Fax: (416) 813-8024
Longitudinal evaluation of neurocognitive function after treatment for central nervous system germ cell tumors in childhood†
Article first published online: 11 MAY 2011
Copyright © 2011 American Cancer Society
Volume 117, Issue 23, pages 5402–5411, 1 December 2011
How to Cite
Mabbott, D. J., Monsalves, E., Spiegler, B. J., Bartels, U., Janzen, L., Guger, S., Laperriere, N., Andrews, N. and Bouffet, E. (2011), Longitudinal evaluation of neurocognitive function after treatment for central nervous system germ cell tumors in childhood. Cancer, 117: 5402–5411. doi: 10.1002/cncr.26127
Portions of this article were presented at the 14th International Symposium on Pediatric Neuro-Oncology, Vienna, Austria, June 2010.
- Issue published online: 18 NOV 2011
- Article first published online: 11 MAY 2011
- Manuscript Accepted: 17 FEB 2011
- Manuscript Revised: 7 JAN 2011
- Manuscript Received: 9 NOV 2010
- germ cell tumors;
- neurocognitive late effects;
Central nervous system germ cell tumors (CNS GCT) are typically localized to midline structures of the brain, including the pineal and suprasellar/pituitary regions. Management of these tumors depends on underlying histology (germinoma or nongerminomatous germ cell tumor). Knowledge about neurocognitive outcome in these patients is limited. Longitudinal neurocognitive outcome in CNS GCT patients seen for neuropsychological evaluation at a single institution was explored.
Thirty-five patients were seen for neurocognitive evaluation after diagnosis and treatment for a CNS GCT. Mean age at diagnosis was 11.66 years. Tumor location was suprasellar in 12 patients, pineal in 9, bifocal in 10, multifocal in 3, and thalamic in 1. Standardized cognitive tests of intelligence, receptive language, visual-motor ability, memory, and academic achievement were administered. Longitudinal and cross-sectional analyses were conducted.
Intelligence, academic functioning, and receptive vocabulary were not significantly compromised in most patients treated for CNS GCT. Working memory, information processing speed, and visual memory declined significantly over time in all patients. Patients with pineal tumors showed early and stable deficits, whereas patients with suprasellar and bifocal tumors showed more protracted declines from initial average functioning. Patients treated with ventricular versus craniospinal radiation displayed better outcome.
Although general cognitive abilities appeared stable and intact after treatment for most children with CNS GCT, a significant decline over time in working memory, processing speed, and visual memory was evident. Tumor location appeared to be important in understanding the trajectory of stability and decline in CNS GCT patients, as did radiation field. Cancer 2011;. © 2011 American Cancer Society.
Central nervous system germ cell tumors (CNS GCT) account for 3% to 6% of all brain tumors in pediatric patients. 1, 2 Two-thirds of all CNS GCT are germinomas. 3 Age at diagnosis tends to be in older childhood and adolescence. 4 The majority of CNS GCT occur in the midline structures of the brain, and the 2 most common sites are the pineal (56%) and the suprasellar (28%) regions. 2, 5 Bifocal involvement of both the pineal and suprasellar regions has been reported in a range of 6% to 26%. 6-9 Patients with tumors in the pineal region can present with increased intracranial pressure due to obstructive hydrocephalus and/or Parinaud syndrome. 10, 11 Patients with suprasellar tumors can present with diabetes insipidus, symptoms of hypopituitarism, and visual disturbances. 11 CNS GCT are typically managed with chemotherapy followed by either focal or craniospinal radiation. 12 If there is evidence of dissemination, then craniospinal irradiation (CSI) is used. 13 Historically, CSI followed by a boost to the primary tumor area has been the standard treatment for intracranial germinoma. 14 Systemic chemotherapy followed by involved field irradiation has been tested as a means of reducing radiation doses and/or volume while maintaining high cure rates. 15 Survival rate in patients with nongerminomatous germ cell tumors is slightly lower, in the range of 60% to 70%.
