St e r e o ta c t i c radiosurgery has been shown to be an

Transcription

1 J Neurosurg 113:53 64, 2010 Neurological complications and symptom resolution following Gamma Knife surgery for brain metastases 2 cm or smaller in relation to eloquent cortices Clinical article Ro b e r t E. El l i o t t, M.D., 1 St e p h e n Ru s h, M.D., 2 Am r Mo r s i, M.D., 1,2 Ni s h a Me h ta, B.A., 1 Je r i Sp r i e t, B.S., 1 As h wat h a Na r aya n a, M.D., 2 Be r n a d i n e Do n a h u e, M.D., 2,3 Er i k C. Pa r k e r, M.D., 1 a n d Jo h n G. Go l f i n o s, M.D. 1 Departments of 1 Neurosurgery and 2 Radiation Oncology, New York University Langone Medical Center, New York; and 3 Department of Radiation Oncology, Maimonides Medical Center, Brooklyn, New York Object. Reports on resection of tumors in or near eloquent cortices have noted neurological complications in up to 30% of patients. This paper contains an analysis of symptom resolution and neurological morbidity following 20-Gy Gamma Knife surgery (GKS) for supratentorial brain metastases 2 cm in greatest diameter. Methods. The authors performed a retrospective analysis of 98 consecutively treated adults (33 men and 65 women with a median age of 61.4 years at the time of GKS) with Karnofsky Performance Scale score 60, who underwent GKS for supratentorial brain metastases 2 cm in diameter. Lesion location was classified as noneloquent (Grade I), near eloquent (Grade II), or eloquent (Grade III), in accordance with the grading system developed by the group at M. D. Anderson Cancer Center. Following treatment, the patients underwent MR imaging and clinical examinations at 6 weeks and every 3 months thereafter. Results. Ninety-eight patients underwent 20-Gy GKS for 131 metastases at initial presentation and 31 patients underwent salvage 20-Gy GKS for 76 new lesions, for a total of 207 lesions (mean lesion volume 0.44 cm 3 ). Lesions were classified as follows: Grade I, 96 (46.4%); Grade II, 51 (24.6%); and Grade III, 60 (29%). Fifteen patients (2 with Grade II and 13 with Grade III lesions) presented with deficits referable to their lesions, yielding pre-gks deficit rates of 7.2% per lesion and 15.3% per patient. The pre-gks deficits improved or resolved in 10 patients (66.7%) at a median time of 2.8 months and remained stable in 3 patients (20%). Two patients (13.3%) experienced worsened neurological deficits. One patient who was neurologically intact prior to treatment developed a new hemiparesis (1 of 83 patients [1.2%]). The rates of permanent neurological deterioration following GKS for Grades I, II, and III lesions were 0% (0 of 96 tumors), 2% (1 of 51), and 3.3% (2 of 60), respectively. The pre-gks neurological deficits and larger lesions were the most significant risk factors for post-gks neurological deterioration. Conclusions. Gamma Knife surgery performed using a 20-Gy dose provides amelioration of neurological deficits from brain metastases that are 2 cm in diameter and located in or near eloquent cortices in nearly two-thirds of patients with a low incidence of morbidity. Consistent with the surgical literature, higher rates of neurological complications were observed as proximity to eloquent regions and lesion size increased. There was no neurological deterioration in patients harboring metastases in noneloquent areas. (DOI: / GKS101047) Ke y Wo r d s stereotactic radiosurgery complication neurological deficit eloquent cortex St e r e o ta c t i c radiosurgery has been shown to be an effective treatment modality for patients with brain metastases. 3,7,14,24,34,39 It is a minimally invasive treatment that offers precise, high-dose radiation to small target volumes and can safely be used to treat multiple lesions in the brain with minimal toxicity. Stereotactic radiosurgery was originally used for lesions 3 cm in maximal diameter and lesions not surgically accessible, Abbreviations used in this paper: GKS = Gamma Knife surgery; KPS = Karnofsky Performance Scale; RPA = recursive partitioning analysis; RTOG = Radiation Therapy Oncology Group; SRS = stereotactic radiosurgery; WBRT = whole-brain radiation therapy. J Neurosurg / Volume 113 / December 2010 but it is now routinely used for patients with multiple metastases regardless of whether WBRT is used. 1,3,4,7,12, 15,20,24,34,35,39,41,42,51,52 Approximately 70 to > 90% of brain metastases can be controlled using SRS, but there is a risk of radiationinduced edema or necrosis related to radiosurgical dose, lesion volume, and prior WBRT. 17,23,26,31,39,40 Radiosurgical dosing varies widely across the literature and often depends on lesion size, history of prior WBRT, tumor pathological characteristics, institutional preference, and proximity of the lesion to critical areas of the brain. The RTOG Protocol posited 24 Gy to be the maximum tolerated dose of single-fraction radiosurgery for solitary, previously irradiated brain metastases 2 cm. 38 Authors 53

2 R. E. Elliott et al. of the protocol recommended lower doses for larger tumors, given the increased rate of moderate-to-severe neurotoxicity associated with these lesions. No controlled studies, however, have produced the optimal dose for brain metastases of any size, and some investigators have reported increased complications when doses > 20 Gy are used. 26,40 Another limitation common to both current radiosurgical and surgical literature is the scant reporting of acute and delayed neurological deterioration and symptom resolution following SRS in relation to the functional location of treated lesions. 30,37,52 To address this issue, Sawaya and colleagues 37 created a 3-tiered grading scale to determine the proximity of brain tumors to eloquent brain regions. Analyzing their own experience with radiosurgical and surgical treatment of brain metastases, they demonstrated higher rates of neurological dysfunction in eloquent regions than in near-eloquent or noneloquent regions. However, investigators from most centers only report overall complications, and few authors have analyzed the rates of neurological improvement and risk of deterioration following SRS for brain metastases in the context of their proximity to major sensory, visual, speech, and motor regions. 