Cognitive sequelae in children with posterior fossa tumors

Introduction 

Posterior fossa tumors represent two thirds of all pediatric brain tumors [1]. Early diagnosis and treatment have led to an improved survival rate in children with malignant posterior fossa tumors. With improved survival rates, attention has shifted to the long-term effects of adjuvant therapy (i.e., radiotherapy). A growing body of evidence has demonstrated that impaired intellectual functioning can be frequently observed among survivors and that this impairment is progressive [2], [3], [4], [5]. Variables identified as important in delineating the relationship between radiotherapy and lower IQ scores include early age at tumor diagnosis [6], [7], radiation dosage [8], [9], and time elapsed since treatment [4], [7]. Although the long-term effects of whole brain irradiation have been suggested as a major cause of cognitive impairment, focal irradiation (i.e., posterior fossa region) has also been associated with cognitive dysfunction [10], [11].

The reliance on Full Scale IQ ratings as a global index of cognitive outcome is of limited utility in predicting the impact of radiotherapy on learning. An assessment of memory functioning would include an examination of a child’s ability to acquire, store, and retrieve information skills that are all necessary for academic success. Given that focal irradiation to the posterior fossa region includes not only the brainstem and cerebellum but also parts of the temporal lobes (the latter believed to be one of the neuroanatomic substrates of memory) an impairment in memory functioning could have a significant and deleterious impact on a child’s academic achievement. Some researchers have demonstrated moderate to severe deficits in visual memory after whole brain or focal irradiation [12], [13]. Research on memory functioning has been limited by the unavailability, until recently, of a comprehensive standardized measure of verbal and visual memory functioning for children. The Wide Range Assessment of Memory and Learning [14] has only been available since 1990.

The aim of this retrospective review was thus twofold: to examine verbal and nonverbal memory functioning using a norm-referenced tool, standardized for children 5-17 years of age, and to confirm previous research findings with regard to impaired intellectual functioning after cranial irradiation. The current sample was comprised of children treated for medulloblastoma and cerebellar astrocytoma (Grades II and III) in a pediatric hospital in Ottawa, Canada.

Patients and methods 

For inclusion in this review all children must have completed a course of radiotherapy after surgery. Of the children also treated with chemotherapy, only those who did not receive methotrexate were included. When used in conjunction with radiotherapy, methotrexate has been associated with substantial neurologic and neurocognitive impairment [7], [15]. In recognition of the clinical concerns regarding their cognitive functioning, survivors under 18 years of age were routinely referred for neuropsychologic assessment as of 1991. The psychologic files of children who met the inclusion criteria and who had been administered all or parts of the Wechsler Scales of Intelligence [16], [17], [18], [19] and the Wide Range Assessment of Memory and Learning (see Appendix 1) between 1991 and 1998 were reviewed. The interval between radiotherapy and assessment was variable. In those instances where children were assessed within 2 years of diagnosis, only the data of those who had remained in continuous progression-free remission for a minimum of 2 years after treatment were included.

The current series was composed of 11 children with medulloblastoma and four children with cerebellar astrocytoma (nine females and six males). The decision to examine all 15 children as a single group was supported by the fact that all tumors had originated in the posterior fossa region, and all had been treated with radiotherapy. Sample characteristics presented in Table 1. The clinical features of the children diagnosed with medulloblastoma are described in greater detail in the report of a larger review conducted within the same institution [20].

Table 1. Sample characteristics

	Characteristics	n = 15	Age at Diagnosis
			<6 yr (n = 6)	≥6 yr (n = 9)
	Medulloblastoma	11 (73%)	3 (50%)	8 (89%)
	Cerebellar astrocytoma	4 (27%)	3 (50%)	1 (11%)
	Hydrocephalus	11 (73%)	6 (100%)	5 (56%)
	Ventriculoperitoneal shunt	4 (27%)	3 (50%)	1 (11%)
	XRT dose (range)	45-56 Gy	45-54 Gy	53.2-56 Gy
	Chemotherapy	9 (60%)	4 (67%)	5 (56%)
	Age at diagnosis (mean, S.D.)	7.9 (4.4)	3.9 (1.8)	10.5 (3.4)
	Interval between XRT and assessment (mean yr, S.D.)	3.5 (3.7)	6.2 (4.5)	1.8 (1.6)
	Assessed ≤2 yr after XRT	6 (40%)	0 (0%)	6 (67%)
	Assessed >2 yr after XRT	9 (60%)	6 (100%)	3 (33%)