Although treatment of CNS GCT has gained more attention, reports on neurocognitive outcome in these patients are sparse. The existing neurocognitive late effects literature is predominately focused on craniospinal radiation for posterior fossa medulloblastoma and to a lesser extent ependymoma. A progressive decline in neurocognitive functioning after cranial-spinal for medulloblastoma is well documented. 16-21 Limited field radiation appears to be associated with better neurocognitive outcome. 7, 22, 23 It is difficult to generalize this previous literature to CNS GCT, however. First, CNS GCT arise in the cerebral midline as opposed to the posterior fossa, and hence has an impact on different cognitive systems. Second, CNS GCT are generally treated with lower doses and/or smaller fields of radiation than medulloblastoma. Finally, on the whole, patients with CNS GCT are commonly diagnosed and treated at an older age than those with medulloblastoma. Considering that the vast majority of patients survive CNS GCT, specific knowledge regarding neurocognitive late effects of this disease is essential.
Survivors of CNS GCT show lower rates of postsecondary education and employment than the normative population. 24, 25 Most studies of neurocognitive function in CNS GCT are restricted to either retrospective reports 26, 27 or very small patient samples (n < 10). 7, 9, 11, 28 Full-scale and verbal intelligence remains relatively intact after treatment, but decreased speed of information processing and visual-spatial intelligence is evident. 7, 9, 11, 26, 27, 29 Intact academic achievement and receptive vocabulary have also been documented. 7, 26 Younger patients show poorer intellectual outcomes than older patients. 26 Memory impairment has been noted in a single case study of a patient with a pineal/thalamic germinoma. 28 In the sole study examining change over time, Merchant et al 9 documented stable and average intelligence in 8 patients treated with craniospinal radiation (dose = 2560 cGy).
Because of the excellent survival chances for patients with CNS GCT, it is important to longitudinally investigate the course and pattern of neurocognitive function. Hence, the primary goal of the current study is to examine patterns of change and stability in multiple domains of neurocognitive function over time in children treated for CNS GCT. Further, the impact of tumor location (ie, pineal, suprasellar, bifocal) has not been considered in prior studies of neurocognitive outcome in CNS GCT. We specifically evaluated the impact of tumor location on longitudinal neurocognitive outcome. Finally, a secondary goal was to examine patterns of neurocognitive functioning at a single long-term time point in children treated for CNS GCT to evaluate the impact of demographic and medical variables on neurocognitive outcome.
MATERIALS AND METHODS
Participants and Procedures
Thirty-five children (21 males) treated for CNS GCT within the last 20 years at the Hospital for Sick Children were included (Table 1). Tumor location was suprasellar in 12 patients, pineal in 9, bifocal in 10, multifocal in 3, and thalamic in 1. Patients with germinoma were treated according to the best clinical practice (1990-2000) (ie, a platinum-based regimen followed by focal or craniospinal radiation, depending on the disease status at diagnosis) and subsequently according to arm B of the International Society of Paediatric Oncology [SIOP] 96 GCT protocol). 30 Arm B consisted of 2 courses of carboplatin-etoposide, alternating with 2 courses of ifosfamide-etoposide, followed by focal radiation for nonmetastatic lesions and craniospinal radiation for metastatic germinomas. The recommended radiation field (involved field) included the initial tumor volume with a 2-cm safety margin. This protocol did not provide specific guidelines for radiation management of bifocal tumors. For patients with bifocal lesions, the radiation field was defined as the periventricular volume, followed or not by a boost encompassing both sites of initial disease. Since 2006, all nonmetastatic germinoma patients were treated with a ventricular field of radiotherapy at a dose of 24 Gy without any boost. For patients with nongerminomatous germ cell tumors, the protocol of chemotherapy varied over time, although all regimens used a combination of etoposide and platinum compound, either cisplatin or carboplatin. Information regarding radiation dose/field is presented in Table 1 for patients with pineal, suprasellar, and bifocal tumors.
