- © 2009 by American Society of Clinical Oncology
Cognitive Deficits and Predictors 3 Years After Diagnosis of a Pilocytic Astrocytoma in Childhood
- Femke K. Aarsen⇑,
- Philippe F. Paquier,
- Willem-Frans Arts,
- Marie-Lise Van Veelen,
- Erna Michiels,
- Maarten Lequin and
- Coriene E. Catsman-Berrevoets
- From the Departments of Pediatric Neurology, Pediatric Neurosurgery, Pediatric Oncology and Hematology, and Pediatric Radiology of Erasmus Medical Center–Sophia Children's Hospital, Rotterdam, the Netherlands; Department of Neurology, University Hospital Erasme; Department of Linguistics, Free University of Brussels, Brussels; and the Unit of Neurosciences, School of Medicine, University of Antwerp, Antwerp, Belgium.
- Corresponding author: Femke K. Aarsen, MSc, Department of Pediatric Neurology, Erasmus Medical Center–Sophia Children's Hospital, PO Box 2060, 3000 CB Rotterdam, the Netherlands; e-mail: f.aarsen{at}erasmusmc.nl.
Abstract
Purpose To prospectively study cognitive deficits and predictors 3 years after diagnosis in a large series of pediatric patients treated for pilocytic astrocytoma (PA).
Patients and Methods Sixty-one of 67 children were grouped according to infratentorial, supratentorial midline, and supratentorial hemispheric site. Intelligence, memory, attention, language, visual-spatial, and executive functions were assessed. Included predictors were sex, age, relapse, diagnosis-assessment interval, hydrocephalus, kind of treatment, and tumor variables.
Results All children with PA had problems with sustained attention and speed. In the infratentorial group, there also were deficits in verbal intelligence, visual-spatial memory, executive functioning, and naming. Verbal intelligence and verbal memory problems occurred in the brainstem tumor group. The supratentorial hemispheric tumor group had additional problems with selective attention and executive functioning, and the supratentorial midline tumor group displayed no extra impairments. More specifically, the dorsal supratentorial midline tumor group displayed problems with language and verbal memory. Predictors for lower cognitive functioning were hydrocephalus, radiotherapy, residual tumor size, and age; predictors for better functioning were chemotherapy or treatment of hydrocephalus. Almost 60% of children had problems with academic achievement, for which risk factors were relapse and younger age at diagnosis.
Conclusion Despite normal intelligence at long-term follow-up, children treated for PA display invalidating cognitive impairments. Adequate treatment of hydrocephalus is important for a more favorable long-term cognitive outcome. Even children without initial severe deficits may develop cognitive impairments years after diagnosis, partly because of the phenomenon of growing into deficit, which has devastating implications for academic achievement and quality of life (QOL).
INTRODUCTION
Astrocytomas account for approximately one third of pediatric central nervous system tumors.1 They are classified according to increasing malignancy grade as pilocytic, fibrillary, and anaplastic astrocytomas. The first-choice treatment for pilocytic astrocytoma (PA) is complete surgical resection. Adjuvant chemotherapy or radiotherapy is reserved for recurrent or progressive, inoperable disease. The long-term survival for children with PA is approximately 90% after 4 years, whereas survival for patients in individual series of cerebellar PA with gross total resection may be 90% to 100%.2
Long-term studies on cognitive deficits in children after treatment of cerebellar PA describe problems with affect regulation, memory, attention, language, speech, visual-spatial, and executive functions in different combinations and to different degrees.3–5 These deficits fit in the spectrum of the cerebellar cognitive affective syndrome.6
Three studies describe long-term cognitive deficits after treatment of PA in the cerebral hemisphere.7–9 In the study of Yule et al,7 two children with cerebral hemisphere PA had subnormal intelligence quotients (IQs) of 87 at 1 year after treatment. After treatment for temporal lobe astrocytomas, five of seven children were reported to have intellectual deterioration, learning disabilities, or psychopathology.8 The majority of 14 children with temporal lobe tumors had an increased risk for memory dysfunction, academic failure, and internalized or externalized behavior problems 3 years after treatment.9
Conclusions in three studies on cognitive sequelae of treatment of a supratentorial midline PA are ambiguous. In one study, children with optic pathway tumors had normal intelligence, memory, and fluency 3 years after chemotherapy, with the exception of children who had neurofibromatosis type 1 (NF1) and children who received radiotherapy.10 In two other studies of children with hypothalamic and chiasmatic tumors, intelligence was significantly compromised at diagnosis,11 remained below average after 6 years,11 and frequently required special education for the children after 3 years.12
Long-term cognitive deficits contribute to impairments that have an impact on functional outcome, such as school achievement and quality of life (QOL). In a consecutive series of 38 children with low-grade astrocytoma, functional outcome was significantly decreased in all domains of the QOL questionnaire except for emotions.13 Predictors for worse functional outcome were surgical damage,3–5,14 hydrocephalus at diagnosis,3 radiotherapy,7 age,13 or relapse.13
The purpose of this study was to prospectively assess cognitive deficits and predictors 3 years after diagnosis in a large consecutive series of pediatric patients treated for PA in different parts of the brain.
