Spinal DAVFs, perimedullary AVFs, and spinal. Clinical features and treatment outcomes of the spinal arteriovenous fistulas and malformations

J Neurosurg Spine 19:207 216, 2013 AANS, 2013 Clinical features and treatment outcomes of the spinal arteriovenous fistulas and malformations Clinical article Won-Sang Cho, M.D., 1 Ki-Jeong Kim, M.D., Ph.D., 4 O-Ki Kwon, M.D., Ph.D., 4 Chi Heon Kim, M.D., Ph.D., 2 Jiha Kim, M.D., 1 Moon Hee Han, M.D., Ph.D., 2,3 and Chun Kee Chung, M.D., Ph.D. 2 1 Department of Neurosurgery, Kangwon National University Hospital, Kangwon National University School of Medicine, Chuncheon, Gangwon-do; Departments of 2 Neurosurgery and 3 Radiology, Seoul National University Hospital, Seoul; and 4 Department of Neurosurgery, Seoul National University Bundang Hospital, Seongnam, Korea Object. Spinal vascular diseases, such as spinal dural arteriovenous fistulas (DAVFs), perimedullary arteriovenous fistulas (AVFs), and spinal arteriovenous malformations (AVMs), are very rare. The authors analyzed the features and treatment outcomes of these conditions. Methods. Data from 64 patients were retrospectively reviewed. There were 33 spinal DAVFs (1 patient had 2 lesions), 20 perimedullary AVFs, and 12 spinal AVMs. Clinical features, radiological findings, treatment results, and clinical outcomes were evaluated according to the diseases, subtypes, and treatment modalities. The median duration of follow-up was 20, 42, and 56 months for spinal DAVFs, perimedullary AVFs, and spinal AVMs, respectively. Results. Spinal DAVFs showed faster progression of symptoms (median 5, 12, and 36 months for spinal DAVFs, perimedullary AVFs, and spinal AVMs, respectively) and worse neurological status at diagnosis (poor neurological status in 56%, 65%, and 33%, respectively). On MRI, signal voids were demonstrated in all except 1 spinal DAVF. At the last follow-up, 94% of spinal DAVFs, 68% of perimedullary AVFs, and 50% of spinal AVMs were completely obliterated. Favorable clinical outcomes were achieved in 91%, 95%, and 58%, respectively. In detail, the majority (78%) of spinal DAVFs were embolized, resulting in complete obliteration in 92% and favorable clinical outcomes in 92%. Most Type IVa and IVb perimedullary AVFs were surgically treated (71% and 88%), with complete obliterations of 86% and 71%, and favorable clinical outcomes in 100% and 86%, respectively. All Type IVc lesions were embolized with a low cure rate of 40%; however, clinical outcomes were satisfactory. Spinal AVMs were generally embolized (67%), and only glomus-type lesions attained a satisfactory cure rate (80%) and clinical outcome (100%). Conclusions. Embolization produced satisfactory outcomes in spinal DAVFs and glomus-type spinal AVMs. Surgery is advantageous in Type IVa and IVb perimedullary AVFs. Palliative embolization can be effective in Type IVc perimedullary AVFs and juvenile spinal AVMs. (http://thejns.org/doi/abs/10.3171/2013.4.spine12732) Key Words spinal dural arteriovenous fistula clinical features perimedullary arteriovenous fistula spinal arteriovenous malformation radiological findings treatment outcome Spinal DAVFs, perimedullary AVFs, and spinal AVMs are vascular diseases that affect the spinal cord. 23,26,31 Although they are very rare, their clinical manifestations of myeloradiculopathy are usually progressive and result in devastating outcomes when untreated. In reports of spinal DAVFs by Aminoff and Logue, 2,3 19% of the patients were disabled at 6 months Abbreviations used in this paper: ASA = anterior spinal artery; AVF = arteriovenous fistula; AVM = arteriovenous malformation; DAVF = dural AVF; PSA = posterior spinal artery. J Neurosurg: Spine / Volume 19 / August 2013 and 50% were disabled at 3 years. The introduction of MRI and spinal angiography has made it possible to understand, diagnose, and treat spinal vascular diseases. Spinal DAVFs form at the dural root sleeve between radicular feeding arteries and draining veins. They were first reported by Kendall and Logue 20 and first denoted as a Type I spinal vascular disease by Di Chiro et al. 11,27 In the early era, stripping and resection of the entire abnormal vein was performed. With the advent of selective spinal angiography by Di Chiro et al. in 1967, 10 an improved understanding of the pathophysiology has made interrup- 207

W. S. Cho et al. tion of fistulas without stripping the entire vein a standard surgical treatment. 20 In addition, since Doppman et al. performed the first embolization of a spinal DAVF in 1968, 13 endovascular therapy has developed markedly and has become one of the major treatment modalities. Perimedullary AVFs are located on the spinal cord, mainly around the conus medullaris, and are fed by intradural axial feeding arteries. They were first described by Djindjian et al. 12 in 1977 and were subclassified into Types IVa, IVb, and IVc based on fistula size and flow. 17,22,23 Because the feeding arteries also supply the spinal cord, prudent treatment planning is needed. Spinal AVMs consist of abnormal arteries and veins without connecting capillaries. According to the contour and extent of the nidus, spinal AVMs have been divided into the glomus type (Type II spinal vascular disease) and juvenile type (Type III spinal vascular disease). 5,10,29 Symptoms tend to occur acutely from the rupture of a nidus, whereas gradual deterioration is less common. 