RESEARCH HUMAN CLINICAL STUDIES

Transcription

1 TOPIC RESEARCH HUMAN CLINICAL STUDIES RESEARCH HUMAN CLINICAL STUDIES Cavernous Malformation of Brainstem, Thalamus, and Basal Ganglia: A Series of 176 Patients Paritosh Pandey, MD Erick M. Westbroek, BS Peter A. Gooderham, MD Gary K. Steinberg, MD, PhD Department of Neurosurgery, Stanford Stroke Center and Stanford Institute for Neuro-Innovation and Translational Neurosciences, Stanford University School of Medicine, Stanford, California Correspondence: Gary K. Steinberg, MD, PhD, R281, Department of Neurosurgery, Stanford University School of Medicine, 300 Pasteur Dr, Stanford, CA gsteinberg@stanford.edu Received, May 11, Accepted, December 4, Published Online, December 20, Copyright ª 2012 by the Congress of Neurological Surgeons SANS LifeLong Learning and NEUROSURGERY offer CME for subscribers that complete questions about featured articles. Questions are located on the SANS website ( Please read the featured article and then log into SANS for this educational offering. BACKGROUND: Cavernous malformations (CMs) in deep locations account for 9% to 35% of brain malformations and are surgically challenging. OBJECTIVE: To study the clinical features and outcomes following surgery for deep CMs and the complication of hypertrophic olivary degeneration (HOD). METHODS: Clinical records, radiological findings, operative details, and complications of 176 patients with deep CMs were reviewed retrospectively. RESULTS: Of 176 patients with 179 CMs, 136 CMs were in the brainstem, 27 in the basal ganglia, and 16 in the thalamus. Cranial nerve deficits (51.1%), hemiparesis (40.9%), numbness (34.7%), and cerebellar symptoms (38.6%) presented most commonly. Hemorrhage presented in 172 patients (70 single, 102 multiple). The annual retrospective hemorrhage rate was 5.1% (assuming CMs are congenital with uniform hemorrhage risk throughout life); the rebleed rate was 31.5%/patient per year. Surgical approach depended on the proximity of the CM to the pial or ependymal surface. Postoperatively, 121 patients (68.8%) had no new neurological deficits. Follow-up occurred in 170 patients. Delayed postoperative HOD developed in 9/134 (6.7%) patients with brainstem CMs. HOD occurred predominantly following surgery for pontine CMs (9/10 patients). Three patients with HOD had palatal myoclonus, nystagmus, and oscillopsia, whereas 1 patient each had limb tremor and hemiballismus. At follow-up, 105 patients (61.8%) improved, 44 (25.9%) were unchanged, and 19 (11.2%) worsened neurologically. Good preoperative modified Rankin Score (98.2% vs 54.5%, P =.001) and single hemorrhage (89% vs 77.3%, P,.05) were predictive of good longterm outcome. CONCLUSION: Symptomatic deep CMs can be resected with acceptable morbidity and outcomes. Good preoperative modified Rankin Score and single hemorrhage are predictors of good long-term outcome. KEY WORDS: Basal ganglia, Brainstem, Cavernous malformations, Microsurgery, Thalamus Neurosurgery 72: , 2013 DOI: /NEU.0b013e318283c9c2 Cavernous malformations (CMs) of the brain are common vascular malformations, accounting for approximately 5% to 10% of all cerebral vascular malformations. 1-6 The prevalence of cerebral cavernous vascular malformations is estimated to be 0.4% to 0.9%. 2,7 CMs in deep locations, including the brainstem, thalamus, and basal ganglia, account for 9% to 35% of all malformations in the brain. 3,7-10 Malformations in ABBREVIATIONS: AOVM, angiographically occult vascular malformation; CM, cavernous malformation; HOD, hypertrophic olivary degeneration; mrs, modified Rankin Score; SRS, stereotactic radiosurgery these locations are of particular interest because their critical location renders them very challenging for surgical resection. The natural history of the CM in deep locations is not completely understood, but most reports suggest thattheyhaveahigherrateofhemorrhagethan superficial CMs as well as high rebleed rates. 7-9,11 Since Dandy 12 first reported excision of a CM from the brainstem, other authors, including our group, have reported their experience with surgery for brainstem and other deep CMs. The optimal treatment of these lesions remains a matter of debate. Most surgeons have reported good long-term results with surgical excision, but there is a high rate of immediate postoperative NEUROSURGERY VOLUME 72 NUMBER 4 APRIL

2 PANDEY ET AL morbidity associated with surgery. 7-10,13-30 A few groups advocate radiosurgery, although whether radiosurgery decreases the risk of future hemorrhage is uncertain, and the risk of radiation complications appears higher than for comparably sized arteriovenous malformations Some centers advocate conservative management of the malformations in critical locations, citing high incidence of immediate postoperative deficits after surgery. 41 Other aspects of surgical decision making, including the timing of surgery, the number of hemorrhages before surgery is offered, and the surgical approach to the lesions, are also not well defined in the literature. Hypertrophic olivary degeneration (HOD) is a rare pathological and radiological phenomenon that occurs after injury to the dentate-rubro-olivary pathway, and is clinically characterized by palatal tremors and oscillopsia. 41 This phenomenon has been described secondary to trauma, hemorrhage, demyelinating disorders, tumor, and degenerative disease. 11,42-44 A few case reports and a small series of patients have been previously described, with HOD secondary to surgical management of brainstem CM. 41,45-47 Objectives We previously published our experience with 56 patients undergoing microsurgical resection of CMs in the brainstem, thalamus, and basal ganglia. 48 We now publish our much larger results for 176 patients with deep CMs, analyzing the clinical and radiological data, the surgical results, and the various factors associated with their long-term outcome. In particular, we studied various statistically significant factors associated with short-term and long-term outcome, and operating through safe corridors for CMs not presenting on the pial surface. We also describe the clinical and radiological course of HOD following surgery for brainstem CM and the management of this complication. PATIENTS AND METHODS Study Design Clinical records, radiological findings, operative details, and complications of 176 patients with deep CMs were reviewed retrospectively. TABLE 1. Anatomic Location and Demographic Details a Setting Between January 1990 and November 2010, 192 patients with 195 brainstem, thalamic, and basal ganglia CMs were operated on by the senior author (G.K.S.) at Stanford University Medical Center. Of these patients, 176 were operated from January 1990 to February 2010, and 170 had at least 6 months of follow-up; these are the patients included in the present analysis. This includes patients described in our previous report with additional clinical follow-up. Clinical records, radiological findings, operative details, postoperative course, and complications were reviewed retrospectively from hospital charts and notes. Patients with poor outcomes and those who worsened following surgery were also analyzed in detail. Follow-up was obtained from hospital charts and, when possible, through telephone interviews with the patients. The preoperative and postoperative modified Rankin Scores (mrs) 49,50 were evaluated for all patients who had a minimum follow-up of 6 months. Participants This retrospective review includes 176 patients (75 male, 101 female; M:F 0.74:1). Ages ranged from 3 months to 72 years (mean age, 39.2 years), and 20 were pediatric patients (,18 years of age). Three patients had 2 separate brainstem, thalamic, or basal ganglia CMs; thus, the number of CMs totaled 179. Of these, 136 were located in the brainstem (28 in midbrain, 94 in pons, and 14 in medulla), 27 were located in the basal ganglia, and 16 in the thalamus (Table 1). Variables Surgical Indications All patients with brainstem CMs had at least 1 clinical hemorrhage episode before they were considered for surgery. Most of the patients with thalamic and ganglionic CMs also presented with hemorrhage, and 57.9% of patients presented with more than 1 clinical hemorrhage. Only 4 patients with large insular CMs extending into the basal ganglia did not have a clinical hemorrhage; these patients presented with seizures. A hemorrhagic episode was defined as a sudden onset or worsening of a clinical neurological deficit, along with evidence of new hemorrhage on magnetic resonance imaging (MRI). Patients with asymptomatic and incidental brainstem CMs were not operated on. In the initial study period, angiography was performed to exclude arteriovenous malformation. Since 1995, however, angiography has not been used if the MR appearance is typical of a CM. In the majority of patients, CMs were located near a pial/ependymal surface. In 19 patients operated on, however, CMs did not present to the pial/ependymal surface. These CMs were located in the basal ganglia (6), insula-basal ganglia-internal capsule (3), thalamus (6), and pons (4). All Basal Ganglia Thalamus Midbrain Pons Medulla P Patients (14.8%) 16 (9.1%) 28 (15.9%) 92 (52.3%) 14 (8.0%) M:F 101 f, 75 m (57.4%f, 14 f, 12 m (53.8%f, 9 f, 7 m (56.3%f, 14 f, 14 m 56 f, 36 m (60.9%f, 8 f, 6 m (57.1%f, %m) 46.2%m) 43.7%m) (50%f, 50%m) 39.1%m) 42.9%m) Age 3 mo-72 y 8-61 y 4-68 y y 3 mo-63 y y Mean age a M, m, male; F, f, female. 574 VOLUME 72 NUMBER 4 APRIL

