The effects of hydrocephalus on intelligence quotient in children with localized infratentorial ependymoma before and after focal radiation therapy

Transcription

1 J Neurosurg (Pediatrics 2) 101: , 2004 The effects of hydrocephalus on intelligence quotient in children with localized infratentorial ependymoma before and after focal radiation therapy THOMAS E. MERCHANT, D.O., PH.D., HEATHER LEE, M.D., JUNHONG ZHU, M.S., XIAOPING XIONG, PH.D., GREGORY WHEELER, M.B.B.S., SEAN PHIPPS, PH.D., FREDERICK A. BOOP, M.D., AND ROBERT A. SANFORD, M.D. Departments of Radiation Oncology and Biostatistics, St. Jude Children s Research Hospital, Memphis, Tennessee Object. The goal of this study was to determine the influence of hydrocephalus on intelligence quotient (IQ) in children with infratentorial ependymoma before and after the administration of focal radiation. Methods. Measurements of ventricular size, including Evans index (EI), cella media index (CMI), frontal horn diameter (FHD), and ventricular angle, were performed using magnetic resonance imaging at the time of diagnosis and again at 3, 6, 9, and 12 months after the initiation of radiation therapy. Of the 59 patients (median age at time of radiation treatment, 4.1 years), the clinical diagnosis established in 50 (85%) was hydrocephalus and 23 (39%) required placement of a cerebrospinal fluid (CSF) shunt. Extent of resection was gross or near total in 50 (85%). Before and after radiation treatment, IQ was measured using age-appropriate testing. The correlation between multiple ventricular measurements and IQ was investigated using standard regression techniques and a generalized linear model. Patients with a higher EI (p = 0.04), CMI (p = 0.001), and FHD (p = ) at the time of diagnosis were more likely to have lower IQ scores before radiation treatment. Patients with higher CMI (p = 0.04) and FHD (p = 0.01) at the time of diagnosis were more likely to exhibit an increase in IQ score after radiotherapy. The rate of change in IQ after radiation treatment was positively correlated with the CMI intercept (p = 0.015) and negatively correlated with the rate of FHD change (p = 0.006). Conclusions. Changes in IQ score before and after radiation treatment are significantly influenced by the extent and treatment of hydrocephalus at the time of diagnosis. Hydrocephalus is an important factor to include when analyzing the effects of treatment. Patients who undergo a second surgery for ependymoma are more likely to require the placement of a CSF shunt (p = ). KEY WORDS ependymoma hydrocephalus cognition radiation treatment pediatric neurosurgery R ADIATION therapy is a mainstay in the treatment of brain tumors in children, including very young children suffering from ependymoma, who incur an increased risk of radiation-related treatment effects. 18 Recent clinical and technical advances in neuroimaging and radiation therapy have improved our ability to treat small target volumes and minimize the dose of radiation to normal tissues. Guidelines for the treatment of smaller target volumes have been developed and tested to demonstrate that the radiation-treated volume can be significantly reduced without affecting the rate of tumor control. 15,17 Documentation of the benefits of these advancements requires Abbreviations used in this paper: CMI = cella media index; CSF = cerebrospinal fluid; CTV = clinical tract volume; EI = Evans index; FHD = frontal horn distance; GTV = gross tumor volume; ICP = intracranial pressure; IQ = intelligence quotient; MR = magnetic resonance; PTV = planning target volume; VM = ventricular measurement. J. Neurosurg: Pediatrics / Volume 101 / November, 2004 the correlation of treatment dosimetry with objective measures of outcome, including those that assess cognitive, endocrine, and audiometric function. Furthermore, accurate modeling of the effects of radiotherapy requires the identification of those clinical and treatment factors that can affect patient outcome in tandem with the effects of radiation. In our investigations concerning the effects of radiation on cognition, endocrine function, and hearing in children with ependymoma, hydrocephalus, which has been known to affect neuropsychological function, 7,21,22 was identified as an important covariate. In most studies the assessment of hydrocephalus has been limited to a determination of its presence or absence at the time of diagnosis by using qualitative interpretation of neuroimaging findings and characterization of the severity of hydrocephalus in terms of the need or lack of need for placement of a CSF shunt. Such a simple, dichotomous assessment of hydrocephalus cannot account for the significance of severe hydrocephalus in a 159