47 In 2001, our center began using a uniform dose of 20 Gy for all brain metastases 2 cm in all patients, regardless of the lesion s proximity to eloquent cortices. We report our experience with a large, consecutive series of functionally independent patients with supratentorial metastases who underwent treatment with 20-Gy GKS. We address the efficacy of the 20-Gy dose in achieving improvements in neurological symptoms and incidence of neurotoxicity, and we analyze the risk factors for neurological deterioration following GKS. Methods Study Design and Eligibility Criteria In 2001, the GKS protocol at New York University Langone Medical Center standardized to radiation doses of 20 Gy for metastases 2 cm and 18 Gy for lesions 2 3 cm in maximum diameter. Following institutional review board approval, the records of all patients with metastatic brain tumors who underwent GKS at our center between 2001 and 2009 were retrospectively reviewed. Similar to the modified Pittsburgh protocol, 14 general eligibility criteria included the following: 1) adults with 1 to 3 cerebral metastases; 2) a maximum tumor diameter of 2 cm for all lesions; 3) KPS Score 70 (functionally independent); 4) estimated life expectancy 4 months; and, unique to our study, 5) no history of prior WBRT. Three patients with KPS Score 60 were included in this study, since their deficits were neurological in nature and directly related to a metastasis causing hemiparesis, and because there was a chance for improvement following treatment. Between February 2001 and January 2009, 101 consecutively treated adults underwent GKS as the sole treatment for between 1 and 3 brain metastases 2 cm in diameter. We reviewed all imaging studies at the time of GKS to determine the size and location of all treated sites of metastasis. Lesion location was classified as noneloquent (Grade I), near eloquent (Grade II), or eloquent (Grade III) in accordance with the grading system posited by Sawaya and colleagues. 37 Adapting this grading system, we categorized noneloquent locations as including frontal and temporopolar regions as well as the right parietooccipital area; near-eloquent locations as including the corpus callosum as well as areas near the motor and sensory cortices, calcarine fissure, and speech center; and eloquent locations as including the motor and sensory cortices, visual and speech centers, and the internal capsule, basal ganglia, hypothalamus, and thalamus. The central sulcus, sylvian fissure, superior and middle frontal sulci, intraparietal sulcus, pars marginalis of the cingulate sulcus, and calcarine fissure were identified using standard multiplanar CT and MR imaging landmarks. 29 The location of all tumors was determined using such anatomical landmarks and in consultation with a neuroradiologist. Data were retrospectively collected by reviewing office and inpatient records as well as pretreatment and serial post-gks MR imaging and/or CT studies. Patient characteristics, KPS scores, RPA classes, and prior treatments were recorded, as were data on tumor progression (the time if took for a treated lesion to progress, the patient s symptom at progression, and what treatment was given). Detailed neurological examinations were performed immediately before treatment and at each followup visit by the treating neurosurgeon, neurooncologist, radiation oncologist, or neurologist. Long-term follow-up information was obtained in 2009 by contacting patients, families, and referring physicians, as well as from patient records of the last follow-up visit. When otherwise unavailable, dates of death were obtained from a social security death index. Radiosurgery Technique All but 3 patients, who were receiving in-patient chemotherapy, underwent radiosurgery on an outpatient basis. On the day of GKS, a local anesthetic agent was applied and a nonrelocatable Leksell stereotactic headframe was affixed to the head of the patient. Our routine MR imaging protocol was performed after injection of triple-dose Gd and acquisition of contiguous 1-mm magnetizationprepared rapid-gradient echo T1-weighted axial images. 11 One patient underwent thin-cut CT studies because there were MR imaging contraindications. All patients received 60 mg phenobarbital and 10 mg dexamethasone intravenously immediately before the procedure. Following the procedure, the patients received 3 additional doses of phenobarbital (1 dose every 12 hours over a 36-hour period) and completed a 7-day course of dexamethasone. Patients with no history of seizures received no further antiepileptic medications unless the treated lesions were in the temporal lobe or precental gyrus. When lesions were in such a location, patients were placed on a regimen of levetiracetam until complete resolution of perilesional edema. Patients with a history of seizures continued antiepileptic therapy indefinitely. Radiosurgery was performed using a Leksell Gamma Knife unit (Elekta AB). Every patient received a maxi- 54 J Neurosurg / Volume 113 / December 2010

3 Radiosurgery of brain metastases in eloquent cortex mum peripheral dose of 40 Gy and a dose of 20 Gy at the 50% isodose line. The 50% isodose line was contoured at the edge of the enhancing tumor and no margin was used. Follow-Up and Treatment at Progression Follow-up consisted of clinical examinations and serial MR imaging with triple-dose Gd 11 at 6 weeks and every 12 weeks thereafter. Local treatment failure for all lesions treated by GKS was defined as a persistent increase in the size of the contrast-enhancing lesion (> 20% in volume based on calculations involving the 3 perpendicular diameters of the lesion) and a concomitant T2 signal change. Distant treatment failure was defined as 1 or more new lesions appearing elsewhere in the brain. Neurotoxicity was graded according to the RTOG toxicity scale: Grade 1, mild neurological symptoms without need for medication; Grade 2, moderate neurological symptoms requiring an outpatient regimen of steroid medications; Grade 3, severe neurological symptoms involving outpatient or inpatient medications; Grade 4, life-threatening neurological symptoms (such as uncontrollable seizures, paralysis, or coma) and radionecrosis that has been histologically proven or is suspected based on clinical or imaging findings; and Grade 5, death. 