A Practical Grading Scale for Predicting Outcomes of Radiosurgery for Dural Arteriovenous Fistulas: JLGK 1802 Study

Data collection

The data of 253 patients with DAVFs who were treated between 1997 and 2016 at 13 Japanese gamma knife centers affiliated with the Japanese Leksell Gamma Knife Society were retrospectively collected. Complete enumeration was mandated to minimize reporting bias aside from the following exclusion criteria: (1) history of extensive intracranial radiotherapy (except for stereotactic radiotherapy for lesions far from the DAVF); (2) no radiographic follow-up; and (3) age below 20 years at SRS. After data merging, 40 patients with multiple DAVFs, 12 who had undergone direct surgery, and one for whom sufficient data were not obtained were further excluded. Thus, a total of 200 patients were included in the study. An itemized list of the extracted information is shown in Supplementary Table 1. NHND was defined as any progressive focal neurological deficit with features other than headache, bruit, orbital phenomena, and cranial nerve deficits due to cavernous sinus (CS) lesions (ophthalmoplegia), which are usually regarded as benign [12,13].

The therapeutic strategy for DAVF was determined by each gamma knife center, referring hospitals, or a combination of these. This determination was made after due consideration of the location, symptoms, neurological deficits, presence/absence of hemorrhage, and CVR, as well as the patient’s preference. In general, endovascular treatment was primarily selected for those who had hemorrhagic strokes and/or impending neurological symptoms. SRS tended to be selected for those who did not have impending neurological symptoms. Surgical treatment was generally reserved for specific situations, such as anterior skull base (ASB) DAVFs.

All aspects of this study were approved by the Clinical Research Review Board of our institution (2018187NI). The study was performed in compliance with the institutional and Strengthening The Reporting of OBservational Studies in Epidemiology (STROBE) guidelines (Appendix 1). This study was registered in a clinical trial registry (University Hospital Medical Information Network Clinical Trials Registry, UMIN000037211) and certified by the Academic Committee of the Japanese Leksell Gamma Knife Society (JLGK1802). The requirement for patient consent was waived owing to the retrospective nature of the study.

Radiosurgical techniques and post-radiosurgical follow-up

A Leksell Gamma Knife (Elekta AB, Stockholm, Sweden) was used for SRS at all institutions. The Gamma Knife model (B, C, 4C, and Perfexion) and treatment software (KULA and GammaPlan, Elekta AB) were selected based on their availability at each institution at the time of SRS. Each patient underwent placement of a stereotactic Leksell frame (Elekta AB) under local anesthesia with or without mild sedation. For stereotactic imaging, a combination of magnetic resonance imaging and digital subtraction angiography (DSA) was generally preferred to computed tomography angiography. However, the technique to be used was determined at the discretion of the attending physicians, and the choice was dependent on the availability of equipment at each center. During radiosurgical planning, arteriovenous fistulas in the dural tissues, appearing as a conglomeration of abnormal small vessels, were circumscribed. Dose selection was performed at the discretion of each center. A prescription dose of 18 to 20 Gy was usually preferred. Because the target location was outside the brain tissues, dose reduction was rarely required. The radiosurgical target tended to be slightly more generous than that for intra-axial lesions.

Patients were followed up regularly at each gamma knife center or referring hospital. In general, magnetic resonance imagings were performed every 6 months until fistula obliteration was confirmed and annually thereafter. DSA was mainly used to confirm fistula obliteration. However, magnetic resonance imaging and/or computed tomography angiography were used alternatively if the patient declined DSA.