2 T. E. Merchant, et al. patient in whom a shunt is not placed, and it cannot be used to determine the effects of change in ventricular size that might occur over time. Using measurements of ventricular size that have been developed for the classification of children with nontumor-related hydrocephalus, 19 we have quantitatively assessed the severity of hydrocephalus in children suffering from infratentorial ependymoma. This method of quantitative assessment allows hydrocephalus to function as a continuous variable in our modeling of the effects of radiation therapy. The purpose of the present study was to determine using a quantitative approach the effects of both hydrocephalus and change in ventricular size on cognitive function before and after radiation therapy in patients suffering from localized infratentorial ependymoma. Analysis of the results presented here indicates that factors other than radiation dosimetry influence cognitive outcome in at-risk patients, such as those with brain tumors and untreated hydrocephalus, who incur an increased risk of reduced cognition. Patient Population Clinical Material and Methods Fifty-nine patients with infratentorial ependymoma were enrolled in a protocol approved by the institutional review board of St. Jude Children s Research Hospital and received radiation therapy between August 11, 1997, and December 17, 2001 at our institution. Criteria for enrollment included the following: an age at the time of radiation treatment between 1 and 21 years, the presence of a histologically confirmed ependymoma, no evidence of dissemination, no prior radiation treatment, no ongoing chemotherapy, adequate cognitive performance status (electrocortigraphy 0 3), and informed consent. The protocol was amended to include one patient older than 21 years. There were 29 female and 30 male patients; median age at the time of radiation treatment was years (range years). Patients were further characterized according to tumor grade (anaplastic ependymoma, 14 cases; differentiated, 45 cases), prior chemotherapy (13 cases), extent of resection (gross total, 45 cases; near total, five cases; subtotal, nine cases), number of resections (one, 39 cases; two, 17 cases; three, one case; and four, two cases), hydrocephalus at the time of diagnosis which was assessed qualitatively (50 cases), and CSF shunt therapy (23 cases). Review of peri- and postoperative records revealed that approximately 39 patients required extraventricular catheter placement at the time of surgery, 10 received treatment for postoperative central nervous system infections (organism isolated or not isolated), and among the patients in whom shunts had been placed, three required one revision and three required three revisions. In no patient was endoscopic third ventriculostomy performed. Among the 13 patients who underwent radiation treatment prior to chemotherapy, most received multiagent chemotherapy that included cyclophosphamide, cisplatin or carboplatin, etoposide, and vincristine. Gross-total resection was defined as microscope-documented tumor visibility only and no evidence of disease on postoperative neuroimaging, near-total resection was defined as residual tumor with a thickness of less than or equal to 0.5 cm on postoperative neuroimaging, and subtotal resection was defined as residual tumor with a thickness of greater than 0.5 cm. In no patient was there residual tumor volume greater than 1.2 cm 3 at the time of radiation treatment. The mean ( 1 standard deviation) interval from the onset of symptoms to diagnosis was months. The median duration of the follow-up period for the entire group was 24 months at the time of analysis. Radiation Therapy Target volume definitions and treatment-planning parameters have been previously described. 15,17 Coregistered computerized tomography scans and MR images were used to define target volumes and normal tissue structures. The following target volume definitions, which are identical to those of the International Commission on Radiation Units and Measurement Report 50, 5 were used. The GTV included the tumor bed, residual tumor, or both. The CTV included the GTV and an added margin of 1 cm allocated for the treatment of subclinical microscopic disease beyond the margin of the tumor bed, residual tumor, or both. The CTV was anatomically confined and therefore limited by the normal tissue structures, such as the calvaria, skull base, tentorium cerebelli, falx cerebri, and sagittal and transverse venous sinuses, through which tumor extension was unlikely. The PTV, the purpose of which was to account for uncertainty in patient positioning, included the CTV surrounded by an additional margin of 0.3 to 0.5 cm in three dimensions. Additional structures defined to assist in the treatment planning process included the entire brain, temporal lobes, eyes, optic chiasm, pituitary gland, hypothalamus, cochleae, and spinal cord. The treatment was planned for most cases with the objective of maximizing the conformity of treatment and minimizing the radiation dose to normal tissues. Intensitymodulated radiation therapy was used to optimize the treatment of those targets with concave surfaces. The number of beams ranged from two to 28. The field edge was 0.6 to 0.8 cm beyond the PTV. The homogeneity of the dose was optimized across the PTV with beam modifiers that included physical or virtual wedges. Target coverage and the dose targeting critical structures, which were in most cases the cervical spinal cord and optic chiasm, were evaluated by examining the isodose distributions throughout these structures in the cardinal planes and with the assistance of dose volume histograms. In view of these evaluations, the optimal beam directions and the prescription isodose surface were selected. Conventional fractionation of 1.8 Gy/day was administered in all patients. The prescribed dose was 59.4 Gy; exceptions included children younger than the age of 18 months, who instead underwent gross-total resection and were treated with 54 Gy. The dose to the upper cervical spinal cord was limited to 54 Gy, that to the optic chiasm was 55.8 Gy, and that to the optic nerves was 50.4 Gy. Most children required a volume reduction to restrict the radiation dose received by the cervical spinal cord. These field reductions were most often accomplished by modifying the fields used in the original treatment plan to exclude the crucial structure. The dose administered to the supratentorial brain in the study group is presented in Fig. 1 to demonstrate the dose-sparing ability of the conformal guidelines and the techniques used to treat these patients. 160 J. Neurosurg: Pediatrics / Volume 101 / November, 2004