PATIENTS AND METHODS
Patients
Between 1994 and 2004, a consecutive series of 67 children without NF1 were treated for PA at Erasmus MC–Sophia Children's hospital. Children with NF1 were excluded because of their specific clusters of deviant behavioral and cognitive impairments.15 Sixty-one patients (including 28 boys) could be tested according to our standard clinical follow-up protocol. Two patients with diencephalic tumors died 6 and 12 months, respectively, after diagnosis, and one patient refused neuropsychologic assessment. Three children were lost to follow-up. Eighteen children and their parents filled in a QOL questionnaire at the time of assessment. Results were described in an earlier study.13
Methods
All children had a standardized pediatric neuro-oncologic follow-up every 6 months and a standardized neuropsychologic assessment at 6 months, 18 to 24 months, and 3 to 4 years after diagnosis in a multidisciplinary outpatient clinic. The data of the last assessment were used, and the interval between two assessments was at least 21 months. Preoperative magnetic resonance imaging (MRI) or computed tomography (CT) scans with and without contrast enhancement were reviewed for ventricle width, presence of ventriculoperitoneal (VP) shunt, ventriculocisternostomy, and the preoperative size and site of the tumor. Postoperative MRI scans at time of assessment were used to determine the ventricle width, resection site, VP shunt, and presence or size of residual tumor at time of neuropsychologic assessment. According to tumor site, three groups were defined: an infratentorial (brainstem and cerebellum), a supratentorial hemispheric, and a supratentorial midline group (ventral and dorsal). The ventral group comprised children with a tumor in the chiasm or adjacent hypothalamus, and the dorsal group comprised children with a tumor in the thalamus, the basal ganglia, or both. The group that comprised left-sided tumors included left-sided supratentorial hemispheric tumors, left-sided midline tumors, and—because of the crossed connections between the cerebral hemispheres and the contralateral cerebellum—right-sided infratentorial tumors. The group that comprised right-sided tumors included right-sided supratentorial hemispheric, right-sided midline, and left-sided infratentorial tumors.
We computed the bicaudate index (BI) as a measure of ventricular dilation at presentation and at time of assessment.3 We distinguished the following hydrocephalus categories: no (BI < 0.19), mild (BI, 0.19 to 0.26), and severe hydrocephalus (BI > 0.26).
A pediatric neurologist (C.C-B.) examined all children to establish residual neurologic impairments and to determine academic achievement at time of neuropsychologic assessment. The neuropsychologic assessment included intellectual functioning with the age-appropriate Dutch version of the Wechsler Scales (CF) (ie, Wechsler Preschool and Primary Scale of Intelligence-Revised [WPPSI-R] or Wechsler Intelligence Scales for Children-Revised [WISC-R]),16 memory,17–20 attention,17,20 language,17,21,22 visual-spatial17,19–20,23 and executive skills17,19,20 (Table 1). QOL was assessed with the parent form of TNO/AZL Children's Quality of Life Parent Form questionnaires for children age 6 to 15 years (TACQOL-PF)24 and the children's form for children age 8 to 15 years (TACQOL-CF).24
Neuropsychologic Outcome in IQ and z Scores for Total, Infratentorial, Supratentorial Midline, and Supratentorial Hemispheric Groups
Statistics
The data of the neuropsychologic assessment were compared with the normative data of the Dutch population and were corrected for age. To compare the performances of all children, all scores were converted into z scores. The χ2 test was used to compare variance of this study group with the normal population. The Kolmogorov-Smirnov test was used to control for normal distribution. Data were analyzed with the Mann-Whitney U test, the Kruskal-Wallis test, and Spearman rank correlations. Data that fulfilled the requirements of parametric testing and had more than 20 entries were analyzed with the two-sided t test, analysis of variance, univariate analysis, and multivariate analysis with Bonferroni's corrections. A two-sided Pearson correlation matrix was computed. In a normal population, 2.3% of children have z scores less than or equal to −2. We considered it clinically important when the percentage of children in this study obtained a z score that exceeded 2.3%. Backward, stepwise, linear regression on the basis of the highest partial correlations obtained in a Pearson correlation matrix was used to identify parameters that predict significantly lower scores on the neuropsychologic tests in the entire PA group. Included risk factors were sex, age at diagnosis or assessment, diagnosis-assessment interval, relapse, kind of treatment, size and site of tumor/lesion, BI at diagnosis or assessment, and presence of VP drain. The maximum number of predictors was set as three because of the sample size. Univariate and multivariate analyses were performed after regression analysis (by using SPSS 16.0; SPSS, Chicago, IL).
