FULL-LENGTH ORIGINAL RESEARCH
Incidence, risk factors, and longitudinal outcome of seizures in long-term survivors of pediatric brain tumors *†‡Nicole J. Ullrich, *†‡Scott L. Pomeroy, §Kush Kapur, †‡Peter E. Manley, ‡¶Liliana C. Goumnerova, and *Tobias Loddenkemper Epilepsia, 56(10):1599–1604, 2015 doi: 10.1111/epi.13112
SUMMARY
Nicole J. Ullrich is director of neurologic neurooncology at Dana-Farber/Boston Children’s Cancer and Blood Disorders Center.
Objective: Seizures are common during and after treatment for a primary brain tumor. Our objective was to describe the incidence and risk factors for seizures in long-term survivors of pediatric brain tumors. Methods: In a retrospective, longitudinal study, we reviewed all consecutive patients during a 12-month period who were at least 2 years post initial diagnosis of a brain tumor. Data collection included age at diagnosis, length of follow-up, extent of initial resection, tumor histology, and treatment modalities. For patients who had experienced seizures at any time, the timing and frequency of seizures, seizure semiology, electroencephalography results, and anticonvulsant use were recorded. Univariate analyses and logistic regression were performed to assess risk factors. Results: The cohort included 298 patients (140 female). Average duration of follow-up was 7.6 years. Initial surgical resection was gross-total in 109 patients, and subtotal for 143. Twenty-nine patients underwent biopsy alone and 17 had no surgical intervention. Tumor location included posterior fossa (104; 36%), midline (98; 34%), cortical (85; 29%), and other (11; 3%). Most frequent diagnoses were low grade glioma, medulloblastoma, and ependymoma. Other treatments included cranial irradiation (N = 163) and chemotherapy (n = 127). Tumor recurrence occurred in 92 patients (30%). Seventy-one patients had seizures (24%). Ongoing seizures at the time of most recent follow-up were present in 42 patients. Risk factors for seizures included tumor location, tumor histology, tumor recurrence, and incomplete resection at time of initial presentation. Significance: Seizures are a frequent comorbidity in pediatric brain tumor survivors, seen at presentation in 24% of patients and ongoing in 14%. Factors predisposing to seizures include tumor pathology (low/high grade glioma, glioneuronal tumor), cortical location, and subtotal resection. These data may assist in identification and management of patients at highest risk for seizures as well as identification of patients for potential treatment trials with antiepileptogenic agents. KEY WORDS: Seizure, Brain tumor, Pediatric, Epilepsy, Oncology.
Accepted July 9, 2015; Early View publication August 31, 2015. Departments of *Neurology and †Hematology/Oncology and Neurosurgery, Boston Children’s Hospital, Boston, Massachusetts, U.S.A.; ‡Dana-Farber/Boston Children’s Cancer and Blood Disorders Center, Boston, Massachusetts, U.S.A.; and §Clinical Research Program, Boston Children’s Hospital, Boston, Massachusetts, U.S.A. Address correspondence to Nicole J. Ullrich, Department of Neurology, Boston Children’s Hospital, 300 Longwood Ave, Boston, MA 02115, U.S.A. E-mail:
[email protected] Wiley Periodicals, Inc. © 2015 International League Against Epilepsy
Seizures are the second most common presenting symptom of supratentorial tumors in childhood; only headaches are more frequent.1 Seizures are also frequent during and after treatment for a primary brain tumor. Overall, the incidence of seizures in patients with brain tumors is nearly 30%.2 Although complete surgical resection is thought to be a predictor of good seizure outcome,3 the risk factors that predispose patients to develop seizures and that define the subpopulation of patients who experience adverse seizure outcome are poorly defined.
