We retrospectively reviewed 793 bAVMs out of the Beijing Tiantan bAVMs prospective registration research database (NCT04572568) from September 2011 to October 2014. The inclusion criteria were as follows: (1) The last treatment was SRS, (2) patients underwent single-session SRS, and (3) patients with more than 2 years clinical and radiological follow-up. The exclusion criteria were as follows: (1) patients who have received intervention other than embolization prior to SRS, (2) patients receiving staged SRS or multiple SRS, and (3) patients missing critical baseline information. Written informed consent for collecting clinical information was obtained from each patient at admission. The study was carried out according to the Helsinki Declaration guideline.

Patients were categorized into two groups, combined upfront embolization and SRS (Em+SRS group) and SRS alone without prior embolization (SRS group).

Baseline demographic, clinical features, and imaging data were collected. The baseline clinical characteristics included age on admission, sex, onset manifestation (hemorrhage, seizure, neurofunctional deficits, and others). The hemorrhagic presentation was defined as hemorrhage that could be ascribed to AVM. In terms of morphological characteristics, deep location was defined as nidus involving basal ganglia, thalamus, or brainstem. The definition of other angioarchitecture features were consistent with the reported terminology provided by the joint committee led by the American Society of Interventional and Therapeutic Neuroradiology (6). Nidus volume was calculated by the ABC/2 method on DSA (7). The Spetzler–Martin Grading System (SM), Virginia Radiosurgery AVM Scale (VRAS), and Modified Radiosurgery-Based AVM Score (mRBAS) were used to predict the long-term neurofunctional outcomes (8–10).

Clinical follow-up was conducted at the first 3–6 months and annually after discharge by clinical visit and telephone interview, and researchers who performed clinical follow-up assessments were blinded to the treatment modalities. In terms of imaging follow-up, magnetic resonance imaging (MRI) was routinely performed semiannual for the first 2 years after SRS and annually thereafter. Confirmatory digital subtraction angiography (DSA) was recommended to patients with complete obliteration on follow-up MRI. AVM obliteration was defined as a lack of abnormal flow voids on MRI or an absence of anomalous arteriovenous shunting on DSA.

The primary outcome was defined as AVM obliteration confirmed by MRI or DSA. The secondary outcomes comprised functional status, post-SRS hemorrhage, all-cause mortality, radiation-induced changes (RIC), including edema and cyst formation. The functional status was assessed by modified Rankin Scale (mRS) score system (favorable: 0–2, poor: 3–6).

The intraoperative embolization strategy depends on the consensus reached by a multidisciplinary meeting composed of senior neurointerventionists and neuroradiologists. A biplane angiography system was used (Siemens, Germany or Philip, Netherland), and the endovascular embolization was performed after induction of general anesthesia. In order to reduce the lesion volume and the risk of hemorrhage, we tend to embolize the lesion supplied by the main feeding artery or target embolization of the high-risk bleeding factors, such as aneurysm, arteriovenous fistula, etc. The main polymeric embolic agent used in this series is Onyx 18 (eV3, Inc.), which contains 6% ethylene vinyl alcohol and 94% dimethyl sulfoxide.

Stereotactic planning neuroimaging results are imported into Leksell Gamma-Plan workstation (Elekta AB, Elekta Company, Stockholm, Sweden) for definition and dose planning. T1 contrast-enhancement sequence and T2 sequence on 3D stereotactic MRI were used to delineate the radiation target. Dose planning was based on the location and volume of bAVM.

Categorical variables are presented as counts (with percentages); continuous variables are presented as the mean ± standard deviations (SD). A 1:1 PSM (with a caliper of 0.02 standard deviations) was performed to match the two groups with similar baseline data, such as age, gender, hemorrhage, volume, location, angioarchitecture, maximum and margin dose, and mRS at admission, mRBAS and VRAS) Absolute standardized differences was used to verify the matching results (Figure 1). At the comparison of baseline characteristics between Em+SRS and SRS, Pearson chi-square test or Fisher exact test was used to compare the categorical variables, and the two-tailed t-test was employed to compare the continuous variables (normal distribution variables). Wilcoxon rank-sum test was applied to compare non-normal distribution continuous variables. Univariable binary logistic regression analysis was applied to assess the odds ratios (ORs) and associated 95% confidence intervals (CI) of outcomes between these two groups. The cumulative rates of obliteration, post-SRS hemorrhage, and all-cause mobility were compared between the two groups using Kaplan–Meier survival analysis (log-rank test) to assess the hazard ratios (HRs) and associated 95% CI. Considering the difference in follow-up time and the cumulative effect of time between the two groups, we used the 5 years and the last follow-up two times nodes to calculate the p-value in the above survival analysis. A value of p < 0.05 was considered to be statistically significant. Statistical analysis was performed using SPSS (version 25.0, IBM, New York, USA).

