All brain AVM patients who underwent resection procedure (including hybrid strategy and surgical resection alone strategy) in Peking University International Hospital between January 2016 and February 2018 were retrieved from our institutional database. The study was performed according to the guideline of the Helsinki Declaration and was approved by the ethics committee of Peking University International Hospital. The inclusion criteria were as follows: (1) Diagnosed with AVM by digital subtraction angiography (DSA) and/or magnetic resonance imaging (MRI); (2) Patients received surgical resection alone or single-stage combined embolization and resection; (3) Patients with SM grade III–V; (4) Followed-up for more than 1 year. Patients with multiple AVMs, hereditary hemorrhagic telangiectasia (HHT), or who underwent invasive procedures (microsurgery, embolization, radiosurgery) before admission were excluded from the study.

Demographic characteristics, clinical features, and imaging data were collected as baseline. Surgical resection alone was defined as resection without embolization. The definition of eloquent area and deep venous drainage were consistent with the Spetzler-Martin (SM) Grading system (2). Diffuseness was determined by preoperative angiograms with TOF images used to identify intervening brain parenchyma within the nidus. The degree of embolization was assessed by two professional neuroradiological physicians based on a three-stage grading system according to the changes of lesion volume and blood flow: Grade A, 0–30%; Grade B, 30–60%; Grade C, 60–100%. Disagreements in the embolization degree were arbitrated by a third physician. The intraoperative blood loss and operational duration were also collected. The perioperative complications such as postoperative hemorrhage, intracranial infection, and new-onset neurological dysfunction were evaluated within one week after the operation. The neurofunctional prognosis was analyzed based on modified Rankin Scale (mRS). The long-term outcomes were evaluated at the last follow-up. MRS > 2 was considered as postoperative neurological disability.

Follow-up was conducted within the first 3–6 months and annually after discharge through clinical visits and telephone interviews. Angiography was performed 1 year after the operation. The mRS scoring was conducted by neurosurgeons with at least 5 years' experience of clinical practice. All radiographic images were interpreted independently by at least two radiologists with equally at least 5 years of clinical experiences in the radiology center of our institute. Researchers who performed follow-up assessments were blinded to treatment modalities.

Neurosurgeons and interventional neuroradiologists prospectively planned the surgical procedure of each AVM at the multidisciplinary vascular conference held every weekday in our center. Large nidus (>3 cm), deep location, multi-source blood supply, and involved critical eloquent areas were considered as complex AVMs. After a comprehensive analysis of the angioarchitectural and hemodynamic characteristics, some complex cases were recommended for hybrid strategy (single-stage combined embolization and microsurgical resection).

The operation was performed in a hybrid angio-surgical suite consisting of a surgical microscope (Pentero 900, Carl Zeiss Surgical AG, Oberkochen, Germany) and a flat biplane panel of digital subtraction angiography unit (AlluraXper20, Philips Healthcare, the Netherlands). All procedures were performed after induction of general anesthesia. Systolic blood pressure was maintained between 120 and 140 mmHg. A 6-Fr catheter was used to access to lesions and support the diagnostic angiography and embolization. The main polymeric embolic agent used on the study participant was Onyx 18 (eV3, Inc.), which contains 6% ethylene vinyl alcohol and 94% dimethyl sulfoxide. A bifemoral approach was preferred when the nidus was supplied by two different circulations (for example, contralateral supply or both anterior and posterior circulation), together with 2 guiding catheters (Envoy 6F, Cordis) placed accordingly.

During the embolization, we preferred to embolize the following parts: (1) the largest and straightest feeding arteries and its supplying unit, (2) part with high rupture risk (flow-related aneurysms or high-flow arteriovenous fistula), (3) expanded deep perforating arteries, (4) extracranial blood supply from the dura mater. When intra-embolization rupture happens, XperCT (AlluraXper20, Philips Healthcare, the Netherlands) of the angiogram machine would be performed to detect the hemorrhage and external ventricular drainage (EVD) placement might be performed according to the hemorrhage location and intracranial pressure, and then, the subsequent resection and hematoma evacuation would be done instantly.