|Medical and Demographic Characteristics|
|No. of patients||35|
|Mean age, y, at diagnosis||11.66 (SD=3.67)|
|Mean age, y, at baseline/first neuropsychological assessment||12.95 (SD=3.05)|
|Mean time, y, from diagnosis to baseline/first assessment||1.01|
|Frequency of patients seen for baseline/first assessment prior to treatment with cranial radiation||17 (48.57%)|
|Frequency of patients seen for baseline-first assessment after completion of chemotherapy||9 (25.71%)|
|Mean time, y, from first to last neuropsychological assessment||3.31|
|CNS GCT diagnosis|
|Immature teratoma||1 (2.86%)|
|Nongerminomatous germ cell tumor||5 (14.28%)|
|Radiation field and volume|
|Focal radiation||3 (3993 cGy)|
|Ventricular radiation||3 (2810 cGy)|
|Whole brain/craniospinal||3 (2883 cGy/)|
|Focal radiation||4 (2660 cGy)|
|Ventricular radiation||4 (2352 cGy)|
|Whole brain/craniospinal||3 (3150 cGy/)|
|Ventricular radiation||9 (2960 cGy)|
|Whole brain/craniospinal||1 (2580 cGy/)|
|Gross total||3 (8.6%)|
|Treatment with external ventricular transient drain or ventriculo-peritoneal shunt for hydrocephalus||11 (31%)|
No differences in age at diagnosis or time since diagnosis was evident as a function of tumor location or radiation field Fs > 1.89, P > .10. Further, the frequency of patients requiring a shunt to treat hydrocephalus was no different between patients with pineal, suprasellar, or bifocal tumors, chi-square s (2, n = 31) = 2.94, P > .10, nor did it vary as a function of radiation field chi-square s (1, n = 31) = 2.48, P > .10. Finally, we examined differences in the size of the tumor (calculated by multiplying the 2 largest measures of diameter on magnetic resonance imaging reported by the neuroradiologist) for the 24 patients for which this information was available: pineal tumors were largest (n = 8, 12.08 cm2, standard deviation [SD] = 13.77), followed by suprasellar (n = 8, 8.77 cm2, SD = 9.50) and bifocal tumors (n = 8, 5.25 cm2, SD = 4.4), although group differences were not statistically significant.
Materials and Procedures
The patients in this series were accrued by clinical referral and were seen for 1 or more neuropsychological assessments (range, 1-3 assessments per individual). Because not all participants were administered all the tests at each assessment point (because of age or time constraints), the number of patients assessed with each measure varies across time. Seventeen patients were seen for a single assessment, 12 had 2 assessments, and 6 had 3 assessments. Measures of intellectual functioning, 31-33 academic achievement, 34, 35 receptive vocabulary, 36 visual-motor function, 37 and memory 38, 39 were most consistently administered and hence are reported here (Table 2). Standard scores were recorded for each test (based on age-related means and SDs from test standardization norms), with a mean of 100 and a SD of 15. For intelligence tests, full scale, verbal comprehension, perceptual reasoning, freedom from distractibility/working memory, and processing speed indices are reported. For academic achievement and memory, different tests were used depending on the year of evaluation or the age of the patient. We created reading, spelling, and arithmetic composite variables by entering the equivalent academic indices and verbal and visual memory composite variables by entering the equivalent visual and verbal memory indices for each patient (depending on which test was administered). This study was approved by the Hospital for Sick Children's Research Ethics Board before the data were extracted from clinical records.
|Domain||No. of Patients||Timea||Estimate||SE||Estimate||SE|
First, the rate of change in standardized test scores over time was determined using mixed-model regression. Because patients were evaluated at various times after diagnosis, yielding unbalanced and missing data, the mixed-model approach was used. Mixed-model techniques were implemented using the PROC MIXED (SAS Institute, Cary, NC). Results were considered significant at P < .05.
To evaluate the impact of medical and demographic variables, mean standard scores for each measure were calculated at a single time point for only those patients seen 1.5 years or longer from diagnosis. This ensured that all patients included in these analyses were seen after the completion of therapy. For patients seen more than once, the assessment conducted after the greatest amount of time since diagnosis was used. The effects of radiation field (ventricular vs whole brain; because follow-up data were limited for those patients treated with focal radiation [n = 3], they were excluded from these analyses), insertion of a shunt/external ventricular drain (EVD) (presence vs absence) to treat hydrocephalus, and surgical resection (yes vs no) on cognitive outcome were examined using 1-way analysis of variance (ANOVA). As well, correlation coefficients were calculated for age at diagnosis and each outcome measure. Finally, patients seen for multiple evaluations may have been most impaired initially and therefore were seen more frequently, leading to referral bias. Consequently, mean scores for each measure were compared for patients seen for a single assessment versus those seen for serial assessment using 1-way ANOVA.
