paediatrics Brussels 17

Pediatr Blood Cancer 2009;52:65–69

Both Location and Age Predict Survival in Ependymoma: A SEER Study

Courtney S. McGuire, MD , 1 Kristin L. Sainani, PhD , 2 and Paul Graham Fisher, MD 1,3,4,5 *

Background. Studies have suggested that supratentorial ependy- momas have better survival than infratentorial tumors, with spinal tumors having the best prognosis, but these data have been based on small samples. Using a population-based registry of ependymomas, we analyzed how age, gender, location, race and radiotherapy influence survival in children. Methods. We queried the Surveillance Epidemiology End Results database (SEER-17) from 1973 to 2003, strictly defining ependymomas by histology. Site codes were used to distinguish between supratentorial, infratentorial, and spinal tumors when available. Outcomes were compared by location, age, gender, race and radiotherapy, using Kaplan–Meier analysis and logrank tests. Cox regression was completed, incorporating all significant covariates from univariate analysis. Results. Six hundred thirty-five children were identified with an overall 5-year survival of 57.1 standard error (SE) 2.3%. Increasing age was associated with

plastic ependymoma, papillary ependymoma and myxopapillary ependymoma). The SEER database captures institutional diagnoses and does not centrally review pathology. Primary tumor locations were distinguished by ICD-0-2 site codes and defined as supraten- torial (700, 702–714), infratentorial (716-717), or spinal (720-721, 701, 725). Those tumors identified as ventricle, overlapping brain, brain not otherwise specified, overlapping or not otherwise specified (715, 718, 719, 728, and 729) were excluded from location analysis because of the inability to assign the tumor to one of the three strata. Survival was defined as the time fromdiagnosis until death due to all causes. Overall survival was calculated by Kaplan–Meier analysis. Treatments coded as beam radiation, radioactive, and radiation were classified as ‘‘radiotherapy’’ while those labeled as none or refused were classified as ‘‘no radiotherapy.’’ Outcomes were compared by age, location, gender, race (blacks and whites), and radiotherapy using univariate logrank tests. Cox proportional hazards multivariate regression was subse- quently used to incorporate all significant covariates from univariate analysis (i.e., location, age, radiotherapy). A pre-planned multi- variate analysis was completed for the subgroup of infratentorial tumors incorporating the significant univariate covariates age and improved survival ( P < 0.0001). Five-year survival by location was 59.5 SE 5.5% supratentorial, 57.1 SE 4.1% infratentorial and 86.7 SE 5.2% spinal. Radiotherapy of the infratentorial tumors resulted in significantly improved survival in both univariate analysis (logrank P < 0.018) and multivariate analysis restricted to this tumor location ( P ¼ 0.033). Using multivariate analysis that incorporated all tumor locations, age ( P < 0.001) and location ( P ¼ 0.020) were significant predictors for survival. Conclusions. Age and location independently influence survival in ependymoma. Spinal tumors are associated with a significantly better prognosis than both supra- tentorial and infratentorial tumors, and may represent a distinct biological entity. Radiotherapy appears beneficial for survival in patients with infratentorial ependymoma. Pediatr Blood Cancer 2009;52:65–69. 2008 Wiley-Liss, Inc. —————— 1 Department of Neurology, Stanford University, Palo Alto, California; 2 Department of Health Research and Policy, Stanford University, Palo Alto, California; 3 Department of Pediatrics, Stanford University, Palo Alto, California; 4 Department of Neurosurgery, Stanford University, Palo Alto, California; 5 Department of Human Biology, Stanford University, Palo Alto, California This work was presented in part at the 2007 Annual Meeting of the American Society of Clinical Oncology, Chicago, Illinois, June 6, 2007. *Correspondence to: Paul Graham Fisher, Room CC22200, Stanford Cancer Center, 875 Blake Wilbur Drive, Palo Alto, CA 94305-5826. E-mail: pfisher@stanford.edu Received 29 February 2008; Accepted 9 September 2008

Key words: brain tumor; ependymomas; epidemiology; pediatric oncology; survival

INTRODUCTION

Brain tumors currently occur at an annual rate reported as high as 4.3 per 100,000 person-years in children, with ependymomas comprising 8–10% of these neoplasms [1–7]. Ependymoma continues to be associated with significant mortality with 5-year overall survival reported about 65% [6,8]. Opinions vary regarding which factors influence outcome. Younger children are generally thought to have a poorer prognosis [9,10], and survival rates appear to be lower in children than adults [6]. Although radiotherapy is the standard of care for adults, current treatment recommendations for children vary and are often limited by the significant neurotoxicity associated with radiation [11]. Survival has also been reported to vary by tumor location: supratentorial tumors lead to higher survival rates than infratentorial tumors, and spinal tumors have the best prognosis [1– 3,12]. Unfortunately, many studies reporting these survival trends are based on small samples and single-institution experiences. Thus, there is a need for more complete and definitive analysis based upon a larger sample in a population-based cancer registry. This study sought to analyze how age, gender, location, race, and radiotherapy influence survival in ependymoma, using such a comprehensive database. The Surveillance Epidemiology End Results Program (SEER-17) was used to identify all ependymoma cases diagnosed at age 18 or younger in 17 United States cancer registries from 1973 to 2003 [13]. The National Cancer Institute’s SEER database is an authoritative source of United States population-based data, incorporating both historical and current cases, and now draws a representative 26% of the population [13]. Approval for research was granted by the Stanford Panel on Human Subjects in Medical Research. Cases were selected from the SEER-17 if there was a diagnosis of ependymoma defined by the International Classification of Disease (ICD-0-3) histology codes 9391, 9392, 9393, and 9394 (ependymoma, ana- METHODS

2008 Wiley-Liss, Inc. DOI 10.1002/pbc.21806 Published online in Wiley InterScience (www.interscience.wiley.com)