3 Effects of hydrocephalus on IQ in childhood ependymoma Neuropsychometric Testing Neuropsychometric testing was conducted before radiation treatment and again at 6, 12, 24, 36, and 48 months after the initiation of radiation therapy. Baseline testing was delayed slightly in cases in which the institution of radiation therapy was considered urgent; in such cases, baseline testing was completed during radiation therapy. Age-appropriate tests included the Bayley Scale of Infant Development, 2 Wechsler Preschool and Primary Scales of Intelligence Revised, 30 Wechsler Intelligence Scale for Children III, 29 and Wechsler Adult Intelligence Scale III. 28 The testing regimen was based on patient age; all tests yielded age-corrected scores. Decisions regarding instrument selection reflected the desire to conduct at least two consecutive evaluations with the same instrument. Investigators used their clinical judgment to select the instrument with which the child was expected to achieve the most accurate performance. The primary outcome measure for this study was the estimated IQ, or Bayley Mental Index score, and not the full-scale IQ, which was obtained only at the 2-year evaluation unless warranted earlier. Ventricular Measurements Ventricular measurements were determined using MR images obtained at the time of diagnosis (prior to intervention); before radiation treatment; and at 3, 6, 9, and 12 months after radiation treatment. Measurements were obtained from MR imaging performed before surgery, before radiation therapy, and every 3 months after radiation therapy for the first 12 months. Measurements included the EI, CMI, FHD, and ventricular angle as described by Mori, et al., 19 and modified by our group to include the FHD instead of the frontal horn radius. A typical case is presented in Fig. 2. We chose linear measurements that were easily reproducible and correlate well with volumetric measures of ventricular size. 20 Statistical Analysis J. Neurosurg: Pediatrics / Volume 101 / November, 2004 FIG. 1. Radiation dose volume histogram demonstrating the supratentorial brain volume measured in 59 pediatric patients with localized ependymoma. The mixed (random and fixed effects) model 23 was used to estimate the trends of VMs in the period between diagnosis (prior to any intervention) and initiation of radiation therapy, and the trends of IQ score in the period before and after radiation therapy, and to explore possible relationship between VMs and the longitudinal change of IQ scores. In particular, we were interested in the following correlations: 1) between single VMs at the time of diagnosis and IQ measured prior to radiation therapy; 2) between single VMs taken prior to radiation therapy and IQ measured prior to radiation therapy; 3) change in IQ score with the change in VMs; 4) change in IQ score with single VMs taken at diagnosis; 5) change in IQ score with single VMs performed before radiation therapy; 6) effect of CSF shunt placement on change in IQ score; 7) effect of shunt placement on change in VMs; and 8) effect of shunt placement on the correlation between change in IQ score and change in VMs. For each patient, the trends of the VMs (EI, CMI, FHD, and ventricular angle) (during a short period from diagnosis to the last recording of VMs after radiation therapy) were characterized by intercepts and slopes of regression lines of VMs. The trend of IQ scores (during a longer period from prediagnosis IQ testing to the last IQ examination conducted after radiation therapy) was characterized by the intercept and slope of regression line of IQ observations. The time for regression for IQ is set at 0 at the start of radiation therapy; the intercept of IQ regression is the mean IQ at start of radiation therapy. We calculated the Pearson correlations between intercepts and slopes of VMs and those of IQ among the patients. As indicated by these correlation coefficients, the short-term trends of VMs are predictive for the long-term trends of IQ score after radiation therapy. To test the effect of CSF FIG. 2. An MR image of the structures whose measurements were used to calculate the EI (ratio of b/a) (A) and the CMI (ratio of B/A) (B) and structures whose measurements were directly evaluated. The bidirectional arrow (C) denotes the FHD; (D), the ventricular angle. 161