Separate linear models were generated to assess change for (1) the entire sample and (2) when the sample was grouped by tumor location (suprasellar, pineal, and bifocal only). (Measures of receptive language and visual-motor function were not included in the tumor site analyses because the sample size within each cell was very small, leading to unstable models.) Intercepts represent the estimated baseline functioning, and slopes characterize change in functioning over time (Table 2 for the whole sample model; Tables 3 and 4 for the analyses including tumor location).
|Neurocognitive Measure||Main Effects and Interactions||df||F Statistic||P|
|Wechsler Intelligence Scales|
|Perceptual Reasoning Index||Tumor location||2, 19||4.89||.02|
|Working Memory Index||Time||1, 18||11.64||.01|
|Processing Speed Index||Time||1, 19||9.68||.01|
|Tumor location||2, 19||3.81||.04|
|Visual immediate||Time||1, 17||7.31||.02|
|Visual delay||Time||1, 17||12.66||.00|
|Tumor site||2, 17||3.56||.05|
|Measure||Standard Risk||High Risk|
|Intelligence (WISC-III, WISC-IV, WAIS-III)|
|Verbal comprehension index||101||92|
|Perceptual reasoning index||114.89||97|
|Working memory index||99.3||85.78|
|Processing speed index||101.78||87|
|Achievement (WIAT-II, WRAT-3)|
|Receptive language (PPVT-III)|
|Peabody picture vocabulary test||103.67||84.13|
|Beery visual-motor integration test||103.67||84.13|
|Memory (CMS, WMS-III)|
Full-scale, verbal comprehension, and perceptual reasoning IQ scores fell within the average range across the entire sample and did not change significantly over time. However, working memory and processing speed declined by approximately 1.5 points per year, P < .01 (Table 2; Fig. 1A). When tumor location was considered, patients with a pineal tumor exhibited perceptual reasoning scores approximately 1 standard deviation below the normative mean and significantly lower than patients with bifocal tumors, F (2, 19) = 6.28, P = .002 (Tables 3 and 4). Processing speed scores for patients with pineal and suprasellar tumors were approximately 2/3 SDs below the normative mean and significantly lower than patients with bifocal tumors, Fs (2, 19) < 4.17, Ps < .05 (Table 4). Examination of group intercepts and slopes revealed that a decline from average scores in processing speed was observed for the suprasellar group, P < .01, whereas the pineal group demonstrated poor performance immediately after diagnosis and treatment, which was then stable over time (Fig. 2A; Table 4). The bifocal group also showed decline in processing speed scores, P < .05, although they started from a higher baseline level. Declines in working memory were evident for the suprasellar and bifocal groups, P < .05, with the pineal group again displaying initial poor performance that remained stable over time (Table 4).
Across the entire sample, receptive vocabulary, reading, spelling, and mathematics fell within the average range and did not change over the modeled time period. A trend toward a decline in visual-motor functioning was present but was not statistically significant (Table 2). No effects of tumor site were observed for academic measures.
A significant decline in the immediate and delayed visual memory indices of approximately 2 points per year was observed across the entire sample (Table 2, Fig. 1B). Mean delayed visual memory for patients with pineal tumors was approximately 1 standard deviation below the normative mean and significantly lower than for patients with bifocal tumors F (2, 17) = 3.88, P = .02 (Tables 3 and 4). Examination of intercepts and slopes revealed that declines from average scores in immediate and delayed visual memory were observed for the suprasellar group, P < .01, whereas the pineal group demonstrated poor performance at diagnosis, which remained stable over time (Table 4; Fig. 2B).