A total of 152 patients were included in this study after PSM analysis (Table 1). The mean age was 29.8 ± 13.5 years old, and 93 (61.2%) patients presented with hemorrhage. One hundred thirty-two (86.8%) lesions were supratentorial, and 35 (23.0%) niduses were classified as deep locations. The mean nidus volume was 12.4 ± 18.4 cm³. Half of the AVMs (77, 50.7%) were Spetzler–Martin (SM) grades I–II. The most common VRAS score noted were VRAS = 2 (48 cases, 31.6%) and VRAS = 3 (51 cases, 33.6%), followed by VRAS = 4 (36 cases, 23.7%). The mean mRBAS score was 2.0 ± 1.9. The median interval between the embolization and SRS was 0.2 (interquartile range = 0.8) years. In terms of intraoperative details, the mean margin dose was 16.7 ± 3.1 Gy, and the maximum dose was 33.2 ± 5.0 Gy. Finally, the mean clinical follow-up duration was 6.2 ± 3.2 years, and the radiographic follow-up lasted for an average of 2.9 ± 2.3 years. After PSM, there were no statistical differences in baseline characteristics between the SRS group and Em+SRS group.

After an average of 2.9 ± 2.3 years of radiological follow-up, 58 (38.2%) patients achieved complete obliteration, and the obliteration rate was similar between SRS and Em+SRS (44.7 vs. 31.6%; OR, 1.754; 95% CI, 0.905–3.401; p = 0.096) (Table 2), and among these patients, 33 (56.9%) patients were confirmed by DSA (Figure 2). However, in a further analysis, the SRS group was superior to the Em+SRS group in terms of cumulative occlusion rate at a follow-up of 5 years (HR, 1.778; 95% CI, 1.017–3.110; p = 0.033) (Figure 2A).

At secondary outcomes, 143 (94.1%) patients achieved favorable functional state, and the favorable functional state was similar between SRS and Em+SRS (96.1 vs. 92.1%; OR, 2.086; 95% CI, 0.502–8.665; p = 0.312). Ten (6.6%) patients suffered subsequent hemorrhages after treatment, the risk of post-SRS hemorrhage was similar between these two groups (3.9 vs. 9.2%; OR, 0.405; 95% CI, 0.101–1.630; p = 0.203). In terms of cumulative post-SRS hemorrhage rate, it was 1.32, 5.11, and 5.11% at 2, 4, and 6 years in the SRS group, and 1.32, 6.77, and 12.48% at 2, 4, and 6 years in the Em+SRS group (HR, 0.437; 95% CI, 0.127–1.510; p = 0.217) (Figure 2C). Four (2.6%) patients died during clinical follow-up, one (1.3%) patient in the SRS group and three (3.9%) patients in the Em+SRS group (OR, 0.324; 95% CI, 0.033–3.191; p = 0.334). The cumulative all-cause mortality had no statistical differences between the two groups in the Kaplan–Meier analysis (log-rank, p = 0.336) (Figure 2B). In addition, edema or cyst after SRS was also similar between these two groups (log-rank, p = 0.991) (Figure 2D). In addition, it should be mentioned that none of the patients had serious embolic complications after upfront embolization.

We conducted a subgroup analysis based on hemorrhage presentation to further identify possible prognostic differences between SRS and Em+SRS. The baseline characteristics of the ruptured subgroup and unruptured subgroup are shown in the Supplementary Tables 1, 2.

In the ruptured subgroup (n = 93) (Table 3), 41 (44.1%) patients achieved complete obliteration, and the SRS group could achieve higher obliteration rate than the Em+SRS group (56.5 vs. 31.9%; OR, 2.773; 95% CI, 1.190–6.464; p = 0.018). In the Kaplan–Meier analysis, the cumulative obliteration rate at 5 years was also higher in the SRS group than in the Em+SRS group (64.5 vs. 41.3%; HR, 2.012; 95% CI, 1.037–3.903; p = 0.038) (Figure 3A). In terms of secondary outcomes, including favorable functional state (95.7 vs. 89.4%; OR, 2.619; 95% CI, 0.482–14.243; p = 0.265), post-SRS hemorrhage (4.3 vs. 10.6%; OR, 0.382; 95% CI, 0.070–2.076; p = 0.265), all-cause mortality (0.0 vs. 4.3%; p = 0.495), edema, and cyst were all similar between these two interventional strategies. In addition, the cumulative post-SRS hemorrhage rate were similar between the two groups in the ruptured subgroup (log-rank, p = 0.221; Figure 3C).

Among 59 patients with unruptured AVMs (Table 4), the overall obliteration rates and the cumulative obliteration rate were similar between the two groups (p = 0.711, p = 0.671, respectively) (Figure 3B). There was no significant difference between the secondary prognostic parameters of the two groups (favorable functional state, p = 0.981; post-SRS hemorrhage, p = 0.542; all-cause mortality, p = 0.981; edema, p = 0.981). In addition, the cumulative post-SRS hemorrhage rate were similar between the SRS and Em+SRS in the unruptured subgroup (log-rank, p = 0.629) (Figure 3D).

Despite our best efforts to improve the design defects of previous researches by adjusting for baseline differences and selection biases using PSM. There were still several limitations: the main limitations were its retrospective nature. Some patients cannot maintain strict and regular imaging follow-up, so we cannot know the exact time of nidus occlusion in these patients. Another limitation is that not all patients had DSA to confirm obliteration. In order to reduce the deviation caused by this limitation, we commissioned two senior neuroradiologists to evaluate the last radiographic follow-up independently, and a third more senior professor-level experts will reevaluate if the result is controversial.