4 T. E. Merchant, et al. shunt placement on the change in IQ scores and VMs, we grouped cases according to shunt status and then compared the mean intercepts and slopes of IQ and VMs for the patients between the two groups (shunt treated and not shunt treated). All analyses in this paper were conducted using the statistical package SAS/STAT (SAS Institute, Cary, NC). Changes in VMs Results A total of 788 VMs were performed using 297 individual MR imaging studies obtained in our 59 pediatric patients with infratentorial ependymoma (Table 1). There was good correlation between the VMs (Table 2): the EI, CMI, and FHD decreased significantly as time after diagnosis increased (p = for EI; p = for CMI; p = for FHD). These changes were inversely proportional to those for the ventricular angle, which increased significantly (p = 0.001) as time after diagnosis increased (Fig. 3). Changes in IQ Score There was no significant change in IQ score for the group overall; however, there was a trend toward an increase in IQ relative to the duration of time after radiation therapy. The mean follow-up period in the entire group was 24 months after the initiation of radiation therapy, and a total of 188 IQ examinations contributed to the analysis (Fig. 4). The measure of IQ before radiation therapy had a significant negative correlation (correlation coefficient 0.38) with a change in IQ score after radiation therapy (p = 0.006); patients with lower IQ scores before radiation therapy had the greatest rates of score increase after radiation had been administered. As a covariate in our analysis, age influenced the intercept (preradiation value) but not change over time. The resulting model demonstrated that IQ = age time; age at the time of radiation therapy (p = ) and time after the initiation of radiation therapy are measured in months, and and are coefficients. Correlation Between VMs and IQ Statistical analysis revealed strong correlations among the four VMs performed at the time of diagnosis (Table 2). Ventricular measurements acquired at the time of diagnosis correlated significantly with IQ measured before radiation therapy and change in IQ score after radiation therapy (Table 3). For example, patients with higher EI (correlation coefficient 0.3; p = 0.04), CMI (correlation coefficient 0.46; p = 0.001), or FHD (correlation coefficient 0.53; p = ) at the time of diagnosis were more likely to have lower IQ scores prior to radiation therapy. Patients with higher CMI (correlation coefficient 0.3; p = 0.04) and FHD (correlation coefficient 0.37; p = 0.01) were more likely to have an increase in IQ after radiation therapy. A change in IQ after radiation therapy was found to correlate with change in VMs; specifically, the rate of change in IQ score had a significant negative correlation with the rate of FHD change (correlation coefficient 0.38; p = 0.006) (Table 4 and Fig. 5). The rate of change in IQ score correlated significantly with the CMI intercept (correlation coefficient 0.34; p = 0.015). There was a strong correlation among VMs performed immediately prior to radiation therapy (Table 5); however, they did not predict IQ results before or after radiation therapy. Effects of CSF Shunt Placement on VMs and IQ In patients who required CSF shunt placement significantly lower IQ scores were observed before radiation therapy than in patients who had not undergone CSF shunt placement (p = 0.03 [t-test]; Table 6). A significant negative correlation was demonstrated between change in IQ score and change in FHD (correlation coefficient 0.5; p = 0.03) in patients with CSF shunts. There was no significant intergroup difference between change in IQ based on shunt status. A significant negative correlation was observed between IQ score before radiation therapy and CMI performed at diagnosis in patients without shunts (correlation coefficient 0.58; p = ). Practical Illustrations of VMs Within the range of VMs (Table 1) and IQ (baseline TABLE 1 Summary of descriptive statistics for the VMs* Mos After Radiation Therapy Before Radiation VMs At Diagnosis Therapy EI mean range CMI mean range FHD (cm) mean range VA ( ) mean range * Mean values are presented as means standard error of the means (SEMs). Abbreviation: VA = ventricular angle. 162 J. Neurosurg: Pediatrics / Volume 101 / November, 2004

5 Effects of hydrocephalus on IQ in childhood ependymoma TABLE 2 Summary of correlation coefficients among VMs throughout study* median IQ 90, range ) recorded for the study group, the preradiation IQ score may be predicted by the graphs presented in Fig. 6. Patients with predictably low IQ scores at the time that radiation therapy is initiated appear to have the most to gain from a reduction in ventricular size and should be considered for shunt treatment in cases in which persistent ventriculomegaly and/or the signs and symptoms of increased ICP are present. Figure 5 serves as a guide for an acceptable rate of change in VMs when CSF flow has been reestablished as a result of resection or shunt placement. Evaluation of Endocrine Function VMs (no. of measurements) VMs CMI VA FHD EI (296) (300) (297) CMI (296) (293) VA (297) * All correlation coefficients are significant (p ). Prior to radiation therapy the patients included in this study underwent evaluation for endocrinopathy, including provocative tests of endocrine function. 