3 SURGERY FOR DEEP-SEATED CAVERNOUS MALFORMATIONS Surgical Adjuncts Surgery was performed under mild hypothermia (32-33 C) in all cases. Electrophysiological monitoring was used in 148 patients (84.1%) and was not used in 28 patients (early in the series). Somatosensory evoked potentials were performed in 148 patients, brainstem auditory evoked potentials in 120 patients, and motor evoked potentials in 76 patients. Cranial nerve monitoring (usually 7th nerve, but also 3rd, 4th, 5th, 6th, 9th, 10th, 11th, and 12th, depending on the location of the CM) was performed in 79 of the 134 patients with brainstem CM (79/134, 59.0%). Mapping of the cranial nerve nuclei was performed to identify the safe entry zone for intrinsic brainstem lesions. 51 Stereotactic localization was performed in 139 patients (79.0%). Intraoperative stereotactic localization of the CM was utilized, with the exception of the patients who underwent surgery in the earlier part of the study. Intraoperative ultrasonography was used in 1 patient. The OmniGuide CO 2 laser was used to resect the lesion in 13 patients. Surgical Approach Surgical approach was determined based on the proximity of the CM to a pial or ependymal surface. The approach was designed to avoid critical structures like long tracts and cranial nerve nuclei and to lead to the pial or ependymal surface closest to the CM. Critical structures, such as corticospinal tract and cranial nerve nuclei, were avoided, and approaches were modified as needed for each surgery. Unlike some previous reports, 8 we did not preferentially use an anterolateral or lateral approach for brainstem lesions; our main guideline was proximity to the pial surface. The goal of surgery was complete excision of the lesion to prevent subsequent hemorrhage. Great care was taken to preserve associated developmental venous anomalies in all cases. The hemosiderinladen brain parenchyma was also preserved, and no attempt was made to resect it. Data Sources/Measurement Each patient s preoperative and immediate postoperative clinical condition was classified as excellent (mrs 0, 1), good (mrs 2), or poor (mrs 3, 4, 5). Whether patients improved, were the same, or had worsened in comparison with their preoperative status was determined. For long-term outcome assessment, in all but 6 patients lost to follow-up, patients and families were asked to evaluate whether patients had TABLE 2. Preoperative Neurological Condition a improved, were the same, or had worsened in comparison with their preoperative status. Outcome was also evaluated by mrs, and poor outcome was defined as mrs.2. Various factors associated with good or poor outcome, including age, sex, location, number of hemorrhages, size and multiplicity of CM, preoperative neurological deficits, preoperative mrs, timing of surgery, surgical approach, and use of electrophysiology and stereotaxy, were evaluated. We specifically analyzed the subset of patients who developed HOD postoperatively. Statistical Methods Univariate analyses were conducted by using x 2 tests, Fisher exact tests, and Student t tests depending on the type of data being compared. Multivariate analysis of obliteration rates and complications were conducted by using standard logistic regression techniques. The level of significance was set at.05. All analyses were performed by using SAS software 9.2 (SAS Institute, Cary, North Carolina). RESULTS The majority of patients presented with multiple signs and symptoms. Cranial nerve deficits (51.1%), hemiparesis (40.9%), face or body numbness (34.7%), and cerebellar symptoms (38.6%) were the most common presenting symptoms. Other presenting features included headache (36.3%), diplopia (38.1%), swallowing difficulty (7.9%), seizures (5.7%), vertigo (11.4%), altered sensorium (6.3%), and involuntary movements (5.1%). Preoperatively, 44 patients (25.0%) were in excellent neurological status, 66 patients (37.5%) in good neurological status, and 66 (37.5%) in poor status. The preoperative mrs of the patients according to CM location is detailed in Table 2. More than onethird of the patients presented with mrs,2. The mean mrs of the patients preoperatively was 2.2, whereas the median mrs was 2. Patients with basal ganglia CMs were in significantly better neurological condition preoperatively than patients with CMs in other locations (P =.001). The mean mrs was also significantly lower in patients with basal ganglia CMs in comparison with patients with CMs in other locations. All (176) Basal Ganglia (26) Thalamus (16) Midbrain (28) Pons (92) Medulla (14) P Preop status, n (%) Excellent (mrs 0,1) 44 (25) 18 (69.2) 2 (12.5) 4 (14.3) 19 (20.6) 1 (7.1).001 Good (mrs 2) 66 (37.5) 7 (26.9) 6 (37.5) 16 (57.1) 32 (34.8) 5 (35.7) Poor (mrs 3-5) 66 (37.5) 1 (3.8) 8 (50.0) 8 (28.6) 41 (44.6) 8 (57.1) mrs, n (%) 0 5 (2.8) 2 (7.7) 1 (6.2) 1 (3.6) 1 (1.1) (22.2) 16 (61.5) 1 (6.2) 3 (10.7) 18 (19.6) 1 (7.1) 2 66 (37.5) 7 (26.9) 6 (37.5) 16 (57.1) 32 (34.8) 5 (35.7) 3 53 (30.1) 1 (3.8) 6 (37.5) 8 (28.6) 32 (34.8) 6 (42.9) 4 12 (6.8) 2 (12.5) 8 (8.7) 2 (14.3) 5 1 (0.6) 1 (1.1) a mrs, modified Rankin Score. NEUROSURGERY VOLUME 72 NUMBER 4 APRIL

