-stage embolization had higher numbers of pedicles (3.14 ± 1.52 vs. 2.40 ± 1.15; p = 0.002). Detailed results are presented in Table 1. Mean fluoro time of the first procedure was 60.56 min for the multiple-stage embolization group and 39.81 min for the single-stage embolization group (p = 0.071). Cumulative mean fluoro time for multiple-stage embolization was 136.06 min. The average amount of Onyx used per session was 8.06 ml in the multiple-stage embolization group and 4.19 ml in the single-stage embolization group (p = 0.120). We observed no spontaneous obliterations.

Fig. 1.

Figure 1.

Table 1.

Baseline characteristics of patients according to treatment strategy.

Variable	Single-stage embolization (n = 67)	Multiple-stage embolization (n = 50)	p value
Female sex	30 (44.8%)	21 (42.0%)	0.760
Age, years (mean ± SD)	32.7 ± 15.0	42.6 ± 16.5	0.052
History of hemorrhage	35 (52.4%)	15 (30.0%)	0.019
Intracerebral hemorrhage	23 (34.3%)	10 (20.0%)	0.130
Intraventricular hemorrhage	19 (28.4%)	7 (14.0%)	0.084
Coexisting aneurysm	13 (19.4%)	12 (24.0%)	0.630
Middle cerebral artery feeder	33 (49.3%)	34 (68.0%)	0.082
Posterior cerebral artery feeder	22 (32.8%)	27 (54.0%)	0.027
Anterior cerebral artery feeder	8 (11.9%)	9 (18.0%)	0.451
Superficial temporal artery feeder	1 (1.5%)	0 (0.0%)	0.393
Vertebral artery feeder	1 (1.5%)	0 (0.0%)	0.393
Pericallosal artery feeder	18 (26.9%)	12 (24.0%)	0.740
Middle meningeal artery feeder	5 (7.5%)	4 (8.0%)	0.875
Anterior choroidal artery feeder	0 (0.0%)	5 (10.0%)	0.007
Occipital artery feeder	2 (3.0%)	2 (4.0%)	0.740
Cerebellar artery feeder	11 (16.4%)	9 (18.0%)	0.676
Deep arterial supply	27 (54.0%)	41 (61.2%)	0.389
Superficial arterial supply	23 (46.0%)	26 (38.8%)	0.472
Number of pedicles	3.18±1.47	2.40±1.15	0.015
AVM size <3 cm	43 (64.2%)	9 (18.0%)	<0.001
AVM size 3–6 cm	18 (26.9%)	32 (64.0%)	<0.001
AVM size >6 cm	7 (10.4%)	10 (20.0%)	0.151
Spetzler–Martin grade I	10 (16.9%)	2 (4.0%)	0.062
Spetzler–Martin grade II	20 (33.9%)	15 (30.0%)	0.837
Spetzler–Martin grade III	19 (32.2%)	15 (30.0%)	1.000
Spetzler–Martin grade IV	7 (11.9%)	13 (26.0%)	0.051
Spetzler–Martin grade V	3 (5.1%)	3 (6.0%)	1.000
Spetzler-Martine grade±SD	2.81±1.12	2.23±1.23	0.012
Eloquent brain area	21 (41.18%)	35 (52.50%)	0.331
Deep drainage	31 (61.22%)	36 (53.13%)	0.389
Epilepsy	14 (28.57%)	14 (21.58%)	0.549
Groin hematoma	0	1 (2.00%)	1.000
Postprocedural seizure	1 (1.49%)	0	0.427
Visual disturbance	0	1 (2.00%)	1.000
Paresis	1(1.49%)	0	0.427

Periprocedural complications

Periprocedural complications were observed in a small proportion of patients. Overall, complications occurred in 4 out of 117 patients (3.4%). These included transient neurological deficits, such as focal neurological symptoms and visual disturbances, postoperative seizures and one case of hematoma in the groin. No cases of procedure-related mortality were observed. Importantly, no permanent ischemic deficits or procedure-related deaths occurred. Due to the low number of adverse events, no statistically meaningful comparison between the single-stage and multiple-stage embolization groups can be performed.