16 Results demonstrated that 27 of 59 patients suffered from a preexisting hormone deficiency. Growth hormone deficiency was observed in 17 patients and evidence of central hypothyroidism or adrenal insufficiency appeared in 10 patients. Our policy was to begin hormone replacement therapy immediately in patients with hypothyroidism and adrenal insufficiency. Those patients with growth hormone deficiency were observed for evidence of growth deceleration and when indicated after 12 months of progression-free follow up received hormone replacement therapy. Discussion After careful observation, we hypothesized that both the severity and the management of hydrocephalus influences patient s performance on IQ tests before and after radiation therapy. This hypothesis was supported by previously reported findings that pointed to hydrocephalus as a possible cause of several potentially reversible neurocognitive conditions, 3,8,11,25,26 poor academic achievement and learning, and poor outcomes, as indicated by various measures of cognition 1,6,10,27,31 including adaptive functioning. 9 Detailed reports with descriptions of the influence of hydrocephalus on cognitive outcomes in children with FIG. 3. Graphs of modeled VMs demonstrating changes in EI, CMI, FHD, and ventricular angle before and after radiation therapy. J. Neurosurg: Pediatrics / Volume 101 / November,

6 T. E. Merchant, et al. FIG. 4. Graph demonstrating pre- and postradiation IQ measurements in pediatric patients with infratentorial ependymoma. Mean values ( 1 standard deviation) are plotted (dashed line) with a modeled curve (solid line) and the estimating equation (IQ = time). Month 0 refers to the time of radiation therapy. brain tumors are limited to our previous study 14 (attention, impulsivity, and reaction time assessed during radiation therapy) and that of investigators who measured presurgical cognitive function in patients with brain tumors. 4 In the present study we proved that ventricular size changes significantly over time, that this change depends on the extent of hydrocephalus at the time of diagnosis and the absence or presence of a CSF shunt, and that the VMs used in this study correlated well with each other and can be reproduced. The most important findings in this study were the following: 1) correlation of IQ tested before radiation therapy with VMs obtained at the time of diagnosis; 2) greatest improvement in IQ scores after radiation therapy in patients with more extensive hydrocephalus at the time of diagnosis; 3) prediction of postradiation IQ scores on the basis of the change in VMs; and 4) dependence of correlations between VMs and IQ on the status of CSF shunt treatment. Our findings demonstrate the magnitude of the effect of hydrocephalus and its management on pre- and postradiation cognition in children with infratentorial ependymoma. Analysis of these findings indicates that the effects of hydrocephalus may be reversible, that hydrocephalus should be a factor included in the modeling of treatment, that the assessment of hydrocephalus as a continuous variable is feasible and valuable, and that untreated hydrocephalus increases the risk of cognitive sequelae in patients with brain tumors. Patients suffering from severe hydrocephalus at the time that infratentorial ependymoma is diagnosed experience an increased risk of cognitive deficits, and careful management is necessary to reduce ventricular size in such patients. For most patients with infratentorial ependymoma, removal of the tumor, with or without decompression of the ventricular system via a temporary extraventricular shunt, relieves the obstruction and reestablishes normal CSF flow. For other patients, lack of improvement in ventricular size and symptoms of increased postoperative ICP necessitate the placement of a permanent CSF shunt. In such cases the resistance setting on the device is selected to achieve drainage and minimize the development of subdural fluid collections. When the slopes and intercepts from the linear modeling of VMs and IQ were correlated, patients with extensive hydrocephalus at the time of diagnosis were found to TABLE 3 Summary of correlation coefficients between the intercept and slope of IQ with the VMs performed at time of diagnosis VM (p value) VM EI CMI VA FHD IQ intercept 0.30 (0.0400) 0.46 (0.0010) 0.09 (0.5300) 0.53 (0.0002) IQ slope 0.18 (0.23) 0.30 (0.04) 0.03 (0.85) 0.37 (0.01) EI 0.60 ( ) 0.28 (0.0400) 0.51 (0.0001) CMI 0.34 (0.0100) 0.48 (0.0004) VA 0.49 (0.0002) 164 J. Neurosurg: Pediatrics / Volume 101 / November, 2004

7 Effects of hydrocephalus on IQ in childhood ependymoma TABLE 4 Summary of correlation coefficients between the intercept and slope of IQ and the VMs VM (p value) EI CM FHD VA IQ Variable Slope (p value) Intercept Slope Intercept Slope Intercept Slope Intercept Slope IQ intercept (0.006) (0.040) (0.370) (0.004) (0.110) (0.290) (0.080) (0.720) (0.170) slope (0.190) (0.900) (0.015) (0.140) (0.630) (0.006) (0.460) (0.310) J. Neurosurg: Pediatrics / Volume 101 / November, 2004 have the lowest IQ scores before radiation therapy, and the rate of change in VMs were correlated negatively and significantly with postradiation IQ scores. This finding reveals the effects of hydrocephalus and its management on outcome. Effective treatment of hydrocephalus is crucial to the cognitive outcome of children with brain tumors. When the VMs performed only at the time of diagnosis were correlated with the slopes and intercepts from the linear modeling of IQ, not only were patients with extensive hydrocephalus at the time of