Statistical analysis

First, the rate of fistula obliteration was analyzed using the Kaplan-Meier method. The curves were stratified based on the baseline characteristics. Each pair of curves was compared using the log-rank test. Continuous variables were dichotomized using median values and were used for stratification. Given the multiplicity, the locations of the fistulas were grouped based on the odds ratios. Patients who underwent any secondary intervention for DAVF were censored at that point (censoring without obliteration). Factors associated with obliteration were examined using bivariate and multivariate Cox proportional hazards analyses. Based on bivariate analysis, factors with a P<0.10 were entered into the multivariate model. Post-SRS hemorrhage was examined using the person-year method. Adverse radiation events (AREs) were summarized and graded using the Common Terminology Criteria for Adverse Events (CTCAE v5.0). Disease-specific mortality was determined using the Kaplan-Meier method. Neurological status prior to SRS and at the last clinical visit was retrieved from patient medical records. Subsequently, the modified Rankin Scale (mRS) [14] score was retrospectively assigned.

Next, to examine the effect of embolization, radiosurgical outcomes were compared between patients treated with SRS alone and those treated with Emb-SRS. The 1:1 nearest available matching method was used. Matching variables were determined based on differences in baseline characteristics between the two cohorts. The rates of obliteration and post-SRS hemorrhage were compared between the matched cohorts.

Finally, to develop a practical grading scale for the prediction of fistula obliteration, factors predictive of fistula obliteration in multivariate analysis were weighted according to their odds ratios. The area under the receiver operating characteristic curve was generated to evaluate the predictive accuracy of the model. Although there is no radiosurgical grading scale for DAVFs, the Virginia Radiosurgery AVM Scale is widely used [15]. The goodness-of-fit of this scale was assessed using the R² value (0.090). Similarly, we also calculated the R² value for the supplemental evaluation of the goodness-of-fit of our scale. All statistical analyses were performed using JMP Pro version 15.0 (SAS Institute Inc., Cary, NC, USA). Statistical significance was set at P<0.05.

Baseline characteristics and overall radiosurgical outcomes

The baseline characteristics are summarized in Table 1. The median radiological follow-up period was 44.6 months (interquartile range [IQR], 24.2 to 73.0). A total of 156, 123, and 92 patients completed the imaging evaluation at 2-, 3-, and 4-year follow-up, respectively. Fistula obliteration was confirmed in 142 (71.0%) patients at a median of 19.9 months (IQR, 12.0 to 28.0); DSA was used in 116 (81.7%) of these patients. The 2-, 3-, 4-, and 5-year cumulative obliteration rates were 49.6%, 66.3%, 78.8%, and 82.7%, respectively (Figure 1A). For residual DAVFs, additional interventions were performed in 15 patients at a median of 38 months (IQR, 12 to 50). Additional interventions included embolization, SRS, direct surgery, and a combination of SRS and embolization in six, five, three, and one patients, respectively. Further data on fistula obliteration following additional interventions were not consistently available.

Table 1.

Patients’ baseline characteristics: summary and comparisons between patients treated with SRS alone and Emb-SRS

Figure 1.

Kaplan-Meier curves of fistula obliteration rates. (A) Entire cohort, (B) female vs. male, (C) presence vs. absence of cortical venous reflux (CVR), (D) sinus type vs. non-sinus type, (E) per fistula location, (F) per location category. SRS, stereotactic radiosurgery; CS, cavernous sinus; TS, transverse-sigmoid; SSS, superior sagittal sinus; ASB, anterior skull base; MF, middle fossa; Tent, tentorial.

The obliteration rate significantly differed according to sex, drainage pattern, sinus versus non-sinus DAVF type, and fistula location. The obliteration rate was higher in women versus men (P<0.01), DAVFs with versus without CVR (P<0.01), and sinus type versus non-sinus type (P<0.01). DAVFs in the CS, transverse-sigmoid (TS) sinus, or other locations had the highest obliteration rates, followed by those in the superior sagittal sinus (SSS), tentorium (Tent), ASB, and middle fossa (MF) (P<0.01) (Figure 1B-E). Other factors such as age (<65 years vs. ≥65 years), prior embolization, hemorrhagic onset, presence of NHND, prescription dose (<20 Gy vs. ≥20 Gy), central dose (<36 Gy vs. ≥36 Gy), and target volume (≤1.5 mL vs. >1.5 mL) failed to show significant differences. Based on the odds ratios for obliteration, fistula locations were categorized into three groups: (1) ASB/MF (lowest expectation of obliteration); (2) SSS/Tent (medium); and (3) CS/TS/others (highest). Significant differences in the obliteration outcomes were observed across the three location categories (P<0.01) (Figure 1F).