4 PANDEY ET AL Cavernous Malformation A solitary CM occurred in 146 patients (83.0%), whereas 30 patients (17.0%) had multiple CMs. One patient had a CM in both the thalamus and basal ganglia, whereas 2 patients had 2 distinct CMs in the pons, both of which were operated on in both patients. Other patients had multiple small CMs in the supratentorial and infratentorial distribution. CMs ranged in size from 4 to 55 mm, (mean size, 15.5 mm). Basal ganglia and thalamus CMs tended to be larger than those located in the brainstem. CMs located in the medulla tended to be the smallest (Table 3). There was no significant difference, however, between mean sizes in various locations (P =.08). Bleeding Rates Of 176 patients, 172 presented with hemorrhage. The remaining 4 patients presented with seizures, all of whom had CMs located in the basal ganglia. There were 70 patients with a single hemorrhage and 102 patients with multiple hemorrhages (Table 4). There was no difference between the number of hemorrhages in various locations (P =.27). A total of 352 hemorrhages were observed in 172 patients over an observation period of patient-years. Thus, assuming that the lesions were present since birth, we calculated the annual retrospective hemorrhage rate as 5.1% per patient per year. This calculation may be incorrect, because it is now increasingly recognized that some CMs are acquired rather than congenital. Continuing with the same analyses, we calculated rebleed rates for every patient. For patients with a rebleed, the observation period is the period from the first bleed to surgery. For patients with a single bleed, the observation period is the interval between the first bleed and surgery. Seven patients underwent surgery without excision of their CM for the reasons mentioned below. None of them had rebleed in their follow-up period, and the follow-up duration was considered as the observation period to calculate the rebleed rate. Taking all of these into consideration, there were 180 rebleed episodes over an observation period of years, thus giving a rebleed rate of 31.5% per patient per year. Management Before coming to our institution, 13 patients (7.4%) received radiation therapy. These included helium-ion radiosurgery (6), conventional radiotherapy (3), and Gamma Knife radiosurgery (4). All these patients had rebleeding after they received radiation TABLE 3. Cavernous Malformation Size a therapy. Before presenting to our institution, 10 patients had undergone attempted surgical resection at another institution. Timing of Surgery Our preference was to perform surgery within 4 to 8 weeks of the last hemorrhagic episode. Owing to the referral pattern, however, some patients were operated on long after their hemorrhagic episodes. Time from the last hemorrhage to surgery ranged from 1 day to 10 years (mean time, 31.1 weeks). There were 122 patients (69.3%) who had surgery within 8 weeks of the last hemorrhage, whereas 73 patients (41.5%) had surgery within 4 weeks of the last hemorrhage. Surgery The surgical approach was determined according to the proximity of the CM to the pial or ependymal surface, and also to avoid critical tracts and nuclei. Standard microsurgical approaches were used for various lesions, depending on their locations. For the purpose of analysis, brainstem approaches were classified as posterior, where the entry point was the floor of the 4th ventricle, and lateral, where the entry point was lateral, dorsolateral, or occasionally anterolateral. Ninety-five patients (95/134, 70.9%) underwent the posterior approach, whereas 39 patients (29.1%) underwent the lateral approach. In 7 patients, the CM was not excised after exploration owing to its deep location in the brainstem and the lack of a safe entry corridor after mapping. In another 4 patients the CM was not resected at the first surgery for similar reasons. However, after another bleed, the CM presented to the pial surface and was then resected. Four patients had their CMs removed in 2 sittings through the same approach. In one of these patients with a thalamic CM, the first surgery was for hematoma evacuation, and the second surgery was for actual resection of the CM. Five patients experienced rebleeds after resection of their CMs. Of these, 1 patient rebled in the immediate postoperative period, and at repeat surgery there was residual CM that was not identified during the first surgery. The other 4 patients rebled from the malformation at 2 months, 8 months, 3 years, and 17 years after their initial surgery, and all were reoperated on for residual malformations. Overall, 159 patients (90.3%) had single surgeries, but 17 patients (9.7%) had 2 surgeries. Outcome Data Following surgery, 55 patients (31.3%) experienced new neurological deficits. The most common complications were increased CM All Basal Ganglia Thalamus Midbrain Pons Medulla P Size, mm Mean, mm Multiple 30 (17.0%) 5 (19.2%) 2 (12.5%) 1 (3.6%) 19 (20.7%) 3 (21.4%).2 a CM, cavernous malformation. 576 VOLUME 72 NUMBER 4 APRIL

5 SURGERY FOR DEEP-SEATED CAVERNOUS MALFORMATIONS TABLE 4. Number of Hemorrhages a No. of Hemorrhages All Basal Ganglia Thalamus Midbrain Pons Medulla P 0 4 (2.3) 4 (15.4) 1 70 (39.8) 13 (50.0) 5 (31.3) 8 (28.6) 38 (41.3) 6 (42.9) 2 48 (27.3) 4 (15.4) 6 (37.5) 11 (39.3) 23 (25.0) 4 (28.6) 3 41 (23.3) 5 (19.2) 4 (25.0) 6 (21.4) 23 (25.0) 3 (21.4) 4 7 (4.0) 1 (6.2) 1 (3.6) 5 (5.4) 5 3 (1.7) 2 (2.2) 1 (7.1) 6 1 (0.6) 1 (1.1) 7 1 (0.6) 1 (3.6) 8 1 (0.6) 1 (3.6) Single 70 (39.8).27 Multiple 102 (57.9) a Values shown are number (%). ataxia (21), diplopia related to 3rd or 6th nerve palsy or internuclear ophthalmoplegia (29), increased motor weakness (22), and 7th nerve paresis (10). Three patients (1.7%) died in the early postoperative period. One patient rebled from the residual CM in pons and was reoperated, but died of poor neurological status. Another patient had a cardiac arrest before we started the resection, presumably from a cardiac etiology. The third patient had a thalamic CM, and a postoperative hematoma, which was evacuated. All these patients were in poor neurological status at the time of presentation. Although almost one-third of patients (55 patients, 31.3%) worsened after surgery in their neurological status, 86 patients (48.9%) remained the same. Any patient with a new or worsened neurological deficit, minor or major, was considered worsened in our analysis. Thirty-five patients (19.9%) improved in their neurological deficits following resection of the hematoma and the lesion, and decompression of the functioning brainstem, basal ganglia, and thalamus. The immediate postoperative neurological status was determined to be excellent (23.9%), good (31.8%), and poor (42.6%). TABLE 5. Long-Term Outcome a,b Long-term Results Long-term follow-up in 170 of 176 patients ranged from 6 months to 19 years, with a mean of 3.5 years (Table 5). Each patient had at least a 6-month follow-up in the clinic. At the last follow-up, 105 patients (61.8%) had improved neurological status and 44 patients (25.9%) remained the same. Thus, 149 patients (87.6%) either improved or remained stable in their neurological status. Only 21 patients (12.4%) were worse than their preoperative status at the last follow-up. Of these patients, 1 worsened because of a glioblastoma, whereas 1 worsened because of the progression of his underlying Parkinson disease/dementia. Thus, 19 patients (11.2%) were permanently worse related to the resection of their CMs (Table 6). The mrs of patients at last follow-up was 0 to 2 in 140 patients (82.4%) and 3 to 6 in 30 patients (17.6%). The mean mrs of the patients was 1.6 in comparison with 2.2 preoperatively (Table 6). Thirty patients (17.6%) were found to have poor outcome as previously defined. Of note, only 2 of 107 patients (1.9%) in good neurological status preoperatively had poor long-term outcome, whereas 28 of 63 All Basal Ganglia Thalamus Midbrain Pons Medulla P Long term Improved 105 (61.8) 16 (66.7) 9 (60) 15 (53.6) 55 (60.4) 10 (83.3).04 Same 44 (25.9) 7 (29.2) 3 (20) 9 (32.1) 24 (26.4) 1 (8.3) Worse 21 (12.4) 1 (4.2) 3 (20) 4 (14.3) 12 (13.2) 1 (8.3) mrs 0 37 (21.8) 17 (70.8) 2 (13.3) 4 (14.3) 11 (12.1) 3 (25) (31.8) 6 (25) 6 (40) 9 (32.1) 29 (31.9) 4 (33.3) 2 49 (28.9) 1 (4.2) 3 (20) 8 (28.6) 33 (36.3) 4 (33.3) 3 15 (8.8) 1 (6.7) 5 (17.9) 9 (9.9) 4 6 (3.5) 2 (13.3) 3 (3.3) 1 (8.3) 5 1 (0.6) 1 (3.6) 6 8 (4.7) 1 (6.7) 1 (3.6) 6 (6.6) a mrs, modified Rankin Score. b Values shown are number (%). NEUROSURGERY VOLUME 72 NUMBER 4 APRIL