diagnosis found to have the lowest preradiation IQ scores, they were also more likely to have improvement in postradiation IQ scores. This finding indicates the reversibility of the effects of hydrocephalus and identifies an at-risk group patients without hydrocephalus who demonstrate low IQ at the time of radiation therapy. Such patients are not likely to improve and may be more vulnerable to the effects of treatment. When VMs performed only at the time of radiation therapy were correlated with the slopes and intercepts from the linear modeling of IQ, these measurements did not predict IQ at the time of radiation therapy nor the change in IQ after radiation therapy. This result is not surprising because the time interval between surgery and radiation therapy is often measured in months and because surgery or shunt placement to treat hydrocephalus results in VMs in the normal range in most patients. When the slopes and intercepts from the linear modeling of VMs and IQ were compared between patients with and without shunts, patients in the former group exhibited a significantly lower IQ before radiation therapy. There was no significant difference in the change in IQ after radiation therapy between patients with and without shunts. Unlike other patients in our study, those with shunts had a significant negative correlation between IQ intercept (IQ before radiation) and change in IQ, and between IQ slope and FHD. That the majority of correlations exist for the patients who had shunts is not surprising. Shunt placement is an indicator of severe hydrocephalus and provides a benefit in terms of cognitive outcome to patients. The only correlation unique to patients without shunts was the negative correlation between CMI intercept and IQ slope. The CMI appears to be more sensitive and specific in predicting IQ outcomes for all patients. Because the IQ scores for the patients in our study fall within the normal range, modeling outcomes on the basis of clinical factors such as hydrocephalus can be difficult. The high degree to which hydrocephalus and its management appear to influence IQ is manifested by the correlations found in the group of patients with limited treatment effects. Although the outcome for most patients was good, we were able to identify those at risk for cognitive deficiency; such patients might benefit from early intervention. In view of the negative correlation between the rate of change in ventricular size and the rate of change in IQ after radiation therapy, we were also identifying patients who might benefit from shunt placement. The results of this study demonstrate that hydrocephalus has a significant influence on cognitive function before radiation therapy additional evidence that factors other than radiation therapy contribute to the long-term effects of therapy in children with brain tumors. The dose and volume of radiation therapy have been reported to affect the cognitive outcome of these children, and significant declines in cognitive function have been documented. The early introduction of radiation therapy has been found to be crucial to disease control, especially in very young children who are most vulnerable to radiation-related treatment effects. The expectation is that children with brain tumors treated with radiation therapy are at risk of a decline in cognitive function. Young age at the time of radiation therapy, increased radiation dose and volume, and time after radiation therapy are the causes most often cited for such decline. Significant declines in cognitive function after radiation treatment have been documented, leading caregivers to FIG. 5. Graph showing IQ slope plotted as a function of FHD slope for all patients and according to the presence or absence of a shunt. 165

8 T. E. Merchant, et al. TABLE 5 Summary of correlation coefficients between the intercept and slope of IQ with the VMs obtained immediately before radiation therapy VM (p value) Variable EI CMI VA FHD IQ intercept 0.21 (0.14) 0.21 (0.15) 0.04 (0.80) 0.12 (0.40) IQ slope 0.12 (0.40) 0.26 (0.07) 0.02 (0.90) 0.13 (0.37) EI 0.65 ( ) 0.27 (0.0400) 0.16 (0.2300) CMI 0.52 ( ) 0.17 (0.2000) VA 0.08 (0.5400) fear radiation therapy as a treatment modality, especially for very young children. Clinical and technical advances in neuroimaging and radiation therapy have improved the ability to treat small volumes with increased precision, thus allowing for significant reduction in the volume of normal brain exposed to radiation during treatment. Criteria for permanent CSF shunt placement in this study included enlarged or enlarging ventricles with persistent signs and symptoms of increased ICP. In no patient was a shunt placed on the sole criterion of ventricular size, and in no patient was a shunt placed during the analysis period when the correlation between VMs and IQ score became apparent. The incidence of shunt-treated patients in this study (39%) exceeds those reported in recent series (25 30%). 