38 In asymptomatic patients in whom surveillance MR imaging showed findings that raised the suspicion of local treatment failure, close observation with more frequent serial MR images and, at the clinician s discretion, a short course of steroid medication were prescribed. In patients in whom progression of disease was confirmed on imaging studies, retreatment of the lesion was undertaken by open resection, local field radiation therapy, WBRT, or, rarely, repeated GKS. When repeated GKS was performed, the radiation dose was most commonly 20 Gy. Exceptions to this included patients who received lower doses for retreatment of metastases that failed to respond to the initial GKS, patients who underwent interval treatment with WBRT, and patients whose tumor was > 2 cm in diameter. There were no limitations on the maximum number of metastases that were treated during each subsequent GKS session or the number of the repeated GKS sessions. Supratentorial lesions that received 20 Gy for initial treatment or subsequent treatment of new metastases are analyzed in this study. Cause of death was determined in all patients who did not survive. Similar to the criteria used by Patchell et al., 32 all patients with progressive neurological disease or dysfunction who died of intercurrent illnesses or advanced systemic cancer were considered to have died of a neurological cause. End Points and Statistical Analyses Primary end points included local treatment failure of GKS for metastases and neurological complications resulting from GKS. Overall survival and local treatment failure were measured from the time of the initial GKS. For each treated site of metastasis the respective GKS session was used as the starting point to determine J Neurosurg / Volume 113 / December 2010 a change in neurological status over time. Neurological symptom-free survival was defined as survival in the absence of any neurological deficits. Patients with fixed deficits prior to GKS were not considered symptom free. Failure to achieve freedom from neurological symptoms occurred when patients developed deficits from local treatment failure, distant treatment failure, or treatment complications. Two patients who died before the 6-week follow-up imaging session were excluded from the local control and neurological outcome analyses. The raw data were entered into an Excel (Office 2008 for Mac, Microsoft Corp.) spreadsheet. Averages are expressed as means ± SDs. Kaplan-Meier survival curves were generated for local disease control and neurological symptom-free survival. Univariate and multivariate logistic regression analyses were performed to determine the effect (expressed as ORs with 95% CIs) of the following variables in neurological disease progression: tumor volume (continuous), specifically a comparison of tumor volume 1 cm 3 and tumor volume < 1 cm 3 (dichotomous); pathological characteristics; number of isocenters used; comparison of homogeneous and heterogeneous enhancement patterns (dichotomous variable); heterogeneous/ cystic nature (none, < 50%, or 50% of tumor volume; trichotomous variable); and the presence of pretreatment neurological deficits referable to the lesion (excludes fixed deficits from previous treatments or preexisting conditions). Student t-tests and Mann-Whitney U-tests were used to analyze any differences between groups for continuous parametric and nonparametric variables, respectively. Fisher exact chi-square tests were used to compare proportions between groups. Analysis of variance was used to compare means of multiple groups. All statistics were calculated using statistical software (SPSS version 17.0 for Mac; SPSS, Inc.). A 2-tailed p value < 0.05 was considered statistically significant for all analyses. Results Patient Demographics, Prior Treatments, and Presenting Characteristics Table 1 summarizes the demographic and clinical data for all patients. Among the 98 patients treated with GKS, there were 65 women (66.3%) and 33 men (33.7%). The mean age of the patients at the time of GKS was 61.4 ± 12.8 years (range years). The most common primary tumors were non small cell lung cancer (49.0%), melanoma (20.4%), and breast cancer (18.4%). Primary site disease was controlled in 61 patients (73.5%), and 72 patients (73.5%) had extracerebral metastases present at the time of the initial GKS. The median KPS score was 90, and 77 patients (78.6%) were RPA Class II. Twenty patients harbored brain metastases that had been treated prior to the initial GKS at our center. Eighteen patients had undergone resection of a brain tumor, and 2 patients had undergone surgery twice for separate lesions. One patient had undergone GKS earlier for a single lesion, and another patient LINAC-based radiosurgery for 2 small lesions. Seventy-two patients (73.5%) were neurologically in- 55

4 R. E. Elliott et al. TABLE 1: Demographic and clinical characteristics of 98 patients with supratentorial brain metastases treated with 20-Gy GKS Characterisic No. of Patients (%)* sex male 33 (33.7) female 65 (66.3) age (in yrs) mean ± SD 61.4 ± 12.8 median 61.4 range primary tumor type non small cell lung cancer 48 (49.0) breast 18 (18.4) melanoma 20 (20.4) renal 7 (7.1) colon 1 (1.0) esophageal 1 (1.0) hepatocellular 1 (1.0) bladder 1 (1.0) unknown primary 1 (1.0) primary tumor controlled yes 61 (62.2) no 36 (36.7) unknown primary 1 (1.0) extracerebral metastases present yes 72 (73.5) no 26 (26.5) KPS score mean ± SD 84.0 ± 9.5 median 90 range RPA class I 15 (15.3) II 77 (78.6) III 6 (6.1) * Unless otherwise specified. tact, and 26 patients (26.5%) had deficits at the time of GKS. Nine patients (9.2%) had fixed deficits from prior surgical treatment of brain metastases (8 patients) or multiple sclerosis (1 patient). These fixed deficits included mild hemiparesis (3 patients), moderate hemiparesis (1 patient), and homonymous hemianopia (5 patients). Fourteen patients (14.3%) had neurological deficits from active supratentorial brain metastases that were treated using GKS. Metastasis-related deficits included mild hemiparesis (6 patients), moderate hemiparesis (3 patients), hemiparesis and aphasia (2 patients), hemihypesthesia (1 patient), quadrantanopia (2 patients), and homonymous hemianopia (1 patient). Two patients had deficits caused by infratentorial lesions (hemifacial numbness and ataxia, respectively). Location and Size of the Tumor and the Patient s Neurological Status At their initial presentation, 98 patients underwent 20-Gy GKS for a total of 131 metastases; later, 31 patients underwent repeated 20-Gy GKS for 76 metatases at distant sites. The mean tumor volume was 0.44 ± 0.62 cm 3 (range cm 3 ). Table 2 summarizes the location of all 207 supratentorial metastases treated with 20-Gy GKS. Fifty-six patients underwent GKS for