Long-term outcomes of 170 brain arteriovenous malformations treated by frameless image-guided robotic stereotactic radiosurgery

Abstract This study was conducted to report long-term outcomes of the frameless robotic stereotactic radiosurgery (SRS) for brain arteriovenous malformation (AVM) at Ramathibodi Hospital. Retrospective data of patients with brain AVM (bAVM), who underwent CyberKnife SRS (CKSRS) at Ramathibodi Hospital from 2009 to 2014, were examined. Exclusion criteria were insufficient follow-up time (<36 months) or incomplete information. Patients’ demographics, clinical presentation, treatment parameters, and results were analyzed. Excellent outcome was defined as AVM obliteration without a new neurological deficit. Risk factors for achieving excellent outcome were assessed. From a total of 277 CKSRS treatments for bAVM during the 6 years, 170 AVMs in 166 patients met the inclusion criteria. One hundred and thirty-nine cases (81.76%) presented with hemorrhages from ruptured bAVMs. Almost two-thirds underwent embolization before radiosurgery. With the median AVM volume of 4.17 mL, three-quarters of the cohort had single-fraction CKSRS, utilizing the median prescribed dose of 15 Gray (Gy). In the multisession group (25.29%), the median prescribed dose and the AVM volume were 27.5 Gy and 22.3 mL, respectively. An overall excellent outcome, at a median follow-up period of 72.45 months, was observed in 99 cases (58.24%). Seven AVMs (4.12%) ruptured after CKSRS but 1 patient suffered a new neurological deficit. Two patients (1.18%) were classified into the poor outcome category but there were no deaths. Negative factors for excellent outcome, by multivariate regression analysis, were the male sex and multisession SRS delivery, but not age, history of AVM rupture, previous embolization, or AVM volume. Despite relatively larger bAVM and utilizing a lower prescribed radiation dose, the excellent outcome was within the reported range from previous literature. This study offers one of the longest follow-ups and the largest cohorts from the frameless image-guided robotic SRS community.


Data collection
After approval by the Institutional Review Committee, a retrospective study of patients with bAVMs who underwent CKSRS from 2009 to 2014 was undertaken. Cases with incomplete data were excluded. Pre-SRS demographics of each patient such as age, sex, history of hemorrhage, previous AVM treatment (surgery, embolization, or radiotherapy), neurological status and clinical presentation were recorded. In those with unruptured AVMs, their initial symptom(s), such as seizure or incidental finding, were documented. Patients' age, AVM volume, by milliliter, and location were taken into account for the calculation of modified radiosurgery-based score (mRBAS) as per the following equation: AVM score = (0.1 Â volume [mL]) + (0.02 Â age [year]) + (0.5 Â location [0 or 1]). One point was assigned for deep locations (brainstem, basal ganglion, and thalamus), whereas 0 points were given for the non-deep areas. [20] The mRBAS were arranged into ranges of scores, 1, 1.01 to 1.5, 1.51 to 2 and >2, for further examination. The CKSRS treatment parameters from the included AVM were retrieved.

Radiosurgery planning and treatment technique
A moldable plastic, custom-made, mask was individually fitted for each patient before obtaining the CKSRS protocol, with 1.2 mm cuts, and a contrast-enhanced computerized tomography (CT) scan. In addition, selected series of magnetic resonance imaging (MRI) consisting of thin-sliced (1-3 mm) gadoliniumenhanced T1, proton-density, and contrast-enhanced magnetic resonance angiography (MRA) were uploaded to the CK planning station. Subsequent integration of the MRI/MRA with the CT was done. The target (AVM nidus) as well as critical structures were delineated before treatment planning. For bAVMs with a diameter <30 mm, or volume <15 mL, a single-fraction SRS was employed. For larger targets, a multisession, in 5 daily deliveries, SRS regimen was utilized. An appropriate prescribed radiation dose, 15 to 20 Gray (Gy), was selected for the single-fraction SRS plan whereas 4 to 6 Gy/ fraction for 5 consecutive fractions was applied for the multisession regimen. The prescribed dose was typically assigned to the 50% to 75% isodose line. A ray-tracing algorithm was exercised for dose calculation.

Outcome assessments
After the CKSRS treatment, appointments for clinical evaluation were scheduled at a 6-month interval. MRI and MRA scans at 24 or 36 months were typically obtained to determine complete obliteration of the AVM. At that time, those with apparent nidus or remaining flow, evident by early draining vein(s), were examined by annual neurologic tests, MRI, and MRA. In patients with nondetectable AVM nidus and no venous outflow by MRI/ MRA, cerebral angiography would have been performed except in the event of patient refusal. For post-CKSRS assessment of outcomes, the included cases must have had at least 36 months of follow-up duration. The outcome of each patient, at his/her latest follow-up, was determined by the classification described by Pollock and Flickinger as follows: excellent, good, fair, unchanged and poor outcome, and death. Complete obliteration of the AVM without a new neurological deficit was defined as excellent outcome. Patients with AVM obliteration were classified into good outcome if they had minor deficit, and into fair outcome if they suffered major deficit that resulted in a decline of their functional status. If the AVM was not obliterated, the unchanged outcome was given to those without a new deficit while poor outcomes were patients who sustained a new deficit. Death was the last category of outcomes if it was believed to be directly related to the AVM or the SRS treatment. [21] The percentage of the aforementioned outcomes were stratified into groups based on the ranges of mRBAS, [20] as described earlier.
Owing to the nature of this retrospective chart review, the clinical and radiographic outcomes, documented by multiple examiners and radiologists, were not blinded.