7,9,29 Because the angioarchitecture is complex, partial obliteration to reduce shunt flow and removal of hazardous components, such as associated aneurysms and varices, are usually the goals of treatment and not complete exclusion. Despite the evolution in the understanding of the pathophysiology, diagnostic tools, and treatment modalities, the rarity and gradual progression of spinal vascular diseases still makes them difficult to diagnose and treat properly. We reviewed our 20-year treatment records to evaluate their clinical characteristics and analyze the treatment outcomes. Methods Patient Selection Under the approval of our institutional review board (Seoul National University Hospital and Seoul National University Bundang Hospital), we retrospectively reviewed 64 patients with available medical records who had been treated at our institutions (Seoul National University Hospital and Seoul National University Bundang Hospital) with spinal DAVFs, perimedullary AVFs, and spinal AVMs during the 20 years between 1988 and 2007. The median duration of imaging and clinical follow-up were 20 months (range 5 111 months), 42 (range 2 160 months), and 56 months (range 3 153 months), respectively. Radiological Evaluation All cases were diagnosed using spinal angiography at admission. Initial MRI was available in 31 of 32 patients with spinal DAVFs, 19 of 20 patients with perimedullary AVFs, and 11 of 12 patients with spinal AVMs. Most cases were diagnosed within 4 weeks of the initial visit. Only 10 cases took more than 4 weeks to diagnose, 9 of which were misdiagnosed at first. The initial diagnoses in these 9 cases were degenerative spondylosis, cerebral palsy, cryptogenic brain disease, transverse myelitis, idiopathic or traumatic myelopathy, varus deformity of the foot, and psychiatric disease. Follow-up MRI was performed in all patients to assess improvement or recurrence, except in 1 patient with a spinal DAVF and 1 with a perimedullary AVF. Follow-up spinal angiography was performed, except in cases that had been completely obliterated at the first treatment and had clinical improvement as well as a decrease in abnormal signals on follow-up MRI. Spinal DAVFs were diagnosed when the fistulas were formed at the dural root sleeve between radicular feeding arteries and draining veins. They were divided into 2 subtypes (Type Ia, 1 feeder; and Type Ib, 2 or more feeders from different levels). 26,31 The origins of radicular arteries in this study were vertebral/ascending cervical/deep cervical arteries in the foramen magnum and cervical spine, intercostal and lumbar arteries in the thoracic and lumbar spines, and lateral sacral arteries in the sacrum. Fistulas in the foramen magnum fed by the cervical radicular artery were also included. Perimedullary AVFs (Type IV) were diagnosed when fistulas were located intradurally and extramedullary and fed by ASAs, PSAs, or the artery of the terminal filum. The thickest ASA from the lower thoracic level was noted as an artery of Adamkiewicz. They were classified into 3 subtypes (Type IVa, a small feeder and small fistula; Type IVb, multiple dilated feeders, intermediate-sized fistula, and dilated draining veins; and Type IVc, multipediculated feeders, giant fistula, and dilated tortuous draining veins). 23,26,31 The origins of axial feeding arteries were ASAs and PSAs from vertebral, intercostal, lumbar, and filar arteries. Twelve lesions were fed by ASAs alone, 3 by PSAs alone, and 5 by both ASAs and PSAs. Six of 17 of the ASAs were the artery of Adamkiewicz. Spinal AVMs were diagnosed when they were composed of feeding arteries, a nidus, and draining veins. They consist of 2 subtypes (glomus type, glomus nidus; and juvenile type, disperse or even extraspinally extended nidus). 26,31 The origins of feeding arteries were ASAs, PSAs, lateral spinal arteries, and radicular arteries. Angiographic results were evaluated immediately after treatment and were defined as follows: complete obliteration (complete surgical interruption or successful embolization of all fistulas, or total resection of the entire nidus); incomplete obliteration (complete surgical interruption or embolization of the proximal feeding arteries but fistulas or nidus remained); and partial obliteration (partial surgical interruption or embolization of fistula, or partial resection of the entire nidus). The results at the last follow-up were divided into complete or partial with MRI and/or spinal angiography, because some of the incompletely or partially treated lesions were newly collateralized or recanalized, and some became completely obliterated during follow-up. Treatment Spinal vascular diseases were treated with surgery, endovascular therapy, or both. Complete obliteration of the fistulas or nidus without complication was a primary goal of treatment. If it was difficult to obliterate the lesions completely, halting disease progression by flow reduction and removal of rupture sources, such as varices and aneurysms, was the next goal. Endovascular therapy was preferred as the primary treatment modality at our institutions. N-butyl 2-cyano- 208 J Neurosurg: Spine / Volume 19 / August 2013