6 PANDEY ET AL TABLE 6. Predictors of Outcome a Factors Good Outcome (140) Poor Outcome (30) Univariate (P) Age,40 71 (87.6) 10 (12.4) (77.5) 20 Pediatric 17 (89.5) 2 (10.5).23 Adult 123 (81.4) 28 (18.6) Sex Male 57 (78.1) 16 (21.9).55 Female 83 (85.6) 14 (14.4) Hemiparesis Yes 48 (69.6) 21 (30.4) No 92 (91.1) 9 (8.9) Cranial nerve deficit 63 (72.4) 24 (27.6), (92.8) 6 (7.2) Ataxia 42 (63.6) 24 (36.4), (94.2) 6 (5.8) mrs Good 108 (98.2) 2 (1.8),.001 Poor 38 (54.8) 28 (45.2) Hemorrhage Single 65 (89) 8 (11).05 Multiple 75 (77.3) 22 (22.7) Time of surgery,8 wk 99 (84.6) 18 (15.4).04.8 wk 41 (77.4) 12 (22.6) Approach Lateral 30 (83.3) 6 (16.7).12 Posterior 74 (78.7) 20 (21.3) EP Yes 115 (80.1) 27 (19.9) No 25 (89.3) 3 (10.1).13 Stereotaxy Yes 108 (81.2) 25 (18.8).15 No 32 (86.5) 5 (13.5) Postop status Improved 32 (94.1) 2 (5.9) Same 70 (86.4) 11 (13.6) Worse 28 (62.2) 17 (37.8) a mrs, modified Rankin Score; EP, evoked potential. patients (44.4%) in poor neurological status preoperatively had poor outcome at long-term follow-up. Eight patients had died at long-term follow-up; 6 died within 6 months of surgery, but 1 patient each died at 18 months and 19 years postsurgery (Table 7). Thus, the long-term mortality rate was 7.6%, and the 6-month mortality rate (presumably attributable to surgery) was 4.7%. Seven patients who had exploration only for their CM had no rebleed over a mean follow-up period of 6.5 years, whereas 4 patients with exploration only had rebleed and underwent resection. As mentioned earlier, 5 patients had rebleeding after purported complete excision of the CM and were operated on again. At their follow-up imaging with MRI, 1 patient had a residual CM in the upper pons, which remained stable over a 6-year follow-up with no evidence of additional clinical or radiological hemorrhage. One patient had a residual CM in the pons, which was treated with radiosurgery. She had a rebleed 15 years later, however, and is planned for repeat surgery and excision. One patient had a recurrent, symptomatic CM in the thalamus, and was operated on again 17 years after the first surgery. In 4 patients with rebleeds after complete resection, MR (3) or CT (1) scans showed no residual CM, and, in 1 patient with an immediate postoperative rebleed, complete resection was based on intraoperative findings. No patients had residual CM or rebleeding on their follow-up radiological examination. Thus, the rebleed rate after presumed complete excision of the CM was 2.9%. Predictors of Outcome Various preoperative factors were analyzed for prediction of poor outcome following surgery for deep CMs. Age, sex, location, number of hemorrhages, size and multiplicity of CM, preoperative neurological deficits, preoperative mrs, timing of surgery, period of surgery, surgical approach, and use of electrophysiology and stereotaxy were evaluated. In the univariate analysis, older age (.40 years), multiple hemorrhages, surgery within 8 weeks of last hemorrhage, presence of preoperative cranial nerve deficits, ataxia or hemiparesis, and preoperative mrs 3 to 5 correlated with poor long-term outcome. Poor preoperative mrs was the strongest predictor of long-term outcome (P,.001). Other factors such as sex, surgical approaches, use of electrophysiology, and stereotaxy did not correlate to long-term outcome (Table 6). Patients with basal ganglia CMs had the best outcome (P =.02). The percentage of patients in various locations having good outcome were basal ganglia (100%), thalamus (73.3%), midbrain (75%), pons (80.2%), and medulla (91.7%), P =.056). In multivariate analysis with logistic regression analysis for all patients, preoperative mrs and time to surgery,8 weeks after hemorrhage correlated with better long-term outcome. For the patients with brainstem CMs, preoperative mrs, age,40 years, and time to surgery,8 weeks correlated with better long-term outcomes (Table 8). Other Analyses Hypertrophic Olivary Degeneration HOD was seen in 10 of the 134 patients with brainstem CMs (7.5%); it was present preoperatively in 1 patient and developed postoperatively in 9. There were 2 males and 8 females with ages ranging from 13 to 55 years (mean, 40.9 years). CMs were located in the pons of 9 patients and in the midbrain of 1 patient. A single hemorrhage occurred in 5 patients, whereas 5 patients had multiple hemorrhages. The mean number of hemorrhages was 2.1. Preoperative neurological status was excellent in 2 patients, good in 2 patients, and poor in 6 patients. Preoperative mrs was 1 (2 patients), 2 (2 patients), 3 (5 patients), and 4 (1 patient). Surgery was performed and approaches used were midline suboccipital (5 patients), lateral suboccipital (1 patient), and 578 VOLUME 72 NUMBER 4 APRIL