a total of 96 metastases in noneloquent cortices (Grade I). The mean Grade I tumor volume was 0.44 ± 0.51 cm 3 (range cm 3 ). The most common Grade I locations were the frontal (48 patients [50%]), temporal (28 patients [29.2%]), and parietooccipital (13 patients [13.5%]) gyri. No patients had neurological deficits referable to tumors within noneloquent locations. Forty-two patients underwent GKS for a total of 51 metastases near eloquent cortices (Grade II). The mean Grade II tumor volume was 0.48 ± 0.68 cm 3 (range cm 3 ). The most common Grade II locations were the posterior parietal region (14 patients [27.5%]), the supplementary motor or premotor area (11 patients [21.6%]), and the occipital region within the optic radiations (9 patients [17.6%]). Two patients with metastases in near-eloquent areas had neurological deficits referable to these lesions, yielding baseline neurological deficit rates of 4.8% per patient and 3.9% per lesion within a neareloquent location. Forty-six patients underwent GKS for a total of 60 metastases in eloquent cortices (Grade III). The mean Grade III tumor volume was 0.51 ± 0.71 cm 3 (range cm 3 ). The most common Grade III locations were the precentral (21 patients [35%]) and postcentral (15 patients [25%]) gyri. Thirteen patients with metastases in eloquent areas had neurological deficits referable to these lesions, yielding baseline neurological deficit rates of 28.3% per patient and 21.7% per lesion within an eloquent location. There was no significant difference in tumor size between tumor groups stratified by location (p = 0.2). Lesions responsible for pretreatment neurological deficits were significantly larger than asymptomatic tumors (1.33 ± 1.2 cm 3 compared with 0.36 ± 0.51 cm 3 ; p = 0.001). Metastases in eloquent cortices were much more likely to cause neurological deficits than lesions in near- or noneloquent areas (p < 0.001). Five patients (5.1%) experienced seizures before GKS treatment. Three of these patients had perirolandic tumors and 2 had temporal lobe tumors. Overall Survival and Local Control The median duration of overall survival was 13.4 months (range 6 weeks 7.8 years). There were no 30-day procedure-related deaths. Survival rates were 54%, 46%, 27%, and 12% at 6, 12, 24, and 36 months, respectively. Twenty-four patients (24.5%) were still alive at a median follow-up duration of 30 months. Thirteen patients (13.3%) died of or with progressing neurological disease. Sixty-one patients (62.2%) died of systemic progression 56 J Neurosurg / Volume 113 / December 2010

5 Radiosurgery of brain metastases in eloquent cortex TABLE 2: Locations of 207 supratentorial brain metastases treated with 20-Gy GKS Location No. of Metastases (%)* Grade I (noneloquent) total 96 (46.4) frontal 48 (50) temporal 28 (29.2) parietooccipital 13 (13.5) lateral ventricle 5 (5.2) cingulate 2 (2.1) Grade II (near eloquent) total 51 (24.6) premotor (abutting precentral gryus) 4 (7.8) posterior parietal 14 (27.5) supplementary motor area 7 (13.7) near Broca area 3 (5.9) near Wernicke area 6 (11.8) optic radiations (lateral to calcarine cortex) 9 (17.6) insular 5 (9.8) corpus callosum 1 (2.0) superior/inferior parietal lobule 2 (3.9) Grade III (eloquent) total 60 (29.0) precentral gyrus 21 (35.0) postcentral gyrus 15 (25.0) carlcarine 14 (23.3) Wernicke area 3 (5.0) Broca area 1 (1.7) centrum semiovale 4 (6.7) basal ganglia/thalamus 2 (3.3) * Total percentages shown for grade are percentages of all metastases; percentages shown for metastases in individual locations are percentages within the grade. or intercurrent illness without progression of brain disease at the time of death. There was no difference in overall survival between patients with and without baseline neurological deficits (p = 0.42, log-rank chi-square test). Local treatment failure occurred in 12 metastases (5.8%) treated with 20-Gy GKS at a median time of 9.8 months (range 3 26 months), yielding a crude local control rate of 94.2%. Local control rates were 93%, 92%, 90%, and 90% at 6, 12, 24, and 36 months, respectively. There were no significant differences in local control between Grades I, II, and III metastases or between tumors producing neurological symptoms and asymptomatic tumors (p = 0.53 and p = 0.11, respectively; log-rank chi-square test). Fourteen patients later required WBRT and 2 others died before WBRT could be done, yielding a 16.3% rate of patients who required WBRT. Indications for WBRT in these patients included widespread intracerebral disease (10 patients), numerous metastases and tumor recurrence at the site of previous resection (1 patient), leptomeningeal disease (2 patients), and tumor recurrence at the site of previous resection (3 patients). J Neurosurg / Volume 113 / December 2010 Neurological Outcomes Table 3 summarizes outcomes in 15 patients with baseline neurological deficits caused by brain metastases, and 1 patient who had no deficit prior to GKS but developed a deficit afterward. The pre-gks deficits resolved completely in 4 patients, improved in 6 patients, remained stable in 3 patients, and worsened in 2 patients after the procedure. Improvement or resolution of deficits occurred between 6 weeks and 6 months (median 2.8 months). Among the 13 patients whose deficits were caused by Grade III lesions, 4 (30.8%) had complete resolution (Fig. 1), 5 (38.5%) had improvement, 3 (23.1%) remained stable, and 1 (7.7%) experienced permanent worsening of symptoms. In the last patient, imaging studies raised the suspicion of radiation necrosis at 3 months. Despite treatment with dexamethasone, she was left with a moderate hemiparesis and died 6 weeks later of systemic progression. One patient with Grade III metastasis (1 [2.2%] of 46 patients) who had been neurologically intact developed a new mild hemiparesis from perilesional edema 6 weeks after she underwent GKS for a lesion in the postcentral gryus. Following treatment with dexamethasone, she continued to have trace difficulties with fine motor function and hand coordination. Three patients experienced neurological deficits due to edema, which developed between 6 weeks and 3 months following GKS; the deficits responded to a short course of steroid treatment. One patient experienced transient expressive aphasia at 6 weeks post-gks, which resolved by 3 months. Two patients with mild preexisting hemiparesis experienced transient worsening of their deficits, which