Treatment outcomes of spinal DAVFs, perimedullary AVFs, and AVMs acrylate was used as an embolic glue in most cases, and platinum coils were used in some cases. Embolization was considered successful when the liquid adhesive embolic agent reached the proximal draining veins via the fistula or nidus. Surgery was selected when the endovascular approach failed or was not feasible. Such situations included feeding arteries that were too small in caliber or too tortuous to advance the microcatheter into the fistulas and a reflux distance that was too short to prevent embolism of glue. Operative procedures consisted of laminectomy/ laminoplasty at the predetermined level and disconnection of the fistula or resection of the nidus. Some complex cases were treated using a combined approach. Clinical Evaluation Clinical status was evaluated in terms of leg motor power, leg pain, other sensory symptoms (hyperesthesia and paresthesia), and sphincter function (urination and defecation), initially and at presentation. The pattern of clinical progression was classified using 3 modes (A, discrete episodes of neurological symptoms with varying degrees of recovery; B, insidious onset and slow progression of neurological symptoms; and C, acute deterioration during chronic slow progression). For the comparison of clinical outcomes during different periods (before and immediately after treatment and at the last follow-up), a modified McCormick grading scale was used (I, neurologically intact, normal ambulation, minimal sensory deficit; II, mild motor or sensory deficit, and functionally independent; III, moderate motor or sensory deficit, functional limitation, independent with aid; IV, severe motor or sensory deficit, functionally limitation, dependent; and V, paraplegia or quadriplegia). 21 Changes in neurological status were measured immediately after treatment and at the last follow-up, and were rated as improved, unchanged, or deteriorated. Improved and unchanged statuses were considered to be favorable, while the other designation was unfavorable. All patients underwent clinical and imaging follow-up at the same time, except 1 patient with a perimedullary AVF who was lost to follow-up after treatment. Results J Neurosurg: Spine / Volume 19 / August 2013 Clinical Features The clinical data of individual patients at presentation are summarized in Table 1. Spinal vascular diseases consisted of 33 spinal DAVFs in 32 patients, 20 perimedullary AVFs in 20 patients, and 12 spinal AVMs in 12 patients. Patients were predominantly male (80%). The median age for patients with spinal DAVFs was the oldest, and symptom duration in spinal DAVFs was the shortest. Severe neurological deficits at presentation (modified McCormick Grades IV and V) were identified in 56% of patients with spinal DAVFs, 65% of those with perimedullary AVFs, and 33% of those with spinal AVMs. The incidence of signs and symptoms, such as leg motor weakness, pain, sensory disturbance, and sphincter change, increased at diagnosis compared with the initial presentation (Table 2). The mode of progression of clinical manifestations was similar in all diseases: A = 21 patients (66%), B = 7 (22%), and C = 4 (12%) in spinal DAVFs; 13 (65%), 6 (30%), and 1 (5%) in perimedullary AVFs; and 9 (75%), 1 (8%), and 2 (17%) in spinal AVMs, respectively. Acute deterioration in spinal AVMs was identified in only 2 cases. Radiological Features Most spinal DAVFs (24 [73%]) were located at the mid- to lower thoracic levels (Fig. 1A). Three fistulas were at the foramen magnum and cervical levels, 4 were at the lumbar level, and 3 were at the sacral level. There were 20 fistulas on the right side, 9 on the left side, and 4 on both sides. Twenty-five lesions were Type Ia and 8 were Type Ib. One patient had 2 simultaneous Type Ia lesions on the left at T-6 and T-7. Thirteen (65%) perimedullary AVFs were located at the conus medullaris (Fig. 1B). One lesion was on the cervical spinal cord, 2 were at the thoracic level, and 4 were at the terminal filum. Seven lesions were on the dorsal side, 8 were on the ventral side, 1 was on the lateral side, and 4 of the terminal filum perimedullary AVFs lay caudally. Seven were Type IVa, 8 were Type IVb, and 5 were Type IVc. Six (50%) of 12 spinal AVMs were located in the cervical spine (Fig. 1C). Three were in the thoracic spine, and another 3 were in the conus medullaris. Five were glomus type, and 7 were juvenile type. Two of the 7 juvenile-type lesions extended to the extraspinal area, one of which was associated with a lipomyelomeningocele. Pretreatment MRI findings are summarized in Table 3. Signal voids were demonstrated in almost 100% of the spinal vascular diseases except for 1 spinal DAVF, which showed only high T2 signal intensity on initial MRI and no lesion on spinal angiography. The lesion was identified on the 8-month follow-up spinal angiogram. Bleeding (cord hematoma and subarachnoid hemorrhage) was identified in 2 patients with spinal AVMs who presented with acute neurological deterioration. Treatment Results The overall complete obliteration rate was highest in spinal DAVFs both immediately after treatment (88% for spinal DAVFs, 70% for perimedullary AVFs, and 42% for spinal AVMs) and at the last follow-up (94%, 68%, and 50%, respectively) (Fig. 2). Imaging follow-up was available for 32 spinal DAVFs in 31 patients because 1 patient refused to undergo follow-up imaging and for 19 perimedullary AVFs because 1 patient was lost to follow-up. There was no follow-up loss in the spinal AVM cases. No case exhibited recanalization after complete obliteration. Each juvenile-type spinal AVM and Type Ib spinal DAVF that was incompletely obliterated reached complete obliteration during the follow-up. In terms of the subtypes, simpler subtypes had a higher complete obliteration rate at the last follow-up visit (Table 4). Rates of complete obliteration were 96% in Type Ia and 86% in Type Ib spinal DAVFs. Those of perimedullary AVFs were 86% for Type IVa, 71% for Type IVb, and 40% for Type IVc. Those of spinal AVMs were 80% for glomus type and 29% for juvenile type. 209