After the embolization, the femoral artery sheath was retained, the guide catheter was removed, and the injection of heparin saline was stopped. The microsurgical procedures would then be performed with the assistance of intraoperative neuronavigation, ultrasonography, indocyanine fluorescence angiography (ICG), continuous monitoring of electroencephalogram, and somatosensory evoked potential. Circumferential dissection was performed along the borders of AVM within the perinidal gliotic tissue. Complete resection of embolized component and residual lesions were attempted when safe. If electrophysiological monitoring abnormalities occurred during the resection, the embolized parts adjacent to the eloquent area were preferably retained. Repeated electrocoagulation of the severed end of the embolized feeding arteries was performed to avoid postoperative recanalization. After the resection and before closing the skull, cerebral angiograms through the retained femoral artery sheath were performed to identify the completely occlusion of the lesion. If residues were detected, then secondary resection would be performed immediately.

A standardized perioperative management protocol was applied to all cases, with postoperative systolic blood pressure maintained between 120 and 140 mmHg, and antiepileptic and anti-cerebral edema treatments performed routinely after the operation. All patients underwent a computed tomography (CT) scan at postoperative day 1 to detect whether postoperative acute hemorrhage or infarction occurred. A MRI scan was performed at postoperative day 3. Repeat CT or MRI scans were performed if patients experienced severe headache or postoperative neurological deterioration.

The categorical variables were presented as counts (with percentages), whereas the continuous variables are presented as the means ± standard deviations (normal distribution variable) or median (quartile) (non-normal distribution variable). Two-tailed t-tests or one-way ANOVA were employed to compare normal distribution continuous variables, and Mann–Whitney U-test was applied to compare non-normal distribution continuous variables. The Pearson chi-square test or Fisher exact test were used to compare categorical variables as appropriate. A stratified analysis of pediatrics and adults was conducted to compare the intraoperative blood loss. Bonferroni correction was adopted in the adjusted post-hoc analysis in order to avoid the occurrence of Type I errors in the subgroup analysis of embolization degree (P < 0.005). Odds ratios (ORs) and 95% confidence intervals (CIs) for potential risk factors of postoperative disability in the hybrid group were calculated by univariate and multivariable logistic regression analyses. P < 0.05 was considered to be statistically significant. Statistical analyses were performed using SPSS (version 25.0, IBM).

A total of 100 consecutive AVM patients (47 corrected using hybrid procedures whereas 53 by surgical resection alone) were included in this study. The mean age was 30.1 ± 13.0 years (range 4.6–61.7 years). Clinically, more than half of the patients (53.0%) were presented with hemorrhage (hybrid vs. surgical resection alone: 55.3 vs. 50.9%). Fifty-nine percent were evaluated as SM grade III (hybrid vs. surgical resection alone: 55.3 vs. 62.3%). In the hybrid group, 33 patients had a reduced SM grade after embolization (17 cases were SM grade II, 24 cases were SM grade III, and 6 cases were SM grade IV). The overall mean mRS scores at admission was 1.2 ± 0.5 (hybrid vs. surgical resection alone: 1.2 ± 0.4 vs. 1.2 ± 0.5). The average size of the nidus was 4.7 ± 1.5 cm (range 2.09–8.62 cm) (hybrid vs. surgical resection alone: 4.7 ± 1.3 vs. 4.8 ± 1.7 cm). Because of the priority of hybrid strategy for deep lesions, the hybrid group was more likely to be accompanied with deep venous drainage (hybrid vs. surgical resection alone: 57.4 vs. 37.7%, P = 0.049). The other angiographic characteristics between these two groups were similar (Table 1).