resolved by 6 weeks in 1 patient and by 3 months in the other patient. Both patients with deficits from Grade II lesions experienced worsening of their deficits accompanied by local treatment failure. One patient had quadrantanopia from a lesion lateral to the calcarine cortex and experienced local treatment failure at 6 months. She was left with a permanent homonymous hemianopia following open resection. Another patient had receptive aphasia from a tumor near the Wernicke area. Following GKS, his aphasia improved but transiently worsened at 6.8 months due to local treatment failure. The patient s aphasia improved once his disease had been controlled by WBRT. Overall, 1 patient (50%) with a symptomatic near-eloquent lesion experienced improvement and 1 patient (50%) experienced deterioration in neurological deficits. No patients with an asymptomatic Grade I or II lesion experienced neurological deterioration following GKS. Two patients with frontal lobe metastases experienced headache due to radiation necrosis that required resection. Another patient with melanoma experienced intracerebral hemorrhage, which required evacuation. The neurological symptom-free survival rates were 80%, 83%, 85%, and 76% at 6, 12, 24, and 36 months, respectively. Table 4 summarizes the neurological outcomes by location. Overall, the rates of permanent neurological deterioration following GKS of Grades I, II, and III lesions were 0% (0 of 96 tumors), 2% (1 of 51), and 3.3% (2 of 60), respectively. Transient deficits occurred in 0% (0 of 96 tumors), 2% (1 of 51), and 5% (3 of 60), respectively. A significantly greater proportion of neurological deterio- 57

6 R. E. Elliott et al. TABLE 3: Evolution of neurological deficits from brain metastases treated with 20-Gy GKS* Case No. Tumor Location Tumor Volume (cm 3 ) Pre-GKS Deficit Post-GKS Deficit Time to Improvement or Resolution (mos) Time to New/ Worse Deficit (mos) Residual Deficit 1 postcentral gyrus 0.99 hemihypesthesia resolved 1.5 NA none 2 precentral gyrus 1.9 mild hemiparesis improved 1.5 NA trace hemiparesis 3 precentral gyrus 1.40 mild hemiparesis worse NA 3 moderate hemiparesis 4 postcentral gyrus 0.12 mild hemiparesis stable NA transient worsen- mild hemiparesis ing at 6 wks 5 precentral gyrus 1.1 mild hemiparesis improved 1.5 NA trace hemiparesis 6 precentral gyrus 0.85 mild hemiparesis stable NA transient worsen- mild hemiparesis ing at 3 mos 7 precentral gyrus 0.13 mild hemiparesis improved 3 NA trace hemiparesis 8 precentral gyrus 0.52 moderate hemiparesis resolved 1.5 NA none 9 precentral gyrus 1.6 moderate hemiparesis improved 3 NA mild hemiparesis 10 postcentral gyrus 3.8 moderate hemiparesis resolved 3 NA none 11 lt superior temporal gyrus 0.45 aphasia (receptive) improved (LTF, tran- mild aphasia (near Wernicke area) sient) 12 lt frontal operculum (Broca area) 2.0 aphasia (expressive), mild hemiparesis improved 3 transient worsening at 6 wks mild aphasia, trace hemiparesis 13 calcarine cortex 0.98 homonymous hemianopia stable NA NA homonymous hemianopia 14 optic radiations (lateral to calcarine cortex) 3.2 quadrantanopia worse (LTF) NA 6 (LTF) homonymous hemianopia 15 cuneus (calcarine cortex) 2.4 quadrantanopia resolved 1.5 NA none 16 postcentral gyrus 1.7 none worse NA 1.5 trace hemiparesis * LTF = local treatment failure; NA = not applicable. ration was associated with Grade III metastases (5 of 60 tumors) than with Grade I or II metastases (2 of 147; p = 0.02, Fisher exact test). Two patients (2.0%, one with a tumor in the postcentral gyrus and the other with a tumor in the temporal region) developed new seizures following GKS and 3 patients with pre-gks seizures had an increase in seizure frequency following treatment. All seizures were successfully controlled using antiepileptic medications. Grade 1 toxicities included headache in 9 patients (9.2%), nausea in 6 patients (6.1%), and dizziness in 4 patients (4.1%). All resolved in a matter of days without further sequelae. Risk Factors for Neurological Deterioration On univariate testing, the following variables had a significant impact on the risk of new or worsened neurological deficits: baseline neurological deficit referable to treated lesion (OR = 47.5; p < 0.001), eloquent location (OR = 6.59; p = 0.027), increasing tumor size (OR = 2.56; p = 0.01), tumor volume 1 cm 3 (OR = 5.77; p = 0.028), and increasing number of isocenters used (OR = 1.82; p = 0.001). On multivariate analysis, only the presence of a preexisting neurological deficit was associated with neurological deterioration. However, given the small number of transient and permanent posttreatment deficits (7 in all), the multivariate analysis was underpowered and the model itself was not significant. Discussion In this consecutive series of nearly 100 patients with supratentorial brain metastases treated with uniform radiosurgical dosing, we have demonstrated that 20-Gy GKS provides excellent local control with low morbidity when used to treat metastases 2 cm in diameter. Gamma Knife surgery also provided good palliation of presenting symptoms, given that 13 (87%) of 15 patients with deficits improved or remained clinically stable. Patients with larger lesions and those with lesions within eloquent cortices tended to have higher rates of neurological deficits before GKS and a greater risk for posttreatment deterioration in their conditions. Moreover, patients with pre-gks deficits were at a markedly higher risk for deterioration following treatment. Such findings may allow for more accurate counseling of patients and their referring physicians prior to SRS treatment for brain metastases. Furthermore, the uniform dose of 20 Gy, its documented efficacy, and the low risk of toxicity may help establish a new standard for the treatment of brain metastases 2 cm. Stereotactic Radiosurgery for Brain Metastases Stereotactic radiosurgery offers a minimally invasive technique that provides excellent rates of local control of metastases, has limited morbidity, and results in decreased length of hospital stay and lower costs. 27,36 Simi- 58 J Neurosurg / Volume 113 / December 2010