RESULTS
Neurologic Data
Median age at diagnosis was 7 years, 8 months (interquartile range, 3.2 to 11.4 years), and median age at assessment was 11 years, 3 months (interquartile range, 8.9 to 15.1 years). The median follow-up period was 3 years, 6 months (interquartile range, 2.0 to 5.0 years). In all children, neuropathologic diagnosis was confirmed by means of biopsy (n = 9) or partial resection (n = 52). Fifteen of 61 children needed other therapy (Fig 1), which included chemotherapy (n = 11) and stereotactic radiotherapy (n = 4). Sixteen children experienced one or more relapses, and treatment of these consisted of re-resection (n = 11), adjuvant chemotherapy (n = 9), or adjuvant radiotherapy (n = 10). The chemotherapy scheme consisted of vincristine and carboplatin according to the International Society of Pediatric Oncology–Low Grade Glioma protocol.
Treatment characteristics after initial diagnosis.
Table 2 lists the patient characteristics and the neurologic impairments at the time of assessment. Thirty-five children had infratentorial tumors (brainstem, n = 6; cerebellum, n = 29). Of 26 children with supratentorial tumors, six had hemispheric tumors, and 20 had midline tumors (ventral, n = 10; dorsal, n = 10). Thirty-four children had no, 11 had mild, and 16 had severe hydrocephalus at first presentation. In five children, a VP shunt was inserted before tumor resection, and a VP shunt was inserted in 18 children after tumor resection. One child was treated with ventriculocisternostomy. At time of assessment, five children still had a mild hydrocephalus.
Clinical Patient Characteristics
Education and QOL
Education results showed that 59% of the children needed special education services (special education, 21%; remedial teaching [RT]) by a specialized Dutch institution in a normal school, 38%). These children had significantly lower scores on tests that measured sustained attention (cancellation test–fluctuations, P = .02), speed (cancellation test–speed, P = .02), long-term memory (Rey Complex Figure Test [CFT]–delayed recall, P = .01; Rey Auditory-Verbal Learning Test [RAVLT]–delayed recall, P = .02), executive functioning (Trailmaking Test [TMT] part B, P = .02) and QOL (TACQOL-CF motor domain, P = .03; cognition domain, P = .04; and TACQOL-PF cognition domain, P = .01; autonomy domain, P = .05) than children without RT or special education. Neurologic independent factors that influenced type of education were relapse or a younger age at diagnosis.
Neuropsychologic Data
Six children (three with cerebellar and three with supratentorial midline PAs) were younger than 6 years of age at the time of assessment. They were too young to perform the tests to measure language, attention, memory, executive, and visual-spatial skills, with exception of verbal fluency and the Beery Visual-Motor Integration test.
Compared with the normal population (Table 1), the entire PA group showed weaker performances on tests that measured sustained attention (cancellation test–fluctuations, P < .001), speed (cancellation test–speed, P < .001), executive functioning (TMT A and B, P < .001; Wisconsin Card Sorting Test: perseverations, P < .001), naming (Boston Naming Test [BNT], P < .02), and long-term visual-spatial memory (Rey CFT–delayed recall, P < .001). Forty-four percent of children showed a discrepancy between performance (total performance IQ [TPIQ]) and verbal intelligence (total verbal IQ [TVIQ]). This is significantly more than in the normal population (P < .001). In 56% of children, TVIQ was lower than TPIQ.
Factors That Influenced Cognitive Outcome
Regression analysis was performed on the significantly weaker performances of these tests, and it showed that there were no significant predictors for impairments of speed (cancellation test–speed) and nonverbal long-term memory (Rey CFT–delayed recall). Independently related predictors that resulted in better scores were presence of VP shunt, chemotherapy, or larger tumor residue in the group with supratentorial hemispheric tumors. Negative predictors were radiotherapy, younger age at diagnosis, older age at assessment, higher BI at diagnosis or assessment, and left-sided tumor (Table 3).