Multivariable analysis

In the prespecified multivariable logistic regression model, adjusted for clinically relevant covariates selected a priori, AVM size < 3 cm was independently associated with a significantly lower likelihood of requiring multiple-stage embolization (OR 0.06, 95% CI 0.02–0.23; p < 0.001) (Table 2).

Table 2.

Prespecified* multivariable logistic regression analysis of factors associated with the need for multiple-stage embolization.

Variable	Odds Ratio (OR)	95% Confidence Interval	p-value
AVM size <3 cm	0.06	0.02 – 0.23	<0.001
Parietal lobe localization	8.34	1.34 – 51.84	0.023
Spetzler–Martin grade	0.59	0.33 – 1.04	0.068
Deep arterial supply (PCA and/or AChA)	2.34	0.90 – 6.10	0.081

*Adjusted model included the following prespecified covariates selected a priori based on clinical relevance and prior literature: AVM size <3 cm, parietal lobe localization, Spetzler–Martin grade, and deep arterial supply (defined as PCA and/or anterior choroidal artery feeders). Variables were not selected solely on the basis of univariate statistical significance.

In contrast, parietal lobe localization was independently associated with a higher likelihood of multiple-stage embolization (OR 8.34, 95% CI 1.34–51.84; p = 0.023).

The Spetzler–Martin grade showed a trend toward an association with multiple-stage embolization but did not reach statistical significance in the adjusted model (OR 0.59, 95% CI 0.33–1.04; p = 0.068). Similarly, the presence of deep arterial supply, defined as feeders originating from the posterior cerebral artery and/or anterior choroidal artery, demonstrated a non-significant trend toward increased odds of staged embolization (OR 2.34, 95% CI 0.90–6.10; p = 0.081).

The number of covariates included in the model was limited to avoid overfitting in relation to the number of observed events. The direction and magnitude of the observed effects were consistent with clinical plausibility and prior literature.

Discussion

In this study, nearly half of the patients undergoing endovascular treatment for brain arteriovenous malformations required a multiple-stage embolization approach. The need for staged treatment was primarily associated with anatomical and angioarchitectural factors, including lesion size, arterial supply, and localization, rather than technical limitations of the procedure itself. These findings support the concept that multiple-stage embolization reflects lesion complexity rather than treatment failure.

Patients requiring multiple embolization sessions more frequently exhibited arterial supply from the posterior cerebral or anterior choroidal arteries and had lesions located in eloquent or deep brain regions. Such vascular configurations are known to increase procedural complexity and limit the safety of aggressive single-session embolization, thereby favoring a staged approach. Similar observations have been reported in prior studies highlighting the role of deep arterial feeders and complex angioarchitecture in determining treatment strategy^10,12.

Consistent with previous reports, higher Spetzler–Martin grade was associated with a greater likelihood of multiple-stage embolization, whereas smaller AVMs were more often treated successfully in a single session^9,17. Importantly, although patients in the multiple-stage group were less likely to present with hemorrhage, this finding should not be interpreted as a protective effect of staged embolization. Ruptured AVMs tend to be smaller and more superficial, characteristics that favor single-session treatment and likely explain this association^9,10,14.

Several studies have shown that excessive embolic volume and extensive nidus penetration during a single procedure increase the risk of hemorrhagic complications¹¹. Consequently, staged embolization may represent a safer strategy in anatomically complex lesions by allowing gradual flow reduction and limiting hemodynamic stress. In line with this, prior reports have demonstrated that partial embolization may result in delayed or spontaneous occlusion in selected cases, particularly in lesions with favorable venous drainage patterns⁸.

Our findings are also consistent with previous observations indicating that AVMs supplied by deep or eloquent arterial territories and those with higher Spetzler–Martin grades are more likely to require staged treatment^10,17. Furthermore, studies have shown that smaller nidus size, superficial location, and fewer feeding arteries are associated with higher rates of complete obliteration following embolization^17–19. These data collectively support a tailored approach to AVM management based on individual anatomical characteristics.