Based on the univariate Cox proportional hazard analysis, sex, location category, CVR, and sinus/non-sinus type were entered into the multivariate model, which demonstrated that location category (adjusted hazard ratio and 95% confidence interval [CI] for CS/TS/others [vs. ASB/MF], 4.0 [95% CI, 1.5 to 10.4]; CS/TS/others [vs. SSS/Tent], 1.8 [95% CI, 1.1 to 3.0]; and SSS/Tent [vs. ASB/MF], 2.2 [95% CI, 0.8 to 5.8]; P-value for overall difference <0.01) and absence of CVR (adjusted hazard ratio, 1.7; 95% CI, 1.2 to 2.5; P<0.01) were significantly associated with obliteration (Table 2).

Table 2.

The results of bi- and multi-variate Cox proportional hazard analyses for factors potentially associated with obliteration

Post-SRS hemorrhage occurred in four patients (2.0%) at a median of 9.3 months (IQR, 5.0 to 61.6). The details are summarized in Supplementary Table 2. All hemorrhages were caused by unobliterated niduses. Three patients (75.0%) presented with hemorrhage or NHND and two (50%) had CVR. Mild/temporary deficits were observed in two patients, whereas no deficits occurred in the remaining patients. Three hemorrhages occurred over 369.3 person-years within 2 years following SRS, and one hemorrhage occurred over 617.5 person-years thereafter. Thus, the annual hemorrhage rates were 0.8% for the first 2 years and 0.5% thereafter.

Regarding symptom improvement, bruit (n=31) improved in 27 patients (87.1%) at a median of 6.0 months (IQR, 3.0 to 9.0), ophthalmic symptoms (n=47) improved in 35 patients (74.4%) at a median of 5.0 months (IQR, 3.0 to 10.5), NHND (n=24) improved in eight patients (33.3%) at a median of 5.5 months (IQR, 2.3 to 11.1), and AREs were observed in six patients. The AREs included right upper extremity weakness due to a DSA complication in one patient (grade 4), temporary abducens and trigeminal nerve deficits in one patient (grade 2; improved with oral corticosteroid), and asymptomatic edema adjacent to the irradiated field in four patients (grade 1). Thus, the incidence of symptomatic AREs was 1.0%. During follow-up, there were four deaths (three due to extracranial malignancy and one due to a dissecting aortic aneurysm). Therefore, the disease-specific mortality rate was 0%. The preoperative mRS score was 0 in 82 patients (41.0%), 1 in 79 (39.5%), 2 in 23 (11.5%), 3 in 10 (5.0%), and 4–5 in six (3.0%). At the last clinical visit, the mRS score was 0 in 146 patients (73.0%), 1 in 20 (10.0%), 2 in 18 (9.0%), 3 in seven (3.5%), 4–5 in five (2.5%), and 6 in four (2.0%). Thus, the mRS scores improved in 75 patients (37.5%), remained unchanged in 117 (58.5%), and worsened in eight (4.0%). Deterioration in mRS scores was largely unrelated to SRS or DAVF (remote malignancy in four patients, aortic dissection in one, Parkinson’s disease in one, and senility in one), except for one patient who developed ataxia from post-SRS hemorrhage.