W. S. Cho et al. TABLE 1: Baseline data for patients with spinal vascular diseases* Parameter SDAVF PMAVF SAVM no. of lesions/no. of patients 33/32 20/20 12/12 male/female 26:6 12:8 9:3 median age in yrs (range) 59 (22 76) 32 (2 61) 24 (9 52) median symptom duration in mos (range) 5 (0.1 24) 12 (0.3 120) 36 (0.1 84) modified McCormick grade I 2 (6) 1 (5) 1 (8) II 6 (19) 3 (15) 4 (33) III 6 (19) 3 (15) 3 (25) IV 11 (34) 11 (55) 2 (17) V 7 (22) 2 (10) 2 (17) subtype Ia, Ib 25, 8 IVa, IVb, IVc 7, 8, 5 glomus, juvenile 5, 7 median follow-up duration in mos (range) 20 (5 111) 42 (2 160) 56 (3 153) * PMAVF = perimedullary AVF; SAVM = spinal AVM; SDAVF = spinal DAVF. One patient had 2 Type Ia lesions. Presented as the number of patients (%). Surgery was generally superior to embolization in terms of complete obliteration; however, feasibility and results were different according to the diseases and their subtypes. In spinal DAVFs, surgery achieved 100% complete obliteration; however, embolization also showed satisfactory results (Table 4). All 5 cases (100%) showed complete obliteration with surgery alone, among which 3 were referred after failure of embolization. Twenty-six (79%) of 33 cases were initially treated with embolization, and 25 (78%) underwent follow-up. One patient with a Type Ib lesion did not undergo imaging follow-up after an incomplete obliteration. Among 25 cases in the last period, 23 (92%) were completely obliterated (19 of 20 cases with a single session and 4 of 5 with 2 sessions). Two cases achieved complete obliteration with the combined approach. There were no significant differences in results between subtypes. In perimedullary AVFs, 7 cases (35%) were initially embolized in a single session, 12 (60%) underwent surgery, and 1 was treated with the combined approach. A patient with a Type IVb lesion was lost to follow-up after complete obliteration with surgery. Eventually, complete obliteration was achieved in 4 (57%) of 7 cases after embolization and 9 (82%) of 11 cases after surgery (Table 4). The majority of Type IVa and IVb lesions were satisfactorily treated with surgery; however, Type IVc lesions were generally embolized. In spinal AVMs, embolization was generally performed (Table 4). Spinal AVMs treated with embolization showed complete obliteration in 38% (3 of 8 lesions). Three of 5 cases treated partially underwent additional embolization for the purpose of flow reduction (2, 5, and 6 sessions each). Those treated with surgery achieved complete obliteration in 67% of cases (2 of 3 lesions), including 1 lesion that was incompletely obliterated initially but that became completely obliterated by the end. Clinical Outcomes A favorable clinical outcome was obtained in 88% of spinal DAVFs, 85% of perimedullary AVFs, and 58% of spinal AVMs immediately after treatment; and in 91%, 95%, and 58% at the last follow-up period, respectively (Table 5 and Fig. 3). The modified McCormick grades from I to V at the last follow-up were 8, 7, 4, 9, and 4 in spinal DAVFs; 3, 4, 6, 6, and 0 in perimedullary AVFs; and 1, 2, 1, 7, and 1 in spinal AVMs. All patients TABLE 2: Signs and symptoms at the initial period and the time of diagnosis in patients with the spinal vascular diseases Sign/Symptom No. of Patients (%) SDAVF PMAVF SAVM Initial At Diagnosis Initial At Diagnosis Initial At Diagnosis leg motor weakness 17 (53) 27 (84) 9 (45) 16 (80) 7 (58) 9 (75) leg pain 15 (47) 18 (56) 11 (55) 12 (60) 3 (25) 3 (25) leg sensory disturbance 14 (44) 28 (88) 3 (15) 16 (80) 2 (17) 6 (50) sphincter change 0 19 (59) 1 (5) 10 (50) 1 (8) 4 (33) 210 J Neurosurg: Spine / Volume 19 / August 2013

Treatment outcomes of spinal DAVFs, perimedullary AVFs, and AVMs Fig. 1. Distributions of the spinal DAVFs (A), perimedullary AVFs (B), and AVMs (C). FM = foramen magnum. with spinal DAVFs and spinal AVMs underwent clinical follow-up. One patient with a perimedullary AVF was lost to follow-up. According to the subtypes, Type Ia spinal DAVFs, Type IVa and IVc perimedullary AVFs, and glomus-type spinal AVMs had better clinical outcomes than the others at the last follow-up period (Fig. 4). Overall, embolization achieved more favorable clinical outcomes than surgery alone (Fig. 5). In spinal DAVFs, completely treated lesions showed better clinical outcomes than those that were partially treated. Favorable clinical outcomes were achieved in 93% of patients (26 of 28 patients) with completely treated lesions at each period, whereas the partially treated group had favorable outcomes in 50% (2 of 4 patients) immediately after treatment and 75% (3 patients) at the last period. In perimedullary AVFs, the group with partial obliteration showed satisfactory outcomes in 100% (6 patients) of patients at each period; however, the group with complete obliterations had worse clinical outcomes than that with partial oblierations (79% [11 of 14 patients] immediately after treatment and 92% [12 of 13 patients] at last). In spinal AVMs, the group with complete obliteration achieved better clinical outcomes than the group with