24 The number of patients who underwent a second surgery was relatively high in this study because of our institutional preference to perform gross-total resection prior to radiation therapy in patients with ependymoma. Sixteen of 22 patients in whom shunts were eventually placed had undergone a second surgery compared with only four of the 37 patients in whom shunts had not been placed (p = ). This finding indicates that patients who undergo a second surgery for tumor resection incur an increased risk of requiring the placement of a CSF shunt. Most children with infratentorial ependymoma present with hydrocephalus at the time of initial diagnosis. Resection of the primary tumor and measures to reduce ICP in the perioperative period usually result in only a small proportion of patients who require the placement of a permanent CSF shunt. A second resection for residual tumor, which is often necessary in the treatment of infratentorial ependymoma, increases the risk that the patient will require the placement of a permanent CSF shunt. This study does not account for the effects of multiple shunt adjustments, postoperative meningitis, or shunt infections which are known to impact psychometric function; these complications occurred in less than 10% of patients in our study. If a patient with persistent ventriculomegaly experiences a decrease in the size of ventricles after shunt placement, the shunt was probably indicated. If the ventricles do not decrease in size, persistent ventriculomegaly is most often related to ex vacuo dilation. Analysis of the data from this study provides new parameters for determining whether shunt placement or shunt revision might be necessary: the correlation between ventricular size and IQ. The standard practice has been to avoid placement of a shunt whenever possible; however, persistent ventriculomegaly is clearly harmful. In the absence of the classic criterion for shunt placement, we would suggest that surgeons consider the effects of tumor and surgery-related morbidity, knowledge of the patient s academic or other performance status prior to diagnosis, and preradiation IQ score. Similar criteria might be developed for a prospective trial to confirm the findings of this study, in which hydrocephalus resolved postoperatively in most patients. We have followed a large series of patients with ependymoma who were treated using radiation therapy and systematic targeting guidelines. We observed no significant cognitive decline even among the youngest patients (age 3 years) during a median follow-up period of more than 3 years. And while decline is always possible, we believe that the side effects of radiation therapy have been significantly reduced using advanced radiation therapy methods. Only through careful observation and analysis of a sufficiently large number of patients treated with radiation therapy in a uniform manner can factors that influence cognitive outcomes be identified. Reducing the side effects of radiation therapy will make the impact of other clinical and treatment factors more prominent. In some patients with hydrocephalus, for example, the impact of ventriculomegaly at the time of diagnosis, or that of untreated ventriculomegaly after diagnosis, may exceed the effects of radiation therapy. Children with hydrocephalus that requires permanent TABLE 6 Summary of the effect of shunt placement on change in IQ No. of Mean SEM Mean w/ and Variable Shunt Cases (min, max) w/o Shunt* SEM p Value IQ intercept yes (34.41, ) no (50.0, ) IQ slope yes ( 0.161, 0.147) no ( 0.067, 0.088) * The difference between mean values for shunt-treated and nonshunt-treated patients. 166 J. Neurosurg: Pediatrics / Volume 101 / November, 2004

9 Effects of hydrocephalus on IQ in childhood ependymoma FIG. 6. Graphs of VMs showing the EI (upper left), FHD (in centimeters; upper right), CMI (lower left), and ventricular angle (in degrees; lower right) at the time of diagnosis modeled according to observed preradiation IQ score for shunt-treated (solid line) and nonshunt-treated (dashed line) pediatric patients with infratentorial ependymoma. CRT = conformal radiation therapy. placement of a CSF shunt incur increased risks of cognitive deficits, 14 hearing loss, 13 and endocrinopathy 16 prior to radiation therapy, according to recent analyses. Children in whom shunts had been placed demonstrated lower baseline scores and higher hearing thresholds than their counterparts without shunts. Patients with shunts appeared to have improvement in their cognitive function during the first 9 months after the initiation of radiation therapy. This observation demonstrates the therapeutic value of CSF shunt placement and leads one to question whether patients who present with severe hydrocephalus would be better served by permanent shunts. In our analysis of audiometric data from a comparison of children who were treated with and without chemotherapy prior to radiation therapy, those with permanent CSF shunts demonstrated a clinically significant increase in their hearing threshold during the first 2 years after radiation therapy. The rate of change over time was accelerated with increasing dose to the cochlea. Children with permanent shunts, however, including those who received chemotherapy before radiation therapy, appeared to have less tolerance to the ototoxic effects of radiation. Conclusions Our findings confirm the influence of hydrocephalus on outcome in children with primary brain tumors. Moreover, it can be inferred from our data that VMs should be assessed at the time of diagnosis and then followed carefully to assist physicians in decisions regarding the management of hydrocephalus. J. Neurosurg: Pediatrics / Volume 101 / November, 2004 References 1. Barnes MA, Pengelly S, Dennis M, et al: Mathematics skills in good readers with hydrocephalus. J Int Neuropsychol Soc 8: 72 82, Bayley N: The Bayley Scales of Infant Development, ed 2. San Antonio: The Psychological Corporation, Boon AJ, Tans JT, Delwel EJ, et al: Dutch normal-pressure hydrocephalus study: prediction of outcome after shunt placement by resistance to outflow of cerebrospinal fluid. J Neurosurg 87: , Brookshire B, Copeland DR, Moore BD, et al: Pretreatment neuropsychological status and associated factors in children with primary brain tumors. Neurosurgery 27: , Committee on Prescribing, Recording, and Reporting Electron Beam Therapy: ICRU Report 50. Prescribing, Recording, and Reporting Photon Beam Therapy. Bethesda, MD: International Commission on Radiation Units and Measurements, Ding Y, Lai Q, McAllister JP II, et al: Impaired motor learning in children with hydrocephalus. Pediatr Neurosurg 34: , Fletcher JM, McCauley SR, Brandt ME, et al: Regional brain tissue composition in children with hydrocephalus. Relationships with cognitive development. Arch Neurol 53: , Hejl A, Høgh P, Waldemar G: Potentially reversible conditions in 1000 consecutive memory clinic patients. J Neurol Neurosurg Psychiatry 73: , Hommet C, Billard C, Gillet P, et al: Neuropsychologic and adaptive functioning in adolescents and young adults shunted for congenital hydrocephalus. J Child Neurol 14: , Lumenta CB, Skotarczak U: Long-term follow-up in 233 patients with congenital hydrocephalus. Childs Nerv Syst 11: ,

10 T. E. Merchant, et al. 11. Mataro M, Junque C, Poca MA, et al: Neuropsychological findings in congenital and acquired childhood hydrocephalus. Neuropsychol Rev 11: , Merchant TE, Goloubeva O, Pritchard DL, et al: Radiation dose-volume effects on growth hormone secretion. Int J Radiat Oncol Biol Phys 52: , Merchant TE, Gould CJ, Xiong X, et al: Early neuro-otologic effects of three-dimensional radiation in children with primary brain tumors. Int J Radiat Oncol Biol Phys 58: , Merchant TE, Kiehna EN, Miles MA, et al: Acute effects of radiation on cognition: changes in attention on a computerized continuous performance test during radiation therapy in pediatric patients with localized primary brain tumors. Int J Radiat Oncol Biol Phys 53: , Merchant TE, Mulhern RK, Krasin MJ, et al: Preliminary results from a phase II trial of conformal radiation therapy and evaluation of radiation-related CNS effects for pediatric patients with localized ependymoma. J Clin Oncol 22: , Merchant TE, Williams T, Smith JM, et al: Preirradiation endocrinopathies in pediatric brain tumor patients determined by dynamic tests of endocrine function. Int J Radiat Oncol Biol Phys 54:45 50, Merchant TE, Zhu Y, Thompson SJ, et al: Preliminary results from a Phase II trial of conformal radiation therapy for pediatric patients with localized low-grade astrocytoma and ependymoma. Int J Radiat Oncol Biol Phys 52: , Merchant TE: Current management of childhood ependymoma. Oncology 16: , Mori K, Shimada J, Kurisaka M, et al: Classification of hydrocephalus and outcome of treatment. Brain Dev 17: , O Hayon BB, Drake JM, Ossip MG, et al: Frontal and occipital horn ratio: a linear estimate of ventricular size for multiple imaging modalities in pediatric hydrocephalus. Pediatr Neurosurg 29: , Ralph K, Moylan P, Canady A, et al: The effects of multiple shunt revisions on neuropsychological functioning and memory. Neurol Res 22: , Reimers TS, Ehrenfels S, Mortensen EL et al: Cognitive deficits in long-term survivors of childhood brain tumors: identification of predictive factors. Med Pediatr Oncol 40:26 34, Rutter CM, Elashoff RM: Analysis of longitudinal data: random coefficient regression modeling. Stat Med 13: , Schijman E, Peter JC, Rekate HL, et al: Management of hydrocephalus in posterior fossa tumors: how, what, when? Childs Nerv Sust 20: , Silver BV, Chinarian J: Neurologic improvement following shunt placement for post-traumatic hydrocephalus in a child. Pediatr Rehabil 1: , Tashiro Y, Drake JM: Reversibility of functionally injured neurotransmitter systems with shunt placement in hydrocephalic rats: implications for intellectual impairment in hydrocephalus. J Neurosurg 88: , Thompson NM, Fletcher JM, Chapieski L, et al: Cognitive and motor abilities in preschool hydrocephalics. J Clin Exp Neuropsychol 13: , Weschler D: The Wechsler Adult Intelligence Scale Revised. San Antonio: The Psychological Corporation, Weschler D: The Wechsler Intelligence Scale for Children, ed 3. San Antonio: The Psychological Corporation, Weschler D: The Wechsler Preschool and Primary Scale of Intelligence Revised. San Antonio: The Psychological Corporation, Yamada J: Neurological origins of poor reading comprehension despite fast word decoding? Brain Lang 80: , 2002 Manuscript received October 14, Accepted in final form June 14, This study was supported by Grant No. RPG CCE from the American Cancer Society and by American Lebanese Syrian Associated Charities (ALSAC) of St. Jude Children s Research Hospital. Address reprint requests to: Thomas E. Merchant, D.O., Ph.D., Department of Radiation Oncology, Mail Stop 220, St. Jude Children s Research Hospital, 332 North Lauderdale Street, Memphis, Tennessee thomas.merchant@stjude.org. 168 J. Neurosurg: Pediatrics / Volume 101 / November, 2004