7 Radiosurgery of brain metastases in eloquent cortex Fig. 1. Magnetic resonance images obtained in the patient in Case 10 (see Table 3). Five years after he had been diagnosed and treated for non small cell lung cancer, this 72-year-old man developed a progressive, moderate right hemiparesis with minimal joint-position sense dysfunction. Magnetic resonance images revealed a 1.9-cm-diameter metastasis in the left postcentral gyrus with surrounding edema (A and B). He underwent uncomplicated GKS with 20 Gy directed to the 3.8-cm3 lesion. Despite worse edema and a slight increase in the lesion size due to central necrosis, the patient s hemiparesis improved by 3 months (C and D). He continued to improve neurologically at the 9-month follow-up visit with concomitant decreases in lesion size and edema (E and F). The patient was neurologically intact by 12 months post-gks, and MR images showed only a faint residual enhancement without edema (G and H). The remaining enhancement on MR images resolved by 24 months (I and J). The patient remained free from intracerebral disease until he died of systemic cancer progression 30 months post-gks. lar to published results of surgical resection of solitary metastases, SRS improves local control, overall survival, and functional survival when combined with WBRT alone.2,22,28 While functional survival is often measured as a function of improvement or stabilization of the KPS score, few studies have systematically analyzed change in neurological function after SRS for brain metastases. Arguably the most critical component of functional status in patients with intracerebral disease is neurological status. The evolution of neurological symptoms and complications over time following SRS for brain metastases, however, may be underreported. Suki and colleagues47 reviewed studies published between 1995 and 2005 on the radiosurgical treatment of brain metastases and found that the reporting of complications in the literature was inadequate. They cited small study populations, failure to stratify lesions by their relationship to eloquent brain regions, heterogeneous pathological conditions (including J Neurosurg / Volume 113 / December 2010 benign lesions, such as meningiomas, malignant gliomas, and vascular malformations), and limited follow-up duration. Table 5 summarizes some of the major published series of SRS for brain metastases and the incidence of radiation necrosis and neurological complications.1,3,4,7,12, 15,20,24,34,35,39,41,42,51,52 Rates of radiation necrosis and neurological deterioration following SRS reported in selected studies range from 0% to 7% and 0% to 17%, respectively. Reported risk factors for these complications include tumor volume, prior WBRT, and increasing radiation dose.17,23,26,31,39,40 Unfortunately, radiosurgical dosing varies widely in the current literature, precluding the establishment of predictive dose-response relationships and standardized probabilities of normal tissue toxicity. In one of the largest studies in which complications were specifically analyzed, Williams and colleagues52 pro spectively reviewed their experience with 273 patients 59

8 TABLE 4: Summary of the evolution of neurological deficits before and after 20-Gy GKS for supratentorial brain metastases Location No. of Tumors Pre-GKS Deficits, No. of Tumors (%)* Post-GKS Changes in Baseline Deficits, No. of Tumors (%) New Deficits, Resolved Improved Stable Worse No. of Tumors (%)* R. E. Elliott et al. Transient Deficits, No. of Tumors (%)* Permanent Deficits, No. of Tumors (%)* Grade I 96 0 NA NA NA NA Grade II 51 2 (3.9) 0 1 (50) 0 1 (50) 0 1 (2) 1 (2) Grade III (21.7) 4 (30.8) 5 (38.5) 3 (23.1) 1 (7.7) 1 (1.7) 3 (5) 2 (3.3) * Percentage of tumors in specific grade. Percentage of tumors related to changes in baseline deficits. New deficits that appeared when the patient s baseline neurological status had been intact. harboring 1 or 2 brain metastases who underwent SRS (median dose 18 Gy). Williams et al. noted new neurological complications in 32% of the 316 lesions treated. Neurological deficits occurred in 17% and new-onset seizures occurred in 13% of the treated lesions. Risk factors for neurological complications included location in eloquent cortex, preexisting neurological deficits, and progression of primary disease at the time of SRS. It is noteworthy, however, that the median tumor size was larger than ours (1.26 cm 3 ), and 29% of the lesions were 3 cm. Such a difference in tumor size and the investigators rigorous, prospective manner of data collection may account for the higher rates of neurological complications than those found in our series. Shehata et al. 40 analyzed the relationship between radiosurgical dose and complications in a retrospective study of 160 patients with 468 metastases 2 cm in diameter that were treated with SRS with or without WBRT. They concluded that 20-Gy SRS combined with WBRT is the optimal regimen for the radiosurgical treatment of brain metastases. Doses in excess of 20 Gy caused a marginally significant trend toward increased Grades 3 and 4 neurotoxicity (5.9% compared with 1.9% for doses 20 Gy) without an improvement in local control. Similarly, Majhail et al. 26 reported on their experience using GKS for various intracerebral pathological conditions (vascular malformations and benign and metastatic brain tumors) in patients who had not undergone previous radiation treatment. Majhail et al. found a lesion size > 2.5 cm and a dose > 20 Gy to be associated with increased complications. We have used 20 Gy for metastases 2 cm for almost TABLE 5: Reported rates of radiation necrosis and neurological complications in cases of brain metastasis treated with radiosurgery* Authors & Year Design No. of Patients Median or Mean Dose (Gy) % Radiation Necrosis per Patient % Radiation Necro sis per Treated Lesion % Seizures per Patient % Permanent Neurological Deficits per Patient Aoyama et al., 2006 PRCT or NA 3.8 NA Chitapanarux et al., 2003 PO NA NA Lutterbach et al., 2003 PO NA 4 Williams et al., 2009 PO NA Kihlström et al., 1993 R NA NA Flickinger et al., 1994 R NA 3.4 Alexander et al., 1995 R NA Shu et al., 1996 R NA NA 3.4 Breneman et al., 1997 R Shiau et al., 1997 R NA NA 5.9 Pirzkall et al., 1998 R or NA NA Petrovich et al., 2002 R NA 2.8 NA Sheehan et al., 2002 R NA NA Hasegawa et al., 2003 R NA NA 2.3 Varlotto et al., 2003 R NA NA 1.5 current study R * PO = prospective observational; PRCT = prospective randomized controlled trial; R = retrospective. Patients with and without a history of prior WBRT received a median dose of 16.9 and 20 Gy, respectively. Patients with and without a history of prior WBRT received a median dose of 15 and 20 Gy, respectively. Three of 5 patients with seizures had seizures prior to GKS. 60 J Neurosurg / Volume 113 / December 2010