partial obliteration (80% [4 of 5 patients] vs 43% [3 of 7 patients] immediately after treatment; 67% [4 of 6 patients] vs 50% [3 of 6 patients] ultimately). Complications In spinal DAVFs, embolization-related complications included 3 cases of temporary thromboembolic symptoms such as Horner syndrome, leg weakness, and flank pain, 1 case of temporary radiculomedullary artery spasm, and 1 case of asymptomatic aortic arch dissection. Surgeryrelated complications were CSF leakage in one patient and neurological deterioration without any surgical problems in another. For perimedullary AVFs, embolization-related complications were temporary flow reduction in an ASA without a distinct cause in 1 patient and ASA/PSA spasm in 2 patients. There was 1 surgery-related complication (asymptomatic rootlet sacrifice) in 2 patients. In spinal AVMs, embolization-related complications included temporary Horner syndrome in one patient and transiently symptomatic vertebral artery dissection in another. There were 2 surgery-related complications (CSF leakage in 1 patient and leg motor weakness in 2 patients). Discussion Spinal DAVFs The demographics, radiological findings, and clinical features of spinal DAVFs in this study are similar to those in previous reports. 19,24 Patients were predominantly male (81%), and lesions were most often first diagnosed in the 6th decade of life (median age 59 years). Twenty-eight (85%) of the 33 lesions were located in the midthoracic to upper lumbar spine, and about one-third of fistulas were on the left side. As symptoms gradually progressed, many TABLE 3: Magnetic resonance imaging findings in patients with spinal vascular diseases No. of Patients (%)* Finding SDAVF PMAVF SAVM T2 hyperintensity 28 (90) 11 (58) 6 (55) cord swelling 17 (55) 8 (42) 2 (18) signal void 30 (97) 19 (100) 11 (100) cord hematoma 0 2 (11) 1 (9) subarachnoid hemorrhage 2 (7) 0 1 (9) cord enhancement 8 (26) 4 (21) 3 (27) cord compression 1 (3) 2 (11) 0 syrinx formation 0 1 (5) 0 * Findings were available in 31 of 32 patients with spinal DAVFs, 19 of 20 patients with perimedullary AVFs, and 11 of 12 patients with spinal AVMs. J Neurosurg: Spine / Volume 19 / August 2013 211

W. S. Cho et al. Fig. 2. Treatment results of the spinal vascular diseases immediately after treatment (left) and at the last follow-up period (right). One patient with a spinal DAVF (SDAVF) refused to undergo imaging during follow-up. One patient with perimedullary AVF (PMAVF) was lost to follow-up. Dark bars indicate complete obliteration, and light bars indicate incomplete or partial obliteration. Numbers indicate the number of cases. SAVM = spinal AVM. types of neurological symptoms (motor weakness, pain, sensory disturbance, and sphincter dysfunction) presented at admission. Interestingly, in this study, spinal DAVFs showed more rapid symptom progression (median 5 months) and poorer neurological status than other spinal vascular diseases. Early diagnosis and prompt treatment should be considered more important in spinal DAVF cases than in the others. The important pathophysiology in these lesions is venous hypertension. A hemodynamic study reported that the venous pressure in spinal DAVFs was 60% 88% of the mean systolic arterial pressure. 16 Gravity as well as the absence of valves in the coronal venous plexus and radial veins of the spinal cord may accelerate venous congestion and cause a subsequent decrease in perfusion pressure, even resulting in infarction. 29 Symptom progression and a high percentage of specific MRI findings (T2 hyperintensity and signal voids) in this study support the concept of venous hypertension as the major pathophysiology. The interval between initial presentation and final diagnosis is still long (range 15 21 months) even after the introduction of MRI and spinal angiography because the spinal vascular diseases have no pathognomonic symptoms or signs, and they progress slowly. 14,19,24,30 In this study, the overall time interval between initial symptom onset and final diagnosis of spinal DAVFs was a median of 6 months. The median interval between initial symptom onset and first visit was 5 months, and most of the cases were diagnosed within 1 month. Signal voids and cord edema on MRI were observed in nearly all cases. Therefore, a high degree of suspicion in progressive myelopathic symptoms with prudent use of MRI would help make a timely diagnosis of other spinal vascular diseases as well as spinal DAVFs. In the early period, surgery was the main treatment modality because endovascular devices and techniques were poor. The success rate of initial surgery alone ranged from 83% to 100%. 1,24,32,33 Although only 7 cases were treated with surgery, this study showed a high success rate of 100%. On the other hand, the complete obliteration rate was lower with endovascular treatment than with surgery, especially when embolization was performed with par- TABLE 4: Treatment results of the spinal vascular diseases at the last follow-up visit according to type of lesion No. of Patients (%) SDAVF PMAVF SAVM Treatment Type Ia Ib IVa IVb IVc Juvenile Glomus embolization 20 5 2 1 4 4 4 complete obliteration 19 (95) 4 (80) 2 (100) 0 (0) 2 (50) 0 (0) 3 (75) partial obliteration 1 (5) 1 (20) 0 (0) 1 (100) 2 (50) 4 (100) 1 (25) surgery 4 1 5 6 0 2 1 complete obliteration 4 (100) 1 (100) 4 (80) 5 (83) 0 1 (50) 1 (100) partial obliteration 0 (0) 0 (0) 1 (20) 1 (17) 0 1 (50) 0 (0) combined approach 1 1 0 0 1 1 0 complete obliteration 1 (100) 1 (100) 0 (0) 1 (100) partial obliteration 0 (0) 0 (0) 1 (100) 0 (0) total 25 7 7 7 5 7 5 212 J Neurosurg: Spine / Volume 19 / August 2013