Effect of embolization on obliteration and hemorrhage

In total, 83 patients were treated with SRS alone, and 117 underwent Emb-SRS. Regarding Emb-SRS, embolization preceded SRS in all but three patients who received SRS and planned adjuvant embolization within 2 months following SRS. The materials used for embolization included the following: n-butyl-2-cyanoacrylate (n=37), coil (n=26), n-butyl-2-cyanoacrylate with coil (n=18), polyvinyl alcohol/gelatin sponge (n=4), coil with polyvinyl alcohol particles (n=3), n-butyl-2-cyanoacrylate with polyvinyl alcohol particles (n=2), Onyx (Medtronic, Dublin, Ireland; n=1), triacyl gelatin microspheres (Embosphere, Nippon Kayaku, Tokyo, Japan; n=1), absorbable sutures (n=1), and unknown (n=24). The baseline characteristics of the patients are summarized in Table 1. There was no significant difference in the baseline characteristics, except for less frequent NHNDs (P<0.01), additional incidental DAVF detection (P=0.03), and a decreased radiation target volume (P<0.01) in the SRS alone group compared to the Emb-SRS group.

Matching was performed using a 1:1 protocol without replacement, with a caliper width equal to 0.2 of the standard deviation of the logit of propensity. Factors that were significantly different between the cohorts (target volume, onset of NHND, and incidental diagnosis) and significantly associated with obliteration (location category and CVR) were used as matching variables. Thus, two matched cohorts comprising 146 patients were included. There were no significant differences in the baseline characteristics between the cohorts (Supplementary Table 3). Fistula obliteration following initial SRS was confirmed in 51 (69.9%) and 53 (72.6%) patients in the SRS alone and Emb-SRS groups, respectively. Emb-SRS was not associated with fistula obliteration (4-year cumulative obliteration rates [Emb-SRS vs. SRS alone], 80% vs. 76%; P=0.54; hazard ratio, 1.13; 95% CI, 0.78 to 1.66) (Figure 2). Post-SRS hemorrhage occurred in two (2.7%) and no (0%) patients in the SRS alone and Emb-SRS groups, respectively. No significant difference in the incidence of post-SRS hemorrhage was observed between the cohorts (Fisher exact test, P=0.50).

Figure 2.

Kaplan–Meier curves of fistula obliteration rates in the matched cohort. Emb-SRS, embolization with SRS; SRS, stereotactic radiosurgery

Development of a practical radiosurgical grading system for DAVFs

The results of the analyses described above showed that the location category and CVR were significantly associated with fistula obliteration. To develop a radiosurgical DAVF grading system, these factors were weighed according to the odds ratios for failed obliteration (presence of CVR [vs. absence], 3.0 [95% CI, 1.5 to 5.9; P<0.01]; SSS/Tent [vs. CS/TS/others], 3.3 [95% CI, 1.6 to 6.9; P<0.01]; ASB/MF [vs. CS/TS/others], 11.8 [95% CI, 3.7 to 38.3; P<0.01]). As such, in the grading system, no points were assigned if the patient had a fistula without CVR and that was not located in the ASB, MF, SSS, or Tent; 1 point was assigned if the fistula had CVR or was located in the SSS or Tent; and 2 points were assigned if the fistula was located in the ASB or MF. Using the total points assigned, patients were divided into three groups based on the expectation of fistula obliteration: highest (0 points), intermediate (1 point), and lowest (≥2 points) (Figure 3). According to this grading system, the 2-, 3-, and 4-year cumulative obliteration rates were 74.2%, 85.1%, and 94.4% in the highest expectation group; 42.3%, 62.7%, and 71.3% in the intermediate expectation group; and 14.6%, 34.7%, and 60.4% in the lowest expectation group (P<0.01), respectively. Logistic regression analysis demonstrated that the area under the curve of the model was 0.75, and the R² value was 0.24, which was comparable to the pseudo R² value (0.090) achieved by the Virginia Radiosurgery AVM Scale [15].

Figure 3.

Radiosurgical grading system for dural arteriovenous fistulas (DAVFs). (A) Schematic drawing of the radiosurgical DAVF grading system. (B) Kaplan-Meier curves showing rates of fistula obliteration stratified by the radiosurgical DAVF grading system. CVR, cortical venous reflux; SSS, superior sagittal sinus; ASB, anterior skull base; MF, middle fossa; SRS, stereotactic radiosurgery.