Data analysis
Patient, treatment, and outcome data would be presented as mean ± standard deviation (SD) or median (interquartile range [IQR]), where appropriate, for continuous variables and as percentage for categorical variables. To investigate patient and treatment factors influencing an excellent outcome, the data were compared using the Student t test or the Mann-Whitney U test for continuous variables, and the x 2 or Fisher exact test for categorical variables. Univariate and multivariate analyses were utilized to identify predictors for excellent outcomes by the logistic regression model with odds ratios and 95% confidence intervals (CI) computation. Kaplan-Meier survival graph of the proportion of AVM obliteration over time would be generated by linear regression analysis. All statistical tests were performed with Stata version 14 software (StataCorp, College Station, TX). Statistical significance was considered with a P value <.05.

Results
From 2009 to 2014, there were 277 bAVM patients who underwent CKSRS at our institute. The excluded subjects were 22 patients whose data could not be retrieved and another 89 cases with insufficient follow-up (<36 months). This resulted in a study cohort of 166 patients, harboring 170 AVMs. Two cases Punyawai et al. Medicine (2021) Table 3). The actuarial AVM obliteration rates from the single-fraction cohort, at 3, 5, 8, and 10 years, were 32.28%, 50.04%, 75.99%, and 75.99%, respectively. Patients with multisession CKSRS did not achieve the same level of success, evident by Log-rank test ( Fig. 1), as the single-fraction treatment group (P < .001). Its actuarial post-SRS AVM obliteration rates were 9.3%, 11.75%, 31.6%, and 43%, at 3, 5, 8, and 10 years, respectively. No patient with AVM eradication suffered a new neurological deficit; hence, the rate of excellent outcome was maintained at 58.24% and none was classified into good or fair categories. Among the nonobliterated AVMs, 69 patients had no new deficit, resulting in the proportion of 40.59% for the unchanged outcome group. After CKSRS, 7  The parameters are presented in median (interquartile range) or mean ± standard deviation unless specified otherwise. † Based on modified radiosurgery-based scoring system (deep AVM locations = brainstem, basal ganglion, and thalamus). [20]   overall occurrences of AVM hemorrhage (4.12%), in 6 patients, were recorded. The events arose at as early as 6, but no later than 67 months after treatment (Fig. 2). Of the 6 patients, only 1 suffered a new neurological deficit with declined functional status. Apart from the hemorrhagic events, a 14-year-old girl developed glioblastoma multiforme (GBM), at the irradiated AVM region, 4 years after embolization and multisession CKSRS of her ruptured AVM. Surgical excision of the tumor was performed with subsequent chemotherapy and radiation. Although she did not have post-SRS hemorrhage or sustained a new neurological deficit, this patient and the aforementioned post-SRS hemorrhage victim were sorted into the same category, making the total number of 2 cases (1.18%) in the poor outcome group. Apart from the mentioned patients, the rest of the study cohort had no documented adverse radiation effect or death. Table 4 details the patients' and treatment's variables for excellent outcome. The significant factors were the male sex (P = .018), AVM volume (P = .021), multisession SRS (P < .001), isodose line (P = .018), and time to AVM obliteration (P < .001). In contrast to many publications, we did not find a history of AVM rupture (P = .672) or previous embolization (P = .632) to correlate with untoward results. Moreover, the age (P = .908) and deep location (P = .338) were not associated with excellent outcome, although the mRBAS appeared to be related (P = .013). Further evaluation by uni-and multivariate analyses for independent predictors of excellent outcome was performed. The AVM volume (95% confidence interval [CI] = 0.98-1.04, P = .488), mRBAS (95% CI = 0.27-2.31, P = .675), and isodose line (95% CI = 0.91-1.02, P = .272) were insignificant, by multivariate examination, whereas the male sex (95% CI = 0.25-0.99, P = .048) and multisession SRS (95% CI = 0.06-0.57, P = .003) were confirmed to negatively affect the outcome (Table 5).