For immediate and delayed verbal memory measures, no changes were evident over time across the entire sample (Table 2). Notably, the intercept for delayed verbal memory was 1 SD below the normative mean.
Demographic and Medical Characteristics
At long-term follow-up (mean, 5.90 years, SD = 3.88, n = 19), treatment with surgery did not have an impact on neurocognitive outcome (P > .10). Patients who required a shunt/EVD to treat hydrocephalus had lower immediate visual memory scores than those who did not require this intervention (74.83 vs 97.56), P < .05. Patients treated with CSI radiation displayed poorer scores than those treated with a ventricular volume for verbal comprehension (87.30 vs 110.50), perceptual reasoning (98.56 vs 121.00), and full-scale IQ (91.78 vs 115.17), reading (85.38 vs 106.17) and receptive vocabulary (85.56 vs 106.20), P < .05. Younger age at diagnosis was associated with poorer verbal comprehension, perceptual reasoning, full-scale IQ, reading, receptive vocabulary, and immediate verbal memory scores (rs = .77, .50, .63 .54, .79, and .49, respectively, P < .05). No differences were evident on any measures for patients seen for a single assessment versus those who were subsequently seen for follow-up assessment (P > .10).
We conducted further qualitative appraisal of the additive effects of variables associated with higher risk of adverse late effects. Variables including female gender, younger age at diagnosis, hydrocephalus, treatment with whole-brain radiation, and higher doses of radiation have all been associated with poorer neurocognitive outcome in children with CNS tumors. For each of these relevant variables, we used a dichotomous coding scheme to characterize level of risk (0 = standard, 1 = high) in each patient, including gender (male = 0, female = 1), presence of hydrocephalus (no = 0, yes = 1), radiation field (focal = 0, cranial-spinal/whole brain = 1), radiation dose (<3000 cGy = 0, >3000 cGy) and age (older than the median age of the sample = 0, younger than the median age of the sample = 1). The frequency of variables that were coded as high risk was then calculated for each patient, and patients were divided into 2 groups: patients who were high risk on 2 or fewer variables (Low) and patients who were high risk on 3 or more variables (High). Mean standard scores across the tests were calculated (Table 5). Based on qualitative inspection of the means, the high-risk group had poorer standard scores than the low-risk group across all tests with the exception of visual memory.
To our knowledge, our study is the first long-term serial evaluation of multiple neurocognitive functions in patients treated for CNS GCT. We report a number of novel findings.
We documented that intellectual, academic functioning, visual-motor ability, and receptive vocabulary were generally intact and remained stable in our sample of patients after diagnosis and treatment for CNS GCT. These findings are consistent with previous studies that show preserved intellectual ability and academic function in patients treated for CNS GCT 7, 26 but for the first time show that preserved ability is stable over a long time. The relatively good outcome for patients with CNS GCT may due to their older age at diagnosis. The mean age at diagnosis for our sample was 11.66 years. Further, we found that younger age at diagnosis predicted poorer performance across measures. Notably, when we examined the impact of radiation field at a single time point (mean, 5.90 years), we found that patients treated with CSI performed poorly, relative to those treated with ventricular volumes on measures of intellectual function, reading, and receptive vocabulary. Hence, the smaller fields of radiation used to treat many of our patients likely resulted in enhanced cognitive outcome relative to patients with posterior fossa tumors. These novel results suggest that attempts at reducing radiation volumes in patients with CNS GCT and in particular germinoma patients are supported by long-term differences cognitive outcomes.
We did identify declines in working memory, information processing speed, immediate visual memory, and delayed visual memory (−1.61, −1.66, −1.80, and −2.21 points per year, respectively) after treatment over the modelled time period of 16 years. Further, performance on tests of immediate and delayed verbal memory was below average at baseline, suggesting early and stable verbal memory deficits.