Treatment outcomes of spinal DAVFs, perimedullary AVFs, and AVMs TABLE 5: Clinical outcomes of the spinal vascular diseases* Outcome No. of Patients (%) Immediately After Treatment Last Follow-Up Visit SDAVF PMAVF SAVM SDAVF PMAVF SAVM improved 8 (25) 5 (25) 1 (8) 16 (50) 11 (58) 4 (33) unchanged 20 (63) 12 (60) 6 (50) 13 (41) 7 (37) 3 (25) deteriorated 4 (12) 3 (15) 5 (42) 3 (9) 1 (5) 5 (42) total 32 20 12 32 19 12 * Clinical outcomes were evaluated using the modified McCormick scale at each period and were compared with pretreatment grades. One patient with a perimedullary AVF was lost to follow-up after treatment. ticles such as polyvinyl alcohol. 15 However, as liquid adhesive embolic agents, such as N-butyl 2-cyanoacrylate, were introduced and the concept of adequate embolization was established by Niimi et al., 25 the success rate of embolization has markedly improved, and embolization has become comparable to surgery. 14,24,25,28,30,32 Embolization has some advantages over surgery. It is less invasive, provides the opportunity to make a diagnosis and treat the spinal DAVFs in a single stage, and helps identify the angioarchitecture. Accordingly, many institutions as well as ours select embolization as the primary treatment modality. During the last follow-up period, the complete obliteration rate was 92% for embolization and 100% for surgery; however, a favorable clinical outcome was achieved in 92% of embolizations and 80% of surgeries. Therefore, active treatment with initial embolization and surgery as an alternative is thought to be a logical treatment strategy for spinal DAVFs at our institutions. Perimedullary AVFs A report about the treatment of 19 patients with perimedullary AVFs was once published from our institutions. 8 Nine patients overlapped, and the other data of the 10 patients in the previous report were unfortunately lost. Therefore, a total of 20 patients with 11 new cases were evaluated in this study. Distinct from spinal DAVFs, perimedullary AVFs are rare and relatively unknown. There was slight predominance in male patients (60%), and the median age was younger than for spinal DAVFs (median age 32 years). In previous reports, 4,8,23 a sex predominance was not clear, and the age at diagnosis ranged from 20 to 40 years, which is younger than that for spinal DAVFs. Most lesions in this study were located at the conus medullaris; however, no relationship was found between location and subtype, as suggested by Mourier et al. 23 Similar to previous reports, in this study most cases of perimedullary AVF presented with gradual progression of symptoms. 4,8,23 The time interval varies from 17 months to 9 years according to the reports. 4,23 In this study, the duration of symptoms (median 12 months) was longer than that for spinal DAVFs and showed similar pretreatment neurological statuses. Preferable treatment modalities differ by subtype. Type IVa and IVb perimedullary AVFs are supplied by small axial feeders, and they have a simpler angioarchitecture than Type IVc lesions. Catheterization of the fistulas through the small feeders is difficult, and disastrous complications such as occlusion of the ASA or PSA may occur. Therefore, surgery is usually preferred. 4,6,8,23 Conversely, the endovascular approach is recommended in Type IVc lesions because the large and tortuous patho- Fig. 3. Clinical outcomes of the spinal vascular diseases immediately after treatment (left) and at the last follow-up period (right). One patient with perimedullary AVF was lost to follow-up. Dark bars indicate favorable outcome, and light bars indicate unfavorable outcome. J Neurosurg: Spine / Volume 19 / August 2013 213

W. S. Cho et al. Fig. 4. Clinical outcomes according to the subtypes of spinal vascular diseases at the last follow-up period. Dark bars indicate favorable outcome, and light bars indicate unfavorable outcome. logical vessels are likely to be dangerous for surgery. In this study, most of the Type IVa and IVb lesions underwent surgery, and all Type IVc lesions were treated with an endovascular approach. In previous reports, 4,23 cases with residual flow after incomplete obliteration showed neurological worsening or no improvement. The angiographic cure rate in this study was high for IVa, IVb, and IVc lesions (86%, 71%, and 40%, respectively). Nonetheless, clinical outcomes were satisfactory with little difference among subtypes (favorable outcomes: 100%, 86%, and 100%, respectively). Only 1 case (5%) of a Type IVb perimedullary AVF deteriorated. In addition, partially obliterated cases achieved better clinical outcomes than completely obliterated cases, with no instances of deterioration during follow-up. Therefore, incomplete or partial obliteration without compromise of normal vessels may be better than aggressive and complete obliteration in some cases. However, close observation is needed, because recanalization and clinical aggravation can occur in partially obliterated cases. Spinal AVMs This series showed a male preponderance (77%) and a young median age of 24 years. The duration of symptoms was the longest (median 36 months), and pretreatment neurological status was best among the types of lesions evaluated. Acute neurological deterioration due to bleeding, such as subarachnoid hemorrhage and intramedullary hemorrhage, is known as the most common clinical presentation of spinal AVMs. 7,9,29 However, only 2 cases (17%) in this study presented with hemorrhage. Slow and gradual progression may be explained by direct cord compression or venous hypertension. 