7 SURGERY FOR DEEP-SEATED CAVERNOUS MALFORMATIONS TABLE 7. Mortality Number Age/Sex Preoperative mrs Location Hemorrhages Cause of Death Time 1 60/f 3 Pons 3 Brainstem hemorrhage Perioperative 2 41/m 3 Pons 3 Brainstem injury Perioperative 3 43/f 4 Pons 4 Locked in syndrome 3 mo 4 60/m 3 Thalamus 3 Postoperative hematoma Perioperative 5 43/f 3 Pons 3 Cardiac arrest Perioperative 6 70/f 4 Pons 7 Brainstem injury, C difficile infection 4 mo 7 60/m 3 Pons 3 Hypertensive ganglionic bleed 18 mo 8 36/f 3 Pons 3 pneumonia 19 y mrs, modified Rankin Score; f, female; m, male. subtemporal transpetrous (4 patients). One patient had a known partial excision of the CM and was operated on again through the same route 2 years later after a repeat hemorrhage from the residual CM. One patient had a repeat hemorrhage after CM resection and was reoperated on 2 months after the first surgery (subtemporal approach) through a suboccipital approach. Postoperatively, 5 patients remained the same and 5 patients worsened in their neurological status. Two patients had significant hemiparesis postoperatively, while 5 patients had internuclear ophthalmoplegia and increased gait ataxia. One patient with a large brainstem hemorrhage, who presented to us 4 years after the hemorrhage, had features of HOD preoperatively on MRI. The other patients developed delayed HOD, first seen on their 6-month follow-up MRI. Three patients demonstrated typical symptoms of HOD palatal myoclonus, nystagmus, and oscillopsia. In addition, 1 patient had a limb tremor and another patient had hemiballismus. The tremor was well controlled with clonazepam; however, the oculopalatal myoclonus and hemiballismus were poorly responsive to therapy. All the patients had poor response to Botox injections. In 2 patients, T2 hyperintensity occurred in bilateral olives, whereas 1 patient had unilateral HOD ipsilateral to the side of the CM. There were 6 patients with residual deficits at the follow-up period, from the preoperative period, and from the increased deficits postoperatively. However, they did not have oculopalatal myoclonus or oscillopsia. Two patients had unilateral HOD ipsilateral to the side of the CM, whereas 4 patients had bilateral TABLE 8. Multivariate Analysis: Predictors of Good Outcome a All Sites HOD. One patient improved to his neurological baseline, and HOD was seen on routine postoperative imaging. He never manifested any signs or symptoms of HOD. Of 5 patients in whom serial imaging was available, 2 showed persistence of T2 hyperintensity, 2 showed decreased but persistent hyperintensity, whereas, in 1 patient, T2 hyperintensity disappeared on followup imaging. DISCUSSION Key Results A total of 176 patients with deep CMs were included in this study. The most common presenting symptoms were cranial nerve deficits (51.1%), hemiparesis (40.9%), numbness (34.7%), and cerebellar symptoms (38.6%). Hemorrhage was the presenting symptom in 172 patients (70 single, 102 multiple). The annual retrospective hemorrhage rate was 5.1% (assuming CMs are congenital with uniform hemorrhage risk throughout life); the rebleed rate was 31.5%/patient per year. Surgical approach depended on the proximity of the CM to the pial or ependymal surface. Postoperatively, 121 patients (68.8%) had no new neurological deficits. Follow-up occurred in 170 patients. Delayed postoperative HOD developed in 9/134 (6.7%) patients with brainstem CMs. HOD occurred predominantly following surgery for pontine CMs (9/10 patients). Three patients with HOD had palatal myoclonus, nystagmus, and oscillopsia, whereas 1 patient each had limb tremor and hemiballismus. At follow-up, 105 Brainstem Sites Factors P OR 95% CI P OR 95% CI Age (,40) Sex (male) Preop mrs (,3), , Single/multiple hemorrhage Surgery,8 wk Approach (lateral/posterior) NA NA NA a OR, odds ratio; CI, confidence interval; mrs, modified Rankin Score; NA, not available. NEUROSURGERY VOLUME 72 NUMBER 4 APRIL

8 PANDEY ET AL patients (61.8%) improved, 44 (25.9%) were unchanged, and 19 (11.2%) worsened neurologically. Good preoperative mrs (98.2% vs 54.5%, P =.001) and single hemorrhage (89% vs 77.3%, P,.05) were predictive of good long-term outcome. hemorrhage rate ranging from 21% to 60% per patient per year. This is similar to our observed rebleed rate of 31.5%, confirming that once these lesions become symptomatic, they have an aggressive course (Table 9). Interpretation and Generalizability Understanding the natural history of CMs, particularly those in the brainstem, thalamus, and basal ganglia, is vital for decision making regarding these complex lesions. Unfortunately, the natural history of the disease, particularly the brainstem and deep CMs, remains unclear, and most of the available data are from retrospective series. 2-5,7 A number of series have presented widely varying bleed and rebleed rates, and hence have proposed a broad range of indications for surgery and recommendations for management. In the absence of any prospective study, the hemorrhage rates are calculated based on retrospective analysis of bleed rates based on the assumption that the CMs are congenital lesions. There is a caveat to this assumption, because many of the CMs are now shown to be acquired lesions. It is unclear whether the hemorrhage rates for brainstem CMs are any different than the supratentorial CMs. In the prospective study by Kondziolka et al, 35 there was no difference between hemorrhage rates in various locations. However, Porter et al 6 demonstrated higher annual rates of symptomatic hemorrhage (10.6%) for deep compared with superficial lesions (0%). This can be explained by higher eloquence of the deeper structures, with even a small bleed causing symptoms, in comparison with the supratentorial lesions, where a small bleed may go unnoticed. We observed a retrospective bleed rate of 5.1%, which is comparable to the rates from Porter et al, 7 Wang et al, 52 and Abla et al, 13 while being higher than some of the other series. We recognize that there are inherent biases in such calculations, such as the assumption that these lesions are congenital, preselection for patients who have bled, and varying definitions of hemorrhage in different series. This calculation also assumes uniform bleeding rates from the CMs, which may not be true. We acknowledge that it is likely that the annual bleeding rate of 5.1% would not apply to small incidental brainstem CMs found on routine MR imaging. For this reason, we do not operate on asymptomatic brainstem CMs. However, it is clear from the literature that once deep CMs become symptomatic, they have a high rate of repeat hemorrhage causing increasing neurological deficits. A majority of the authors have found a repeat TABLE 9. Literature Review: Bleed and Rebleed Rates Limitation The limitation for this analysis is that this high rate of repeat hemorrhage may be overestimated by the presence of a few patients who have repeated hemorrhages over a short period of time. It must be acknowledged, however, that 7 patients in our series had their malformations explored, but not excised, and none had any hemorrhage over a follow-up period. Thus, some lesions remain quiescent after 1 bleed, but other lesions bleed repeatedly and cause significant neurological worsening. Unfortunately, there are no predictors of repeat hemorrhage described in the literature. Surgical Indications Many authors have proposed indications for intervention and surgery. Barring a few reports, most authors recommend surgery for symptomatic and accessible lesions. Tarnaris et al 53 suggested that conservative management for brainstem CMs have a better long-term outcome as compared to the surgical outcome. This analysis, however, was based on a small group of patients (21), of which only 6 underwent surgery. Also, 8 of the 15 patients undergoing conservative treatment had worse long-term outcome, which resulted in poorer neurological outcome than what is described in various surgical series, including ours. Most of the CMs we operated on were near a pial or ependymal surface. However, as Mathiesen et al 26 reported, some CMs, especially in the basal ganglia and thalamus but also the brainstem, did not present to the pial/ependymal surface, yet resection could be done with acceptable morbidity. We operated on 4 pontine CMs deep in the pons that did not present close to any CSF-brainstem interface by using the safe entry corridors described before. 27,54 Previously, many authors suggested waiting for 2 symptomatic bleeds before any treatment was considered for these deep lesions. 37,39 However, consistent with more recent reports, we now often prefer to operate on an accessible CM after a single clinical hemorrhage, keeping in mind the aggressive natural history of symptomatic deep CMs. We also found that patients with single bleeds had a 10.9% chance of poor long-term Series No. of Patients Bleeding Rate, % Rebleeding Rate, % Fritschi et al, Porter et al, Kupersmith et al, Wang et al, Ferroli et al, Abla et al, (in patients with single bleed), 35 in patients with.1 bleed Current series VOLUME 72 NUMBER 4 APRIL