Risk Factors of the Cognitive Profile of the Total Group
The same risk factors also were negatively influencing other cognitive domains: radiotherapy (selective attention: Stroop Color-Word test, P < .03), younger age at diagnosis (performance intelligence: TPIQ, P < .05), older age at assessment (visual-spatial skills: Rey CFT–copy, P < .05), higher BI at assessment (language comprehension: Tokentest, P < .001), left-sided tumors (language comprehension: Tokentest, P < .001), and right-sided tumors (performance intelligence: TPIQ, P < .001; visual-spatial skills: Beery Visual-Motor Integration, P = .02). In addition to these factors, a longer interval between diagnosis and assessment negatively influenced results on tests of language reception (Tokentest, P < .04), selective attention (Stroop Color-Word test, P < .04), and verbal fluency (P < .03). An interaction effect was found between diagnosis-assessment interval and radiotherapy on selective attention (P < .01).
Compared with the normal population the infratentorial group had significantly lower scores on verbal intelligence (TVIQ, P < .02), sustained attention (cancellation test–fluctuations, P < .001), speed (cancellation test–speed, P < .001), long-term visual-spatial memory (Rey CFT–delayed recall, P < .001), executive functioning (TMT A, P < .001; TMT B, P < .01; Wisconsin Card Sorting Test perseverations, P < .001), and naming (BNT, P < .01). Further sub analysis showed that children with a cerebellar PA had significantly low scores on all these tests with exception of verbal intelligence in comparison to the norms. Children with a brainstem PA had problems with sustained attention (Cancellation test–fluctuations, P < .001), verbal long-term memory (RAVLT–delayed recall, P < .01), and naming (BNT, P < .02) compared with the norms. The z scores of −2 or less were found in the domains of speed (cancellation test–speed; 60%), verbal intelligence (TVIQ < 70; 40%), and verbal short-term memory (RAVLT, 1 to 5; 17%).
The supratentorial hemispheric group scored significantly lower on tests that measured selective (Stroop Color-Word test, P < .02) and sustained (cancellation test–fluctuations, P < .02) attention compared with the norms. The z scores of −2 or less were found in the domains of speed (cancellation test–speed; 50%), and executive functioning (TMT B; 33%).
The supratentorial midline group performed significantly weaker on speed and sustained attention (cancellation test–fluctuations and speed, respectively, both P < .01) compared with the norms. Children who had dorsal tumors scored significant lower on language comprehension (Tokentest, P < .04) compared with the norms, and 25% had z scores of −2 or less in verbal short-term memory (RAVLT, 1 to 5).
DISCUSSION
Cognitive deficits are well described in children with a cerebellar PA but not in those with a PA in other areas of the brain. This study shows that problems with sustained attention and speed are present in all children with PA. This implies that these cognitive deficits are independent of the site of the tumor, and it reflect a more global brain dysfunction.17 Acquired brain injury in children often has diffuse effects that result in more basal cognitive function problems.24,25
In addition to speed and attention impairments, a distinct cognitive profile is found in each defined tumor group. In agreement with our findings in an earlier and smaller series3 in the infratentorial cerebellar tumor group, we find deficits in language, visual-spatial memory, and executive functioning. In another previous study, we described long-term social and behavioral problems.13 The cluster of disturbances of executive function, impaired spatial cognition, linguistic difficulties, and personality changes is often called the cerebellar cognitive affective syndrome.3–7,14 This is thought to be caused by disruption of neural circuits that connect the cerebellum with the prefrontal, posterior parietal, superior temporal, and limbic areas.6
Traditionally, the brainstem is not regarded as a structure that mediates cognitive or behavioral functions. In the brainstem group, we recorded deficits in verbal intelligence, verbal memory, naming, and—in a previous study13—behavioral problems. Cognitive deficits restricted to the language domain have been described in only one other study in children with brainstem tumors.26 No significant language disturbances were found in this study, but two children had mild impairments in lexical generation or phonologic awareness. In a study that compared adult patients with mild and severe olivo-ponto-cerebellar atrophy, patients with severe atrophy had significantly lower scores on verbal memory (RAVLT).27 In children with pontine glioma, behavioral changes, such as separation anxiety, school phobia, pathologic laughter, and aggression, have been described.28 A possible explanation could be disruption of the reciprocal cerebello-ponto-cerebral circuitry that results in an interruption of fibers that modulate behavior and cognition.27,28
In the group with supratentorial hemispheric tumors, we found specific problems with selective attention and executive functioning. This is in contrast to earlier-described general intellectual deterioration and memory dysfunction in children with a hemisphere PA.7–9 All children in the supratentorial group of this study had normal intelligence, and only one child had memory dysfunction as a result of a resection of the hippocampus. Sparing of global intelligence may be explained by the lack of radiotherapy and well-controlled epilepsy of children in the supratentorial hemispheric group of this study. These factors may lead to intellectual decline and specific memory and attention problems.29,30
In the dorsal supratentorial midline tumor group, we found specific impairments in the domains of language comprehension and verbal short-term memory. The language and memory problems were recorded in five children with left thalamic tumor localization, and we did not observe apraxia or neglect. These findings are in agreement with the problems of attention, memory, executive, and visuospatial functions and with the aphasia, transient neglect, and transient apraxia in a small series of five children who underwent resection for left thalamic tumors.31
The statistical analysis showed that severity of the ventricle dilation at times of diagnosis and assessment is a risk factor for the development of deficits of attention, language comprehension, and executive functioning. Treatment of ventricle dilation with a VP drain resulted in better scores. This is in agreement with our earlier study, in which we computed a correlation between cognitive functioning and preoperative ventricle dilation in children who underwent operation for a cerebellar PA.3 The findings in this study seem to suggest that treatment of hydrocephalus, even if ventricle width is not progressing, could prevent long-term cognitive deficits after brain tumor treatment, but additional research is necessary.