33 In this study, three-quarters of the spinal AVMs were treated with embolization (80% of the glomus type and 57% of the juvenile type). The rate of complete obliteration was high (80%) in glomus spinal AVMs, whereas it was low (29%) in the juvenile type. Although surgery achieved a higher rate of angiographic cure than embolization (67% vs 38%), clinically favorable outcomes were lower in surgery alone than in embolization alone (33% vs 75%), and a minority of cases underwent surgery. Treatment results of spinal AVMs have rarely been reported, most of which were surgical cases in the early period. 18,34 The angiographic cure rate was about 70%, and clinical improvement or unchanged status ranged from 80% to 90%. The results for glomus-type spinal AVMs were better, with a long-term cure rate of 80% and favorable clinical outcomes in 93% of cases. 9 However, reports of embolization of spinal AVMs are relatively rare, and most are case reports or series. In a clinical study with embolization using particles, 7 63% of cases improved clini- Fig. 5. Clinical outcomes of the spinal vascular diseases at the last follow-up period according to treatment modality. Dark bars indicate favorable outcome, and light bars indicate unfavorable outcome. E = embolization; S = surgery. 214 J Neurosurg: Spine / Volume 19 / August 2013

Treatment outcomes of spinal DAVFs, perimedullary AVFs, and AVMs cally and 17% had an unchanged clinical status during a mean follow-up of 6 years, although no cases maintained a complete occlusion, which was comparable to the results with surgery. In addition, discrepancies between angiographic and clinical outcomes have been observed in some reports as well as in this study; 7,18 that is, longterm clinical outcomes are satisfactory even after incomplete obliteration. Therefore, although not in glomus-type spinal AVMs, a reasonable treatment strategy may be an incomplete/partial embolization or repeated embolization if necessary. Study Limitations There were some limitations in this study. First, the small number of cases made it difficult to perform a statistical analysis. Because spinal vascular diseases are very rare, it would not be easy to evaluate them statistically without a multicenter, long-term study. Second, this study used 20 years of data, during which there have been technical advances. The possibility exists that cases treated later would achieve better outcomes than those treated earlier, although the relationship between the treatment period and results was not investigated. Major diagnostic tools such as MRI and selective spinal angiography were already established, pathophysiology of the spinal vascular diseases was identified, and surgical techniques have not changed much in the past 20 years. However, considerable improvement in endovascular techniques is thought to have a great influence on treatment outcomes. Third, some cases overlapped with those in the previous reports from our institution. 8,28 However, less than 50% of our previous reports were the same, and the analysis in this study was made from a different point of view, such as comparison of clinical features, radiological findings, and treatment outcomes among each disease. Conclusions Although rapid progression of spinal DAVFs requires early treatment, a relatively simple angioarchitecture is advantageous in achieving complete obliteration and favorable clinical outcomes with initial embolization and surgery as an alternative. Surgery is suitable for Type IVa and IVb perimedullary AVFs in terms of accessibility and treatment outcomes. In cases of Type IVc perimedullary AVFs and spinal AVMs, deliberate embolization is needed for complete obliteration, flow reduction, or removal of the hazardous vascular structures. Disclosure The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. Author contributions to the study and manuscript preparation include the following. Conception and design: Chung, Cho, KJ Kim, Han. Acquisition of data: Chung, Cho, KJ Kim, Kwon, CH Kim, Han. Analysis and interpretation of data: Chung, Cho, KJ Kim, CH Kim, Han. Drafting the article: all authors. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Chung. Study supervision: Chung, KJ Kim, Kwon, CH Kim, J Kim, Han. J Neurosurg: Spine / Volume 19 / August 2013 References 1. Afshar JK, Doppman JL, Oldfield EH: Surgical interruption of intradural draining vein as curative treatment of spinal dural arteriovenous fistulas. J Neurosurg 82:196 200, 1995 2. Aminoff MJ, Logue V: Clinical features of spinal vascular malformations. Brain 97:197 210, 1974 3. Aminoff MJ, Logue V: The prognosis of patients with spinal vascular malformations. Brain 97:211 218, 1974 4. Antonietti L, Sheth SA, Halbach VV, Higashida RT, Dowd CF, Lawton MT, et al: Long-term outcome in the repair of spinal cord perimedullary arteriovenous fistulas. AJNR Am J Neuroradiol 31:1824 1830, 2010 5. Baker HL Jr, Love JG, Layton DD Jr: Angiographic and surgical aspects of spinal cord vascular anomalies. Radiology 88:1078 1085, 1967 6. Barrow DL, Colohan AR, Dawson R: Intradural perimedullary arteriovenous fistulas (type