9 SURGERY FOR DEEP-SEATED CAVERNOUS MALFORMATIONS outcome, compared with patients with multiple bleeds who had a 22.7% chance of poor outcome (P =.05). We do not recommend operating on asymptomatic incidental deep CMs, unlike some authors, primarily because the natural history of these lesions is still poorly understood and mostly calculated from retrospective series, and these are formidable lesions, with significant short- and long-term morbidity. Timing of Surgery Mathiesen et al 26 found a significant difference in outcome between the results of early and late surgery, and reported that neurological deterioration was detectable in 15 of 17 patients treated late, compared with 4 of 12 patients treated early. However, they did not analyze the details of surgery timing on long-term patient outcomes. In the present series, the rate of immediate neurological worsening was higher in patients operated on after 4 weeks following their last hemorrhage (37.5%) in comparison with those operated on within 4 weeks of hemorrhage (26%). Also, the rate of immediate neurological improvement was higher in patients operated on,4 weeks from ictus (26%) in comparison with those operated on.4 weeks from ictus (15%). It is possible that removal of the hemorrhagic mass leads to improvement of neurological symptoms, which is more likely in early rather than in delayed surgery. In the present series, patients operated on within 8 weeks of the last hemorrhage had a better outcome compared with those operated on after 8 weeks (P =.04). Hence, similar to other authors, we advocate early surgery between 4 to 8 weeks of ictus, unless there is a large hemorrhage or an acutely worsening patient in whom immediate surgery might be indicated. The presence of a subacute hematoma aids in the surgical dissection. The evacuation of subacute blood also hastens neurological improvement in some cases. Owing to the referral pattern, however, only 42.5% of the patients in this series had surgery within 4 weeks of ictus, whereas 69.3% underwent surgery within 8 weeks of ictus. Surgical Approach Surgical approach to deep CMs depends on various factors, including the location, the proximity to the pial/ependymal surface, and the presence of safe trajectory and entry zones for resection of these lesions (Figure 1). Superficial lesions are usually associated with good outcome, and at surgery the malformation is easily visible as a dark reddish-blue area. It is important to perform brainstem functional mapping before an incision on the pial surface of the brainstem to avoid the cranial nerve nuclei. As mentioned previously by us and other authors, T2-weighted images provide a false sense of proximity of the malformation to the surface because of ferromagnetic properties of the hemosiderin with blooming of blood products, and hence we rely on T1-weighted images to plan the surgery. Some authors have reported better results by using an anterolateral pontine approach in comparison with a posterior approach through the floor of the fourth ventricle. Ferroli et al, 8 in their analysis of 52 patients with brainstem CMs, reported that the incidence of new postoperative deficits was lower when the malformation was resected through an anterolateral pontine surface in comparison with the floor of the fourth ventricle (5% vs 29%, P =.035). We have not experienced the same benefit for this approach. Through the use of brainstem mapping to localize the cranial nerve nuclei and intraoperative navigational assistance, we have achieved excellent results performing surgery through the floor of the fourth ventricle (Figure 2). We compared long-term results for brainstem CMs operated on via the lateral or posterior route and found similar rates of good outcomes (P =.12). We prefer the approach that traverses through the least amount of brainstem tissue safely. Both the suprafacial and infrafacial triangles described before are good entry points for pontine lesions, whereas, for mesencephalic lesions, a triangle between the facial colliculus and oculomotor nuclei has been described as a safe entry zone. 54 It is critical to respect the midline in mesencephalic and pontine lesions, because injury to the medial longitudinal fasciculus may lead to bilateral internuclear ophthalmoplegia, which can be devastating for the patient. For anterolateral pontine/mesencephalic lesions, we use the subtemporal transpetrous approach with an entry point either superior to the 5th nerve exit, or between the 5th and 7th nerves. We will occasionally use an anterior transsylvian exposure for CMs residing in the anterior medial mesencephalon, often using a contralateral approach to take advantage of a better trajectory (Figure 3). We approach thalamus and basal ganglia lesions either through the ventricle with the use of a transcallosal approach, or through the least eloquent cortex (Figure 4). 15,22,26,55 As mentioned before, these lesions can be resected even though they do not present to the ependymal surface. Surgical Adjuncts Surgical adjuncts such as intraoperative electrophysiology and neuronavigation are invaluable tools to achieve good results for these difficult lesions. Intraoperative neuronavigation is useful for localization of the brainstem CMs that do not present to the pial surface, for planning a safe trajectory to the deep thalamic and ganglionic CMs, and for avoiding injury to critical brain tissue during resection. 56,57 We have not found brain-shift with cerebrospinal fluid drainage a major concern, because these are central lesions. Intraoperative electrophysiology is also extremely valuable. Our monitoring routinely includes electroencephalogram, bilateral somatosensory evoked potential, and brainstem auditory evoked potential for every patient, and cranial nerve nuclei mapping when appropriate. In the second half of the study, we routinely used bilateral motor evoked potentials. Continuous monitoring of sensory, motor, and brainstem pathways has allowed early detection of excessive retraction and manipulation of critical structures. As has been the experience of other authors, brainstem mapping of cranial nerve nuclei and fibers has reduced morbidity by localization of these critical structures and has minimized injury to them caused by pial incisions, resections, and excessive retraction and by NEUROSURGERY VOLUME 72 NUMBER 4 APRIL