9 Radiosurgery of brain metastases in eloquent cortex 10 years and believe it offers a high rate of sustained local control (nearly 90% or greater, even 3 years following treatment) with a very low toxicity profile. On univariate analysis, the presence of a preexisting neurological deficit was the most prominent risk factor for posttreatment deterioration. Increasing tumor size, location in eloquent brain regions, and number of isocenters were also significant but less prominent factors. However, given the low number of patients who experienced neurological deterioration and the high correlation of these variables, we have to point out that the multivariate model was underpowered and not statistically significant. Larger studies with uniform dosing are needed to determine tumor characteristics and technique-related risk factors that may impact normal tissue toxicity and resultant neurological worsening. In contrast to studies reporting worse overall survival for patients with gliomas, 5,6 surgically resected metastases 48,53 in eloquent locations, or lesions causing neurological deficits, we found that the presence of neurological deficits from brain metastases did not correlate with decreased survival. The impact of these deficits on the patients quality of life, however, was not assessed in our study. Seizures have been reported to occur in 3% 29% of patients following SRS for brain metastases. 1,3,4,7,34,52 We agree with investigators at other centers 13,52 about the use of anticonvulsant medications in high-risk populations. We have noted seizures in patients with perirolandic and temporal metastases and, therefore, keep these patients on anticonvulsant medications (regardless of seizure history) until the surrounding edema disappears. Larger, prospective studies are needed to confirm the efficacy and resource utilization of our protocol. In parallel with the evolution of the treatment of cerebral metastases, the goals of therapy have changed over the past few decades. No longer is the goal of treatment simply palliation to relieve suffering for a few months prior to patients succumbing to their diseases. The goals are now to control intracerebral tumor growth indefinitely, prolong survival in selected cases, and improve or maintain neurological and functional status of patients with brain metastases. To best achieve that end and to determine the optimal course of treatment for a given patient, consistent recording and reporting of complications are required and, ideally, randomized controlled trials to help determine the ideal dosage for tumors of various sizes and functional locations. Resection of Brain Metastases Although numerous centers have attempted comparisons of resection and radiosurgery for brain metastases, a direct comparison of these modalities is inherently fraught with selection bias. Compared with lesions selected for SRS, tumors that are resected tend to be larger, causing more mass effect or elevated intracranial pressure. 30,46,49 This makes a meaningful and direct comparison of these treatment modalities difficult, if not impossible. Using current stereotactic techniques, the perioperative mortality rate is < 2%, the complete resection rate is 90%, the neurological deterioration rate is between 0% and J Neurosurg / Volume 113 / December %, and the rate of local recurrence ranges from 0% to 16%. 19,30,37,48 50 Few studies in the surgical literature have addressed neurological complications stratified by proximity to eloquent cortex. Sawaya and colleagues 37 studied neurological outcomes in 327 patients who underwent a total of 400 craniotomies for intraparenchymal tumors (194 metastases and 206 gliomas). These investigators found increasing risks of incomplete resection, neurological deterioration, and overall complications in patients with Grade III tumors. Similarly, Paek et al. 30 analyzed complications as they relate to the functional grade of the tumor s location in 208 patients who underwent surgery to remove 1 brain metastases. They reported fewer complete resections, higher rates of neurological complications, longer hospitalizations, higher rates of KPS score deterioration, and shorter overall survival in patients with Grade III metastases. Tan and Black 49 reported on their experience with stereotactic resection of 55 brain metastases in 49 patients. There was no operative death and gross-total resection was achieved in 96% of cases. No patient who was neurologically intact prior to surgery exhibited a new postoperative deficit. Only 2 (3.6%) of the patients with a preoperative deficit experienced worsening of the deficit after surgery, and 70% of patients with deficits had improvement or complete resolution of their deficits. Tobler and Stanley 50 reported on 14 patients with brain metastases within eloquent cortices who underwent stereotactic resection. These authors reported no new neurological deterioration as a result of surgery and improvement or resolution of presurgical deficits in 11 (85%) of 13 patients. Although Tobler and Stanley contend that their results demonstrate that stereotactic resection should be the primary mode of treatment for eloquent brain metastases, their patient population is too small to support that conclusion. An often overlooked aspect in the comparison of surgery and SRS for brain metastases is the need for postoperative WBRT. Randomized controlled studies have demonstrated that adjuvant WBRT following resection of brain metastases results in improved local control, overall survival, and functional status. 28,32 Thus, it has become the standard of care, consistent with the rate of WBRT use in 90% 100% of patients in the surgical literature. However, as the deleterious effects of WBRT have become increasingly recognized, many centers are attempting to limit its use in the initial treatment of patients with brain metastases. Stereotactic radiosurgery has been vital to achieving that end. It avoids most of the side effects of WBRT, including alopecia, skin changes, fatigue, hypopituitarism, cognitive deterioration, ataxia, leukoencephalopathy, and interruption of systemic chemotherapy. 8 10,16,21,44 Unlike WBRT, SRS can be repeated should new lesions arise, which occurs in nearly half of patients. 18 In most studies in which SRS alone was the initial treatment of patients with limited numbers of brain metastases, WBRT was avoided in 70% 90% of patients. 7,14,24,34 Only 16.3% of patients in our series required WBRT, sparing almost 85% of patients potential adverse physical and neurocognitive effects. 61