IV spinal cord arteriovenous malformations). J Neurosurg 81:221 229, 1994 7. Biondi A, Merland JJ, Reizine D, Aymard A, Hodes JE, Lecoz P, et al: Embolization with particles in thoracic intramedullary arteriovenous malformations: long-term angiographic and clinical results. Radiology 177:651 658, 1990 8. Cho KT, Lee DY, Chung CK, Han MH, Kim HJ: Treatment of spinal cord perimedullary arteriovenous fistula: embolization versus surgery. Neurosurgery 56:232 241, 2005 9. Connolly ES Jr, Zubay GP, McCormick PC, Stein BM: The posterior approach to a series of glomus (Type II) intramedullary spinal cord arteriovenous malformations. Neurosurgery 42:774 786, 1998 10. Di Chiro G, Doppman J, Ommaya AK: Selective arteriography of arteriovenous aneurysms of spinal cord. Radiology 88:1065 1077, 1967 11. Di Chiro G, Wener L: Angiography of the spinal cord. A review of contemporary techniques and applications. J Neurosurg 39:1 29, 1973 12. Djindjian M, Djindjian R, Rey A, Hurth M, Houdart R: Intradural extramedullary spinal arterio-venous malformations fed by the anterior spinal artery. Surg Neurol 8:85 93, 1977 13. Er U, Yigitkanli K, Simsek S, Adabag A, Bavbek M: Spinal intradural extramedullary cavernous angioma: case report and review of the literature. Spinal Cord 45:632 636, 2007 14. Eskandar EN, Borges LF, Budzik RF Jr, Putman CM, Ogilvy CS: Spinal dural arteriovenous fistulas: experience with endovascular and surgical therapy. J Neurosurg 96 (2 Suppl): 162 167, 2002 15. Hall WA, Oldfield EH, Doppman JL: Recanalization of spinal arteriovenous malformations following embolization. J Neurosurg 70:714 720, 1989 16. Hassler W, Thron A, Grote EH: Hemodynamics of spinal dural arteriovenous fistulas. An intraoperative study. J Neurosurg 70:360 370, 1989 17. Heros RC, Debrun GM, Ojemann RG, Lasjaunias PL, Naessens PJ: Direct spinal arteriovenous fistula: a new type of spinal AVM. Case report. J Neurosurg 64:134 139, 1986 18. Hurth M, Houdart R, Djindjian R, Rey A, Djindjian M: Arteriovenous malformations of the spinal cord: clinical, anatomical and therapeutic consideration a series of 150 cases. Prog Neurol Surg 9:238 266, 1978 19. Jellema K, Canta LR, Tijssen CC, van Rooij WJ, Koudstaal PJ, van Gijn J: Spinal dural arteriovenous fistulas: clinical features in 80 patients. J Neurol Neurosurg Psychiatry 74: 1438 1440, 2003 20. Kendall BE, Logue V: Spinal epidural angiomatous malformations draining into intrathecal veins. Neuroradiology 13: 181 189, 1977 21. McCormick PC, Torres R, Post KD, Stein BM: Intramedullary ependymoma of the spinal cord. J Neurosurg 72:523 532, 1990 215

W. S. Cho et al. 22. Merland JJ, Reizine D: Embolization techniques in the spinal cord, in Dondelinger RF, Rossi P, Kurdziel JC, et al (eds): Interventional Radiology. New York: Thieme, 1990, pp 433 442 23. Mourier KL, Gobin YP, George B, Lot G, Merland JJ: Intradural perimedullary arteriovenous fistulae: results of surgical and endovascular treatment in a series of 35 cases. Neurosurgery 32:885 891, 1993 24. Narvid J, Hetts SW, Larsen D, Neuhaus J, Singh TP, McSwain H, et al: Spinal dural arteriovenous fistulae: clinical features and long-term results. Neurosurgery 62:159 167, 2008 25. Niimi Y, Berenstein A, Setton A, Neophytides A: Embolization of spinal dural arteriovenous fistulae: results and followup. Neurosurgery 40:675 683, 1997 26. Oldfield EH, Doppman JL: Spinal arteriovenous malformations. Clin Neurosurg 34:161 183, 1988 27. Ommaya AK, Di Chiro G, Doppman J: Ligation of arterial supply in the treatment of spinal cord arteriovenous malformations. J Neurosurg 30:679 692, 1969 28. Park SB, Han MH, Jahng TA, Kwon BJ, Chung CK: Spinal dural arteriovenous fistulas: clinical experience with endovascular treatment as a primary therapeutic modality. J Korean Neurosurg Soc 44:364 369, 2008 29. Rosenblum B, Oldfield EH, Doppman JL, Di Chiro G: Spinal arteriovenous malformations: a comparison of dural arteriovenous fistulas and intradural AVM s in 81 patients. J Neurosurg 67:795 802, 1987 30. Song JK, Vinuela F, Gobin YP, Duckwiler GR, Murayama Y, Kureshi I, et al: Surgical and endovascular treatment of spinal dural arteriovenous fistulas: long-term disability assessment and prognostic factors. J Neurosurg 94 (2 Suppl):199 204, 2001 31. Spetzler RF, Detwiler PW, Riina HA, Porter RW: Modified classification of spinal cord vascular lesions. J Neurosurg 96 (2 Suppl):145 156, 2002 32. Steinmetz MP, Chow MM, Krishnaney AA, Andrews-Hinders D, Benzel EC, Masaryk TJ, et al: Outcome after the treatment of spinal dural arteriovenous fistulae: a contemporary singleinstitution series and meta-analysis. Neurosurgery 55:77 88, 2004 33. Symon L, Kuyama H, Kendall B: Dural arteriovenous malformations of the spine. Clinical features and surgical results in 55 cases. J Neurosurg 60:238 247, 1984 34. Yaşargil MG, Symon L, Teddy PJ: Arteriovenous malformations of the spinal cord, in Symon L (ed): Advances and Technical Standards in Neurosurgery. Vienna: Springer-Verlag, 1984, Vol 11, pp 61 102 Manuscript submitted July 30, 2012. Accepted April 16, 2013. Please include this information when citing this paper: published online May 24, 2013; DOI: 10.3171/2013.4.SPINE12732. Address correspondence to: Chun Kee Chung, M.D., Ph.D., Department of Neurosurgery, Seoul National University College of Medicine, 28 Yongon-dong, Chongno-gu, Seoul 110-744, Korea. email: chungc@snu.ac.kr. 216 J Neurosurg: Spine / Volume 19 / August 2013