10 PANDEY ET AL FIGURE 1. Microsurgical approach without CM presentation to pial or ependymal surface. This 42-year-old man presented with headaches suggestive of migraines. Sagittal (A) and axial (D) T1-weighted MRI images revealed a small CM in the pons, not reaching a pial or ependymal surface. He was advised to be followed clinically and radiologically. Over the next 2 years, he experienced 2 additional clinical hemorrhages, which manifested as right face and left body numbness, right side facial paresis, and diplopia. Sagittal (B) and axial (E) T1-weighted MRI image of the brain revealed a pontine CM, which had enlarged in size, but still did not present to the pial or ventricular surface. In view of repeated hemorrhages and neurologic deficits, the patient was operated via a right subtemporal approach with the pial incision on the lateral surface of the brainstem, lateral to the V nerve entry zone. Postoperatively, the patient had transient left-sided weakness. However, he made a complete neurologic recovery over the first postoperative month. Postoperative sagittal (C) and axial (F) MRI showed complete excision of the CM. manipulation of the pial surface during resection. 51,56 We also believe mild intraoperative hypothermia may improve outcome, although this is unproven. Mild hypothermia is achieved by applying a cooling blanket and decreasing the core body and brain temperature to 32 to 33 C. This technique has been used for the majority of our patients who undergo craniotomies to treat intracranial vascular lesions with good results overall. 58 No previous authors have analyzed the effects of surgical adjuncts on long-term patient outcomes. Many authors have reported good results without the use of these adjuncts. We did not find any significant difference in long-term outcome of patients treated with or without the use of stereotaxy or electrophysiology. However, we believe that this lack of significance may be due to a low rate of complications. Surgical Results and Complications Surgical results and complications reported by other authors are detailed in Table 10. It is difficult to compare outcomes of our series with other series because of different patient populations, timing of surgery, hemorrhage and rehemorrhage rates, follow-up period, and outcome scales. In the present series, 7.9% of patients were in a worse neurological status at follow-up in comparison 582 VOLUME 72 NUMBER 4 APRIL

11 SURGERY FOR DEEP-SEATED CAVERNOUS MALFORMATIONS FIGURE 2. Supracerebellar/infratentorial approach. A 51-year-old man presented with headache, confusion, and diplopia. On examination, he had a left 3rd nerve palsy and a right hemiparesis. Axial (A) and sagittal ( B) T1-weighted MRI images revealed a cavernous malformation in the left dorsal midbrain. The patient was operated on via a paramedian supracerebellar infratentorial approach, and complete excision of the CM was achieved. He did not have any new neurological deficits postoperatively and at 6-month follow-up revealed residual 3rd nerve paresis. C, Postoperative axial T1-weighted MRI images showed complete excision of the lesion. FIGURE 3. Contralateral approach to anterior midbrain CM. A 48-year-old woman presented with sudden-onset right hemiparesis, dysarthria, and a partial left 3rd nerve palsy. A, axial T1-weighted MRI demonstrated a cavernous malformation with acute hemorrhage located in the left cerebral peduncle of the midbrain. The CM presented to the pial surface of the medial aspect of the cerebral peduncle. Microsurgical resection of the lesion was performed 6 weeks after the initial hemorrhage via a right (contralateral) orbitozygomatic transsylvian approach and complete resection of the malformation was achieved, as seen on the postoperative axial T1-weighted MRI ( B). She was unchanged neurologically immediately postoperatively and showed marked improvement in her right hemiparesis over the next 3 months. NEUROSURGERY VOLUME 72 NUMBER 4 APRIL

12 PANDEY ET AL FIGURE 4. Transsylvian approach after failed radiosurgery treatment. A 49-year-old man presented with sudden onset of left hemiparesis, which improved over the next 2 months. An MRI image of the brain done at that time at an outside institution revealed a right basal ganglia CM. He was treated with Gamma knife radiosurgery. Seven years after radiosurgery, the patient developed new-onset left hemiparesis, which progressively worsened. Axial T1-weighted MRI image of the brain following the onset of hemiparesis (A) and 3 months later after further deterioration (B), showing increase in the size of the malformation, as well as evidence of blood in different stages of chronicity. He underwent a right frontotemporal craniotomy with transsylvian approach and complete excision of the malformation. Postoperatively, the patient s hemiparesis was unchanged and subsequently improved over the next 8 months to the point he was ambulating with a walker. Postoperative axial T1-weighted MRI (C) revealed complete excision of CM. with their preoperative condition. We have used a strict definition for deficits and have included both mild and severe residual deficits. Eight patients died, 6 within 6 months of surgery, and 1 each died at 18 months and 19 years follow-up. Thus, the 6-month mortality rate was 3.4%. Ferroli et al, 8 in their experience with 52 patients with brainstem CMs, described a morbidity and mortality rate of 19% and 1.9%, respectively. They used the Karnofsky Performance Scale as the outcome scale. In a large series of 137 patients by Wang et al, 29 morbidity and mortality rates were 27.7% and 0%, respectively. In their series of 100 patients with brainstem CMs, Porter et al 7 reported a morbidity and mortality rate of 10% and 4%, respectively. In the largest series of 300 patients (260 adults and 40 pediatric TABLE 10. Literature Review: Morbidity and Mortality patients), Abla et al 13,14 reported that 112 patients experienced a new neurological deficit (37.3%). The mortality in their series was 3%. Other authors have reported similarly good results with surgery for deep CMs. In our series at long-term follow-up, 83% of patients had good neurological status, and only 17% of patients were mrs $ 3. Two of these 30 patients were in poor neurological status related to other diseases (Parkinson disease with dementia in 1 patient, and glioblastoma multiforme in another), whereas 1 patient died of a hypertensive bleed unrelated to the CM. Importantly, only 2 patients who were in good neurological condition preoperatively worsened and had a poor mrs. All other patients with poor outcomes had an mrs $3 preoperatively, and retained a poor mrs after surgery. Series Patients Surgical resection Morbidity, % Mortality, % Zimmerman, Fritschi, Porter, Steinberg, Wang, Ferroli, Hauck, Abla, Present series VOLUME 72 NUMBER 4 APRIL