Continuous clinical or angiographic follow-up was performed for an average of 2.3 ± 0.6 years (range 1.2–3.3 years) for all the patients. In the ruptured cohorts, 26 patients (55.3%) underwent hybrid procedures, whereas 27 patients (50.9%) underwent surgical resection alone. Hybrid procedures (embolization+resection) lasted significantly longer than the surgical resection alone (7.3 ± 2.5 vs. 5.7 ± 2.0 h, P = 0.012; odds ratio, 95.918; 95% CI, 21.945–169.892). However, the actual surgical resection for the hybrid strategy was significantly shorter (3.7 ± 1.2 vs. 5.7 ± 2.0 h, P = 0.000; odds ratio, 119.735; 95% CI, 66.247–173.223). There were no significant differences in intraoperative blood loss (P = 0.059) and short-term outcomes (P = 0.453) between the two strategies. In the pediatric subgroup, the intraoperative blood loss was also similar between these two strategies (ruptured: P = 0.254; unruptured: 0.629). Intraoperative angiography revealed residual lesion in 1 patient (3.8%) in the hybrid group, which was successfully corrected by a subsequent resection before the skull closure of the first resection. During the perioperative period, 3 patients (11.5%) in the hybrid group presented with delayed postoperative hemorrhage. One of them received emergency hematoma evacuation surgery, and no recanalization was observed, and the other two patients received conservative medical treatment. In addition, one other patient in the surgical resection alone group experienced postoperative hemorrhage arising from residual lesion (found by postoperative DSA), and this patient subsequently received conservative treatment. Overall, after follow-up of on average 2.3 years, no significant differences were found regarding long-term mRS (0.9 ± 1.0 vs. 1.1 ± 1.1, P = 0.373) and long-term obliteration rates between these two groups (100.0 vs. 96.2%, P = 1.000) (Figure 1). Eight patients (15.1%) in the ruptured cohorts suffered long-term disabilities (mRS > 2) (hybrid surgery vs. surgical resection alone: 3 [11.5%] vs. 5 [18.5%], P = 0.745).

For the cohort with unruptured AVM, 21 cases (44.7%) received hybrid procedures whereas 26 patients (49.1%) underwent surgical resection alone. Although the total surgical duration for the two strategies was similar (P = 0.615), the actual resection duration for the hybrid group was significantly shorter than the surgical resection alone (3.5 ± 1.4 vs. 5.7 ± 2.7 h, P = 0.002; odds ratio, 132.447; 95% CI, 52.994–211.899). There was no significant difference between both short-term (1.5 ± 0.9 vs. 1.7 ± 1.2, P = 0.599) and long-term mRS (0.5 ± 1.0 vs. 0.8 ± 1.4, P = 0.313) and long-term disability (mRS > 2, 9.5 vs. 7.7%, P = 1.000) between these two groups. One patient in the surgical resection alone group succumbed to intraoperative hemorrhagic shock (intraoperative blood loss: 11,500 ml) and died (Table 2).

Univariate analysis revealed that diffuse nidus (P = 0.002) and deep perforating arteries (P = 0.012) contributed to long-term disability (mRS > 2) of the 5 patients (10.6%) in the hybrid group. Excessive embolization (Grade C: embolize 60–100%) may be associated with long-term disability (P = 0.073). However, multivariable logistic regression analysis revealed that excessive embolization (P = 0.041; odds ratio, 24.000; 95% CI, 1.140–505.194) was significantly associated with long-term disability in the hybrid group (Table 3). Further subgroup analysis showed that the embolization degree significantly correlated with intraoperative blood loss (P = 0.033). Compared with minimal embolization (Grade A, 0–30%), moderate embolization (Grade B, 30–60%) could significantly decreases intraoperative blood loss (P = 0.009; odds ratio, 157.64; 95% CI, 110.068–745.487; Bonferroni correction: P = 0.018) (Figure 2). No significant differences were found in short-term (P = 0.213) and long-term mRS (P = 0.228) between different embolization degrees (Table 4). However, there was a trend of higher disability rate in the excessive embolization group (Grade C; 20.0 vs. 0–5.6%, P = 0.128).