Spontaneous or delayed complete occlusion of brain arteriovenous malformations following partial embolization has been reported but appears to be relatively uncommon. Most series describe this phenomenon as rare, with reported rates ranging from approximately 0.8% to 2%²⁰. However, higher rates have been observed in selected cohorts. For example, Sultan et al. reported spontaneous complete occlusion in up to 12.85% of partially embolized AVMs during follow-up, particularly in lesions with favorable angioarchitectural features such as limited nidus size and specific venous drainage patterns²¹. These findings suggest that while spontaneous obliteration after partial embolization is not common, it may occur in selected cases and should be considered when planning staged treatment strategies. In our study we did not observed any spontaneous obliterations.

Lastly procedural planning in our center favored staged embolization for nidus diameters exceeding approximately 2.5 cm, the analytical size categories were defined independently, and the observed association between AVM size < 3 cm and single-session embolization reflects an intrinsic anatomical effect rather than procedural allocation bias. A study by van Rooji et al. demonstrated that AVMs thath can be cured with single session embolziation are 2.2 cm on average²². Plasenica et al. described that decreasing AVM < 2.5 cm decreases the risk of hemorrhage and makes it more suitable for radiosurgery²³.

Limitations

This study has several limitations that should be acknowledged. First, its retrospective and single-center design may limit the generalizability of the findings and introduce selection bias related to institutional treatment strategies and operator experience. Second, only patients treated with endovascular embolization with curative intent were included, which may not reflect the full spectrum of brain arteriovenous malformations encountered in multimodality treatment paradigms.

Third, the decision to perform staged embolization was partially based on clinical judgment and procedural safety considerations, including an institutional threshold of approximately 2.5 cm for nidus diameter. Although this reflects real-world practice, it may have influenced treatment allocation and should be considered when interpreting the association between AVM size and the need for multistage embolization. Importantly, AVM size categories used in the statistical analysis (< 3 cm, 3–6 cm, > 6 cm) were predefined and consistent with commonly used classifications, mitigating potential analytical bias.

Fourth, although a prespecified multivariable model was applied, residual confounding by unmeasured anatomical or hemodynamic factors cannot be excluded. Additionally, the relatively limited sample size, particularly in subgroup analyses, may have reduced statistical power to detect weaker associations.

Finally, angiographic follow-up was limited to a minimum of six months after the final embolization session, and longer-term outcomes such as delayed recanalization or late hemorrhage were not assessed. Future prospective, multicenter studies with standardized treatment criteria and extended follow-up are warranted to validate these findings and further refine predictors of staged endovascular treatment.

Conclusions

In endovascular embolization of bAVMs, a staged approach seems to be relatively often required, which was primarily associated with anatomical factors such as lesion size, arterial supply, and location. Larger AVMs and those supplied by posterior or deep arterial branches were more likely to require staged treatment, whereas smaller lesions were more often amenable to single-session embolization. The lower prevalence of hemorrhage in the multistage group likely reflects underlying lesion characteristics rather than a protective effect of staged treatment. These findings support an individualized, anatomy-driven approach to treatment planning in patients with brain arteriovenous malformations.

Taken together, our results emphasize that multiple-stage embolization should be viewed not as a limitation but as a deliberate strategy for safely managing complex AVMs. Individualized treatment planning—incorporating lesion size, arterial supply, venous drainage, and eloquence—may optimize outcomes and reduce procedural risk. Future prospective, multicenter studies with standardized follow-up are warranted to further refine patient selection and to evaluate long-term outcomes of staged versus single-session embolization strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 ^{(120.2KB, xlsx)}

Author contributions

Roger M. Krzyżewski- study conceptualization, study design, statistical analisis, manuscript writting and editing, fundingSandra A. Pilawska- data collection, data suppervisionGabriela A. Brożek- data collection, data suppervisionKornelia M. Kliś- study design, manuscript editing and corrections, statistical analysisBorys M. Kwinta- neurosurgical overview, data curractionKrzysztof Stachura- neurosurgical overview, data curractionTadeusz J. Popiela- Radiological subpervision, study overview, manuscript correctio and editing.

Funding

Study was supported by Jagiellonian University Statutory Funds number N41/DBS/001516.

Data availability

Data is available as a supplementary file. Full dataset available upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Human ethics and consent to participate

Approved by Jagiellonian University Bioethical Committee.