Chemotherapy as treatment for PA results in better executive functioning. The effect of radiotherapy versus chemotherapy on cognitive functioning is well known.29 The more favorable effect of chemotherapy versus surgery may be explained by restriction of surgery to a diagnostic biopsy or to a relatively small resection in children treated with chemotherapy. This neurosurgical procedure limits the traumatic damage compared with a more extensive resection.
This study demonstrates that there are three inter-related age effects: age at diagnosis, interval between diagnosis and assessment, and age at assessment.32 A younger age at diagnosis as a risk factor for more severe impairment of cognitive functions and academic achievement is also described in studies on survivors of childhood cancer33 and in children with traumatic brain injury.32 An explanation of this phenomenon could be the vulnerability theory33: children with a diagnosis and treatment at a younger age have to learn new skills with already defective basal functions. For the other age effects, there could be a biologic34 or a behavioral explanation.32 In children who received radiotherapy, the late effects of radiotherapy on processes that interfere with neuronal development lead to cognitive decline,34 and this is supported by the statistical finding of an interaction effect between age and radiotherapy on attention. However, in children without radiotherapy, cognitive problems become apparent years after diagnosis of PA. This finding may be explained by the theory that “early brain damage may have a cumulative effect on ongoing development, with increasing deficits emerging through childhood as more functions are expected to mature and need to be subsumed within the undamaged tissues.”35(p107) This behavioral phenomenon of growing into deficits has already been described in a study of children with PA and their functional outcomes.13
Educational data show a high percentage (59%) of children after PA treatment that needed special education services (special education, 21%; RT, 38%) compared with national averages of 5% (special education)36 and 15% (RT)37 in the Netherlands. Children who experience relapse and those who have a younger age at diagnosis are especially more at risk for educational problems. This is in agreement with the study of long-term survivors of brain tumors, in which 24% to 70% received special education services, depending on younger age of diagnosis.33 In addition to educational problems, these children also have a lower QOL than children who receive normal education. This strongly suggests that cognitive deficits in children after treatment of PA have clinical implications not only on their school careers but also on their subjective perceptions of daily QOL.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: Femke K. Aarsen, Willem-Frans Arts, Coriene E. Catsman-Berrevoets
Administrative support: Femke K. Aarsen, Coriene E. Catsman-Berrevoets
Provision of study materials or patients: Femke K. Aarsen, Marie-Lise Van Veelen, Erna Michiels, Coriene E. Catsman-Berrevoets
Collection and assembly of data: Femke K. Aarsen, Maarten Lequin, Coriene E. Catsman-Berrevoets
Data analysis and interpretation: Femke K. Aarsen, Philippe F. Paquier, Willem-Frans Arts, Marie-Lise Van Veelen, Maarten Lequin, Coriene E. Catsman-Berrevoets
Manuscript writing: Femke K. Aarsen, Philippe F. Paquier, Willem-Frans Arts, Marie-Lise Van Veelen, Erna Michiels, Maarten Lequin, Coriene E. Catsman-Berrevoets
Final approval of manuscript: Femke K. Aarsen, Philippe F. Paquier, Willem-Frans Arts, Marie-Lise Van Veelen, Erna Michiels, Maarten Lequin, Coriene E. Catsman-Berrevoets
Footnotes
-
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
- Received August 13, 2008.
- Accepted January 27, 2009.