A further novel finding was that cognitive function was associated with tumor site. Perceptual reasoning for patients with pineal tumors was poor and significantly lower than patients with bifocal tumors. Previous reports have documented compromised perceptual reasoning, although location was not accounted for. 9, 11, 29 Our study provides a more detailed extension of these findings. Furthermore, patients with pineal tumors displayed early and stable deficits in working memory, information processing speed, and visual memory. These findings support the hypotheses that the impact of a tumor within the pineal region disrupts neural networks important for these specific cognitive functions. This disruption may take the form of infiltration into healthy tissue by the tumor or compression and subsequent damage of healthy tissue from mass effect. 40, 41 Notably, the pineal gland lies within the thalamic-epithalamic region that has been demonstrated to be important in working memory 42 and declarative memory, 43-46 likely through thalamo-frontal and thalamo-hippocampal circuits. 40, 41 Deficits in working memory have been reported in children treated for thalamic/epithalamic brain tumors. 40, 41
A different trajectory was evident for patients with suprasellar and bifocal tumors, as estimates of their baseline functioning (intercepts) were within the average range for working memory, information processing speed, and visual memory. Significant declines in working memory, information processing speed functioning, and visual memory were observed over time, however. These emerging deficits may reflect the impact of the treatment with cranial radiation on long-term functioning, 20, 47-49 rather than the initial impact of the tumor on local cognitive networks. The relatively better initial functioning of these patients, particularly those with bifocal tumors, may be related to the smaller tumor size compared with patients with pineal tumors.
We note some common limitations of clinical research that are present in this study. Our data were not gathered prospectively and hence our study is limited by the format of a retrospective review. Because the sample was accrued clinically over an extended time span, some tests were revised and therefore different versions were used across assessments for some individuals. This may decrease the reliability of some of the scores recorded across assessments. We did not evaluate behavioral/emotional functioning, and we recognize the potential confounding influence of this on our results. Our sample was not large enough to evaluate the interaction between tumor location and radiation field in relation to cognitive outcome. The size of our patient sample also precluded the statistical examination of the interaction between variables that are independently associated with poor neurocognitive outcome. Further, for our analyses of long-term follow-up, we recognize that there is the increasing possibility of more confounding variables influencing our findings. Finally, the generalizability of this study sample may be an issue. It may be that some children were more impaired initially, so they were seen more frequently for neuropsychological evaluation; hence, there may be referral bias in our sample.
General cognitive abilities appear stable and intact after treatment for most children with CNS GCT, but a significant decline over time in working memory, processing speed, and visual memory is evident. Furthermore, tumor location appears important in understanding the trajectory of stability and decline in CNS GCT patients. Finally, patients treated with ventricular versus whole-brain radiation displayed better outcome. These findings justify further collection of neurocognitive data in prospective trials for patients with CNS germ cell tumors.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.
- 10Pineal region tumors. In: Handbook of Neurosurgery. 5th ed. New York, NY: Theime; 2001: 455-457..
- 28CNS germinoma: the effects of central tumors on neuropsychological function. In: Morgan JE, Baron IS, Ricker JH, eds. Casebook of Clinical Neuropsychology. New York, NY: Oxford University Press; 2001.
- 30Malignant GNS germ cell tumors: interim analysis after 5 years of SIOPCNS GCT 96. Med Pediatr Oncol. 2002; 39: 227., , , et al.
- 31Manual for the Wechsler Intelligence Scale for Children. 3rd ed. San Antonio, TX: The Psychological Corporation; 1991..
- 32The Wechsler Intelligence Scale for Children. 4th ed. New York, NY: The Psychological Corporation; 2003..
- 33The Wechsler Adult Intelligence Scale. 3rd ed. New York, NY: The Psychological Corporation; 1997..
- 34The Wechsler Individual Achievement Tests. 2nd ed. San Antonio, TX: The Psychological Corporation; 2002..
- 35Wide Range Achievement Test. 3rd ed. Wilmington, DE: Jastak Associates; 1993..
- 36Peabody Picture Vocabulary Test-III. Circle Pines, MN: American Guidance Service; 1997., .
- 37The Visual-Motor Integration Test. 4th ed. Administration, scoring and teaching manual. Austin, TX: Pro-Ed; 1997..
- 38The Wechsler Memory Scale. 3rd ed. San Antonio, TX: The Psychological Corporation; 1997..
- 39Children's Memory Scale. San Antonio, TX: The Psychological Corporation; 1997..