13 SURGERY FOR DEEP-SEATED CAVERNOUS MALFORMATIONS Radiosurgery Radiosurgery has been advocated as an alternative treatment to brainstem, thalamic, and ganglionic CMs, which are located in high-surgical-risk areas of the brain. Many authors have reported their experience with stereotactic radiosurgery (SRS) with varying success Our combined Stanford University-University of California, Berkeley, program previously treated 57 angiographically occult vascular malformations (AOVMs) in these locations, using Bragg peak helium ion radiosurgery (47 AOVMs) or linear accelerator radiosurgery (10 AOVMs), with mixed results. 32 Of 57 patients, 18 (31.6%) experienced symptomatic bleeding (20 hemorrhaging episodes) after radiosurgery; 16 hemorrhages occurred within 36 months after radiosurgery (9.4% annual bleeding rate; 16 hemorrhaging episodes/171 patient-years), and 4 hemorrhages occurred more than 36 months after treatment (1.6% annual bleeding rate; 4 hemorrhaging episodes/257 patient-years) (P,.001). Recently, Lunsford et al 39 reported their experience of Gamma Knife radiosurgery with 103 patients with CMs considered at high risk for resection. Before SRS, the annual hemorrhage rate was 32.5%. The hemorrhage rate in the first 2 years following SRS was 10.8%, but it dropped to 1.06% subsequently, with 2 patients dying of repeat hemorrhage. Also, 13.5% of patients developed new neurological deficits owing to radiation effects following SRS. Almost one-fourth (24.4%) of patients who had radiosurgery as their primary management experienced rebleeds after SRS. Amin-Hanjani et al 31 described 95 patients with 98 CMs treated with stereotactic Bragg-peak proton beam therapy and found that the annual hemorrhage rate was reduced from 17.3% before radiation to 4.5% following radiosurgery after a latency period of 2 years, although a permanent neurological deficit developed in 16%, and 3% died. It is unclear if the protection from hemorrhage described 2 years after SRS is an effect of radiation or simply the natural history of the CM, which undergoes repeated hemorrhage for some interval and then ceases to bleed. Matheisen et al 26 did not report any benefit in their 5 patients who underwent radiosurgery and reported that the incidence of bleeding in radiosurgically treated CMs was even higher than that documented in untreated incidental or untreated symptomatic CMs. Gewirtz et al 59 analyzed 11 surgically removed CMs that were previously irradiated and did not find histological evidence of CM vessel occlusion in any patient. In other authors opinions, 40 some CMs suitable for radiosurgery exist, although most of the so-called surgical high-risk CMs would have a good surgical result at centers that have substantial experience in managing these lesions. We prefer microsurgery over radiosurgery for symptomatic CMs in the brainstem, thalamus, and basal ganglia. Predictors of Outcome Other authors have discussed the predictors of outcome for CMs involving the brainstem, thalamus, and basal ganglia. Hauck et al 23 studied 44 patients with brainstem CMs and found that presentation with hemiparesis, more than 1 pretreatment hemorrhage, poor preoperative mrs, and location in pons/ medulla predicted poor outcomes. In multivariate analysis, only preoperative mrs scores predicted outcome (P =.015). Samii and colleagues 10 reported that higher preoperative Karnofsky Performance Scale scores and smaller volume lesions had better outcome, but timing of surgery and multiple hemorrhages did not influence outcome. Bruneau et al 18 reported a series of 22 patients and found that late surgery and multiple hemorrhages predicted poor outcome. In our study, older age (.40 years), presence of preoperative cranial nerve deficits, ataxia or hemiparesis, poor preoperative mrs, surgery more than 8 weeks after last hemorrhage, and multiple hemorrhages correlated with poor long-term outcome in univariate analysis. In multivariate analysis, age.40 (for brainstem CM sites), poor preoperative mrs, and surgery. 8 weeks after last hemorrhage were significant predictors of worse outcome. Poor preoperative mrs was the strongest predictor of long-term outcome (P,.001). Other factors such as sex, surgical approaches, use of electrophysiology, and stereotaxy did not correlate to long-term outcome. Patients with basal ganglionic CMs had the best outcome; however, there was no difference in outcome between various brainstem sites. Hypertrophic Olivary Degeneration HOD is a rare phenomenon that appears after a variety of insults to the dentate-rubro-olivary pathway of the brainstem. It has been previously reported for brainstem hemorrhage, trauma, tumor, demyelination, and infarcts. HOD was first described by Oppenheim, as pathological enlargement of the inferior olivary nucleus. Anatomically HOD has been associated with injury to the dentate nucleus, superior cerebellar peduncle, or the central tegmental tract. Guillain and Mollaret 60 described a triangle, that connects the dentate nucleus of cerebellum, the red nucleus of mesencephalon, and the inferior olivary nucleus of the medulla, and insults to these tracts can lead to HOD. The proposed mechanism for this pathology is transsynaptic degeneration, and the inferior olivary nucleus shows hypertrophy of both neurons and glia. 61,62 Although HOD has been described following trauma, hemorrhage, metastasis, and degenerative pathology, 11,42,43,62 few reports exist of this complication following surgery for CMs. Before this, there were 8 patients reported with HOD following surgery for brainstem CMs (all located in the pons) in 4 case reports and a small series of 4 patients. 41,45-47,63 In the previously mentioned large series of surgery for brainstem CMs, this complication was not mentioned separately. In the present series, 9 of 10 patients with HOD had pontine CMs, whereas 1 patient had a mesencephalic CM. Four had unilateral HOD and 3 had bilateral HOD. Typical symptoms, including oculopalatal myoclonus and tremors, occurred in 4 of the 8 cases previously reported. In the largest series reported by Hornyak et al 41 3of4 patients had palatal tremor. In our series, 10 of 134 patients with brainstem CMs (7.5%) had HOD, making it the largest series to report this particular pathology. However, only 3 patients NEUROSURGERY VOLUME 72 NUMBER 4 APRIL

14 PANDEY ET AL presented with new-onset symptoms like oculopalatal myoclonus and tremor, whereas other patients had significant deficits, either preoperative or postsurgical, which persisted in the follow-up period. Interestingly, most of the patients (70%) with HOD presented preoperatively with profound deficits and poor neurological status. The temporal evolution of MR findings accompanying HOD has been described before. 64 The essential MR features are increasing T2 signal in the inferior olives, which starts around 3 to 4 weeks following the insult, and may persist indefinitely or at least for 3 to 4 years. The other feature is olivary hypertrophy, which starts 6 months after the insult, and persists up to 4 years (Figure 5). In the current series, serial MR scans were available in 5 of the 10 patients, and increased T2 signal persisted in 4 patients, whereas it subsided after 1 year in 1 patient. The management of symptoms following HOD is controversial and not very effective. Benzodiazepines, carbamazepine, and 5-hydroxytrytophan all have been tried with limited efficacy. 41 In our patients, we tried clonazepam and Botox injections with limited efficacy. Shepherd et al 65 described a patient with refractory tremor and HOD managed with deep brain stimulation. Indeed, we offered deep brain stimulation to 2 of our patients FIGURE 5. Hypertrophic olivary degeneration. A 57-year-old woman presented with 3 distinct episodes of clinical hemorrhage resulting in diplopia, left body numbness, and left hemiparesis. Sagittal (A) and axial (D) T1-weighted MRI images revealed a pontomedullary CM, presenting to the 4th ventricular surface. The patient was operated on via a midline suboccipital approach, and complete excision of the lesion was performed. Postoperatively, she developed internuclear ophthalmoplegia, facial nerve palsy, and right hemisensory deficits. She had some improvement in extraocular movements and sensation over the initial 3 postoperative months. She developed a delayed onset of ataxia and tremor of the right upper extremity approximately 4 months after surgery. On follow-up, these deficits did not improve, and she remained neurologically worse in comparison with her preoperative status. Postoperative sagittal (B) and axial (E) T1-weighted MRI images revealed complete excision of the CM. A 2-month postoperative T2-weighted axial MRI of the medulla (C) showing no evidence of hypertrophic olivary degeneration. On repeat MRI at 6 months postoperatively, there was focal hyperintensity on T2-weighted MRI (F) in the ventral medulla on the left side characteristic of hypertrophic olivary degeneration (arrow). 586 VOLUME 72 NUMBER 4 APRIL