In the surgical resection alone group, 7 patients (13.2%) suffered long-term disability (mRS > 2). Deep venous drainage (P = 0.048) and deep perforating arteries (P = 0.003) were significantly associated with long-term disability in the univariate analysis. Multivariable logistic regression analysis revealed that deep perforating arteries (P = 0.025; odds ratio, 15.389; 95% CI, 1.412–167.669) were significantly associated with long-term disability (Table 3).

Currently, intraoperative cerebrovascular angiography has two applications in hybrid surgery: (1) intraoperative diagnostics and (2) intraoperative embolization. In general, the intraoperative diagnostic angiography was considered to be effective in improving the obliteration rate of lesions and offering fast quality control (19). However, studies on intraoperative embolization strategies are limited (Table 4). At present, there are three main types of intraoperative embolization procedures: (1) nidus embolization, (2) nidus and feeding artery embolization, and (3) retrograde venous embolization. Many previous studies reported that the nidus+feeding artery embolization could provide a bloodless surgical plane for surgical resection, and the embolic agents could be used to mark arterial feeders during the resection procedure (9, 12, 15). Besides, intraoperative retrograde venous embolization was an emerging hybrid strategy for treatment of cerebral AVMs and has been postulated to achieve a high rate of complete AVM obliteration and excellent functional outcomes (11, 14). In this study, we adopted the nidus+feeding artery embolization strategy. After embolization with Onyx, the embolized lesions appear black, and the feeding arteries and lesion boundary could easily be identified in the subsequent resection procedure.

To our knowledge, the relationship between the embolization degree and clinical prognosis for the hybrid surgery has not been fully evaluated. One previous study suggested that simultaneous occlusion of more than 60% of AVM volume could induce significant redistribution of local blood flow (20), which might develop secondary NPPB and hyperemia of the adjacent brain (21). Besides, the increased pressure of feeding artery after embolization was reported to be proportional to the degree of AVM obliteration (22). Excessive blood stasis might occur in the residual lesions when the increased feeding artery pressure transmitted to the residual nidus, which might increase the intranidal pressure and finally lead to the subsequent rupture. Therefore, limited embolization was recommended to avoid excessive high pressure in the feeding arteries (23). In this study, we found excessive embolization (Grade C, 60–100%) was significantly associated with the postoperative long-term disabilities in the SM grade III–V AVMs. The underlying mechanism might be the unpredictable acute neurological impairment caused by the intraoperative rupture during the embolization, or difficulty in resecting deep lying lesions caused by the constraint of lifting the rigid embolized nidus, or the intraoperative hemostasis difficulty caused by congestion in the remaining lesions after embolization, which damages more functional brain tissues. In addition, the moderate embolization (Grade B, 30–60%) eases subsequent resection by effectively reducing intraoperative blood loss.

Surgical resection is the most critical step in hybrid surgery, and it is controversial whether the embolized lesions should be removed. Kocer et al. (9) suggested that if deemed safe, then the embolized component of the lesion should be completely removed. However, the excision of embolic lesions near the eloquence presents a high risk of neurological impairment. Accordingly, we suggest retention of a properly embolized lesions to maintain a sufficient lesion-to-eloquence distance as recommended in a previous study (15). Even so, there is a 0 to 13% risk of recanalization in the embolized occluded lesion after the operation (24–26). In this study, we retained the embolized occluded lesions with critical functions monitored by the intraoperative electroencephalogram and somatosensory evoked potential, and the embolized artery residues were repeatedly coagulated to avoid recurrence of AVM. And as expected, no recurrence was observed at the last follow-up.

Our study design presented potential limitations. Firstly, because this is a single-center retrospective study, the operative indicators and intraoperative manipulation may vary across institutions. In addition, although the degree of intraoperative embolization in this study was evaluated by three experienced neuroradiologists based on the changes in lesion volume and blood flow, the results were still relatively subjective and lacked quantitative measures. Subsequently, we recommend for further multicenter prospective trials on hybrid strategy to explore the optimal intervention approach for AVMs.