Discussion
Stereotactic radiosurgery is an established treatment modality for bAVM. [1,2] Ideal SRS ought to yield high rates of AVM obliteration with a trivial proportion of complications. Despite advances in imaging studies and radiation delivery techniques, AVM obliteration rates remained relatively unchanged. However, newer technologies appeared to have lowered the overall sequelae of radiotherapy. [22] Thanks to the frameless immobilization, CyberKnife SRS permits the option of administering treatment, by either single-or multisession, for varying sizes of AVMs. However, unlike ample data by GK and LINAC series, the literature search, for full-text documents published in the English language, produced just <10 publications from CK series. [14][15][16][17][18][19] Moreover, there was only 1 publication that had a median follow-up duration >60 months. Unfortunately, only 9 patients comprised this published data by Gupta et al. [15] Our study, therefore, included both the large number of AVM subjects and the long follow-up time. The overall AVM obliteration rate in this study was in line with the previously reported range of 50% to 90%, [3,4,8,11,12,23,24] albeit on the lower end of the spectrum, possibly due to the lesser-than-average prescribed dose (15 Gy) along with the larger target volume. The overall rate of post-SRS hemorrhage was also within the reported 1% to 5% range. [2,7,14,[17][18][19] Recognized negative factors, for AVM eradication by SRS, include patient factors, such as age, deep location, history of hemorrhage or large AVM volume, and treatment factors, that is, prescribed dose. [7,8,25] This study concurred with previous publications with regard to the AVM volume and multisession SRS. It should reflect the substantial magnitude of targets that automatically mandated fractionation of radiotherapy rather than the poor selection of treatment options. On the other hand, the excellent outcome was not inversely affected by age, history of AVM rupture and deep location, as previous studies validating radiosurgery-based AVM scoring systems might suggest. [26][27][28] Although most of the SRS series did not find sex to be associated with outcomes, this study, in a multivariate analysis, identified the male sex as an independent negative predictor. Frager et al published a similar observation [29] and Bir et al found that female patients had a higher proportion of AVM obliteration. [30] On the contrary, Yang et al found the male sex to be one of the protective factors against the post-SRS rebleeding [31] and Liscak et al [3] demonstrated that male patients achieved a higher percentage of AVM obliteration.
In addition to the above-mentioned negative factors, it is largely well-known that pre-SRS embolization hinders the probability of AVM obliteration. [32,33] At Ramathibodi Comprehensive Neurovascular Center, there has always been a significant proportion of AVM patients whose pre-SRS embolization were necessary. In contrast, our statistical analysis contradicted those facts. Similar findings, of no untoward effect from prior embolization, were previously published by few centers. [34,35] Oermann et al noted that the previously embolized AVMs had a significantly worse rate of obliteration after SRS. However, upon multivariate analysis, it failed to prove the case but, instead, the AVM architectural complexity was the actual negatively-affecting variable. [36] We have not explored this particular matter in the present study.
By not including abstract-only information, an English language literature search for full-text, from the PubMed, Scopus, and Google Scholar databases, returned 4 CKSRS for bAVM series with at least 20 cases. [14,[17][18][19] Despite the largest number of subjects in their study, Colombo et al performed the assessments of outcomes from 102 patients who had at least 36 months of follow-up. [14] Therefore, with 170 AVMs and the  Table 5 Regression analyses of the predictors for excellent outcome by SRS treatment of brain AVM. .272 CI = confidence interval, mRBAS = modified radiosurgery-based AVM score, [20] deep AVM locations = brainstem, basal ganglion, and thalamus, SRS = stereotactic radiosurgery. Table 4 Analyses of variables for excellent outcome by SRS treatment of brain AVM. AVM = arteriovenous malformation, Gy = Gray, IQR = interquartile range, m = month, mRBAS = modified radiosurgery-based AVM score [20] , N/A= not applicable, SD = standard deviation, SRS = stereotactic radiosurgery, y = year. * Number with percentage in brackets unless specified otherwise.  (Table 6). Although the overall AVM obliteration from our study, compared with others, appeared to be relatively low, there are several conceivable explanations for it. First, the median AVM volume of 4.17 mL in the single-fraction group was the largest among the CKSRS series. In addition, this study comprised a higher proportion of larger AVMs than other CK series, with the median volume of 22.3 ml. Due to sizeable AVMs, the prescribed doses for single-and multisession CKSRS were relatively lower than in other studies. Considering this, lower than average obliteration rates were rather predictable. Nevertheless, the incidence of post-SRS AVM hemorrhage appeared to be within the reported range. [14,[17][18][19] Unfortunately, because different bodies of literature described various but not standardized outcomes, it was rather difficult to directly compare complication rates among the CKSRS series, other than the post-SRS AVM rebleeding. None detailed their results based on RBAS or mRBAS systems; hence, this study was the first, among CKSRS cohorts, to stratify results by standardized method. One patient with GBM was observed in the study. She was the second case who developed this malignancy after CKSRS for bAVM, after the first patient report from Xhumari et al. [37]

Study limitations
The presented study has some limitations. First, the "criterion standard" cerebral angiography to determine complete obliteration was not used in all cases. Due to the fact that some patients refused to take part in the post-SRS cerebral angiographic study, the outcome assessment is less than ideal because of the nonuniform post-treatment radiographic evaluation. Another constraint was the exclusion of 111 cases (39%) for lack of data or insufficient follow-up duration. It could have affected the overall obliteration rate or the incidence of complications as Table 6 CKSRS for brain AVM series with at least 20 cases including the present study.

Conclusion
These results, with a considerable number of patients and extensive follow-up duration, confirmed the efficacy and safety of the frameless image-guided robotic stereotactic radiosurgery for brain AVM. Identified risk factors hindering achievement of excellent outcome were the male sex and multisession treatment.