1599
1600 N. J. Ullrich et al.
Key Points • • • • •
Retrospective, longitudinal study of 298 pediatric patients at least 2 years post diagnosis of brain tumor, seen during 12-month period Incidence of seizures at presentation was 24%; 14% of cohort had ongoing seizures at most recent follow-up Factors predisposing for seizures included tumor pathology (low/high grade glioma, glioneuronal tumor), cortical tumor location, and subtotal resection Presence of residual tumor and increased time since diagnosis predicted poor seizure control Data may help identify both patients at highest risk for seizures and those that can be tapered more quickly from anticonvulsant treatment
Few studies have evaluated tumor-related seizures in children. It is estimated that 1–3% of children with new onset of afebrile seizures are found to have a brain tumor.4 By corollary, seizures are the presenting sign of brain tumors in children in approximately 10% of cases.5 As long-term survival continues to improve, a better understanding of the natural history and risk factors for poor seizure control in children with brain tumors is crucial. The objectives of our study were to examine the rates of seizures in a large population of long-term survivors of pediatric brain tumors and to identify predictors of adverse seizure outcomes.
Methods Patient population and data collection We performed a retrospective, longitudinal study on all patients presenting for follow-up evaluation to the Pediatric Brain Tumor Program Outcomes Program at Dana-Farber Cancer Institute during a consecutive 12-month period who were at least 2 years after initial diagnosis of a CNS tumor. Our institution maintains a high follow-up rate of survivors of primary brain tumors, each of whom is followed for at least 10 years after completion of treatment. For the first year following end of treatment, patients undergo follow-up neuroimaging studies every 3 months and then every 6 months for the second year and yearly thereafter. Only at 10 years following completion of treatment is the time between scans advanced to every 2 years. Therefore, we have a large cohort of patients with longitudinal outcome data. Eligibility criteria included a diagnosis of primary brain tumor, at least one unequivocal witnessed seizure, and treatment and follow-up at our institution. Medical records were abstracted using a standardized data acquisition tool by a single investigator (NJU) for demographic, clinical, and radiologic data. Clinical data collected included patient gender, age at diagnosis, length of follow-up, extent of iniEpilepsia, 56(10):1599–1604, 2015 doi: 10.1111/epi.13112
tial resection, tumor grade and location, tumor histology, and treatment modalities. Patient data were analyzed for tumor pathology and grade according to individual World Health Organization (WHO) histologic diagnosis and were coded by tumor type.6 Tumors with fewer than three cases each were grouped as “other.” This designation included teratoma and sarcoma. Information on tumor location was obtained from review of imaging studies and operative reports. Tumor location was analyzed by individual sites and grouped according to four categories: posterior fossa, midline/deep midline, cortical, and other. Neuroimaging studies were reviewed for the presence of residual tumor. The presence or absence of residual tumor was based on a combination of review of the neurosurgeons operative notes, discussion with the neurosurgeon, review of pre- and postoperative imaging (either intraoperative magnetic resonance imaging [MRI] or immediate postoperative MRI), and review of the first MRI done at follow-up, typically at 3 months. For patients who had experienced seizures at any time from initial presentation to the most recent follow-up visit, timing of seizure onset, seizure frequency, electroencephalography (EEG) abnormalities, and neuroimaging correlates were also obtained. Seizures were classified as simple or complex partial, generalized/secondarily generalized, and unclassified (seizures not adequately described to be classified according to the guidelines of the International League Against Epilepsy [ILAE]).7 Breakthrough seizures were defined as an unequivocal seizure occurring any time after antiepileptic drug initiation. Seizure frequency was defined as at least once daily, at least weekly, at least monthly, less than once per month, and unclear. Refractory seizures were defined as those that required three or more anticonvulsants concurrently. Ethical approval This retrospective, the Health Insurance Portability and Accountability Act (HIPAA)–compliant study was approved by the institutional review board of our institution with waiver of individual consent. Statistical analysis Univariate analysis was used to identify factors associated with seizure occurrence at any time from presentation to most recent follow-up. Individual risk factors as predictors in these models included glial or glial-neuronal tumor histology, cortical tumor location, subtotal resection at time of initial surgery, and recurrence of tumor. Candidate predictors of seizures were the clinical measures described above and measures of the treatment modalities. To establish bivariate associations between the response and the predictors, we fit a separate model for each candidate predictor. We concluded that there was a significant bivariate association between occurrence of seizures and a candidate predictor if the p-value for the likelihood ratio test was 2 years prior to diagnosis At time of diagnosis During treatment After completion of treatment 2 years from end of treatment Seizure frequency Daily Weekly Monthly