10 R. E. Elliott et al. A further sequela of surgical removal of brain metastases is the risk of leptomeningeal dissemination. Numerous centers have demonstrated a higher incidence of leptomeningeal dissemination in patients who underwent resection of metastases than in those who underwent SRS. 25,43,45,46 Although this difference is more pronounced in patients harboring posterior fossa metastases, it has also been shown to be significant in the setting of supratentorial lesions. 46 Some preliminary data indicates en bloc resection (in which the dissection remains peripheral to and does not violate the internal capsule) may be superior to piecemeal resection and confers a risk of leptomeningeal disease not dissimilar to SRS. 33,46 Given the extremely poor prognosis for patients with leptomeningeal disease (median survival 4 months) and the questionable efficacy of WBRT for this condition, surgical and radiosurgical measures to limit its occurrence should be readily used. We believe there is a distinct, but at times overlapping, role for resection and SRS for brain metastases in all locations. Similar to other centers, 49 we use surgery for larger tumors (> 3 cm); tumors causing significant mass effect, edema, or hydrocephalus; cases of unknown primary cancer; and patients with severe neurological deficits that may be improved with immediate excision. Smaller lesions, those deep in the parenchyma, and lesions in patients with significant medical comorbidities may be optimally treated with SRS. There is overlap, however, for lesions of intermediate size in eloquent regions. Both SRS and resection have been shown in predominantly retrospective studies to offer excellent rates of local control and a relatively low risk of neurological deterioration. The need for WBRT and the risk of leptomeningeal disease must be considered in this comparative analysis. Controlled studies with adequate follow-up and resource utilization analyses are needed to determine the optimal treatment of such lesions. Study Limitations Although this study is retrospective in nature, it provides data on a relatively large series of consecutive patients treated with an identical radiosurgical protocol using a fixed radiation dose. Unlike most other series in which radiosurgical dosing varies widely, all patients received 20 Gy for metastases of all primary tumor types. These data, which are based on a uniform patient selection and identical treatment, should be of help to others in choosing SRS for their patients. These data may help establish dosing standards, although the possibility of using lower doses for 2-cm lesions to further decrease neurological risk should be considered and evaluated carefully. Our data, and those of others 30,37,52 describing neurological status as a function of tumor location, may allow better counseling to patients about their expected recovery or risk of neurological deterioration, and may help determine the optimal treatment strategy for a given patient. Conclusions Gamma Knife surgery with 20 Gy provides palliation of neurological deficits from brain metastases 2 cm that are located in or near eloquent cortices in most patients, and it does so with low rates of morbidity. Consistent with findings reported in the surgical literature, higher rates of neurological complications were observed as proximity to eloquent regions and tumor size increased. Patients with pretreatment deficits should be counseled about their moderately increased risk of temporary or permanent neurological worsening. Since larger lesions were more often associated with symptoms, early and reasonably frequent posttreatment MR imaging should be performed to detect new metastases that are still small and asymptomatic. Such a protocol may result in fewer patients with neurological deficits and reduce the use of WBRT. 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: Elliott, Golfinos. Acquisition of data: Elliott, Morsi, Mehta, Spriet. Analysis and in terpretation of data: all authors. Drafting the article: Elliott, Rush, Morsi, Narayana. Critically revising the article: Elliott, Rush, Nara yana, Donahue, Parker, Golfinos. Reviewed final version of the manuscript and approved it for submission: Elliott, Rush, Donahue, Parker, Golfinos. Statistical analysis: Elliott. Administrative/technical/ material support: Rush, Golfinos. Study supervision: Rush, Narayana, Donahue, Parker, Golfinos. References 1. Alexander E III, Moriarty TM, Davis RB, Wen PY, Fine HA, Black PM, et al: Stereotactic radiosurgery for the definitive, noninvasive treatment of brain metastases. J Natl Cancer In st 87:34 40, Andrews DW, Scott CB, Sperduto PW, Flanders AE, Gaspar LE, Schell MC, et al: Whole brain radiation therapy with or without stereotactic radiosurgery boost for patients with one to three brain metastases: phase III results of the RTOG 9508 randomised trial. Lancet 363: , Aoyama H, Shirato H, Tago M, Nakagawa K, Toyoda T, Hatano K, et al: Stereotactic radiosurgery plus whole-brain radiation therapy vs stereotactic radiosurgery alone for treatment of brain metastases: a randomized controlled trial. JAMA 295: , Breneman JC, Warnick RE, Albright RE Jr, Kukiatinant N, Shaw J, Armin D, et al: Stereotactic radiosurgery for the treatment of brain metastases. Results of a single institution series. Cancer 79: , Chaichana K, Parker S, Olivi A, Quiñones-Hinojosa A: A pro posed classification system that projects outcomes based on preoperative variables for adult patients with glioblastoma multiforme. Clinical article. J Neurosurg 112: , Chang EF, Smith JS, Chang SM, Lamborn KR, Prados MD, Butowski N, et al: Preoperative prognostic classification system for hemispheric low-grade gliomas in adults. Clinical article. J Neurosurg 109: , Chitapanarux I, Goss B, Vongtama R, Frighetto L, De Salles A, Selch M, et al: Prospective study of stereotactic radiosurgery without whole brain radiotherapy in patients with four or less brain metastases: incidence of intracranial progression and salvage radiotherapy. J Neurooncol 61: , Chow E, Davis L, Holden L, Tsao M, Danjoux C: Prospective assessment of patient-rated symptoms following whole brain 62 J Neurosurg / Volume 113 / December 2010