Abbreviation XRT  Radiation therapy

Treatment for medulloblastoma consisted of suboccipital craniectomy, followed by irradiation to the craniospinal axis in doses ranging from 25.2 to 36 Gy, with a boost to the primary tumor site of 16-23.4 Gy. Total tumor dose ranged from 45 to 56 Gy. A boost to the spinal cord was indicated for two children (9 Gy and 30 Gy). Children diagnosed with cerebellar astrocytoma received 50-56 Gy to the posterior fossa region after surgery. Two children had gross total resections and two underwent subtotal excisions. Three tumors were classified as Grade III and one was classified as Grade II. Chemotherapeutic agents (e.g., vincristine, CCNU, prednisone) were administered to nine children.

Age at the time of irradiation (mean = 8.1 years, S.D. = 4.5) was nearly equal to the age at diagnosis (mean = 7.9 years, S.D. = 4.4) for all children except one. The location (i.e., extension into pons and medulla) of this child’s tumor precluded surgical intervention, and its growth was therefore closely monitored for approximately 1 years before the initiation of radiotherapy. This child also had the distinction of being the only one in this series born prematurely (32 weeks gestation). A visual inspection of sample means revealed no differences when this child’s data were removed from analysis. The decision was therefore made to include this child to increase the generalizability of the findings.

Analyses were conducted using SPSS, Version 7.5. All test scores were standardized based on a mean = 100, and a S.D. = 15, and sample means were compared with the normative means of the population. Independent samples t tests were performed to examine differences between means for Verbal and Performance IQ, and verbal and visual memory. A stepwise regression analysis was conducted to examine the amount of variability in outcome measures accounted for by two independent variables: age at diagnosis and time since treatment. A third independent variable, total radiation dosage, could not be examined in greater detail because of the restricted range of doses used within this small series.

To assess the effect of age at diagnosis further, the group was divided into two groups, using an arbitrary cut-off of 6 years of age. A paired-samples t test was performed to compare the performance of children who were younger at diagnosis with those who were older.

Results 

Children were tested at a mean age of 11.6 years (S.D. = 3.7) with a range of 6.2-17.8 years. The time interval between radiotherapy and neuropsychologic assessment ranged from 2 months to 15.2 years (mean = 3.5 years, S.D. = 3.7). One child in this series was assessed more than 10 years after craniospinal irradiation (i.e., 15.2 years later). This child’s data were initially excluded from analysis but then replaced when a visual inspection of their scores revealed no differences when compared with the group means. Six children were identified as English speaking, whereas seven others claimed French as their first language. Two children were fluently bilingual in both English and French. These children were tested in French, which was their preferred language. Analysis revealed that language of assessment did not have a differential effect on Verbal IQ ratings for those children who were not bilingual [t(11) = 0.015, P = n.s.].

Baseline neuropsychologic assessments before radiotherapy were not performed with most children because of their postoperative irritability and general malaise. Where more than one assessment was available for any given child, the most recent assessment was included in this review. The sample means for each dependent variable are reported in Table 2. The sample means were significantly lower than the means of the normative population in all instances.

Table 2. IQ and memory index scores compared with normative population*

		Mean	Standard Deviation
	IQ†
	Verbal (n = 14, 93%)§	88.07‡	12.06
	Performance (n = 13, 87%)§	86.46‡	16.1
	Full Scale (n = 13, 87%)§	86.31‡	13.92
	WRAML¶
	Verbal (n = 15, 100%)	77.67‡	17.0
	Visual (n = 14, 93%)§	87.36‡	21.94
	General (n = 12, 80%)§	88.92‡	20.62

* The standardized tests used with this series have a mean = 100, S.D. = 15. † IQ scores are from the age-appropriate Wechsler scale. ‡ P < 0.006. § The variability in sample size reflects conditions (e.g., optic atrophy) that interfered with the child’s ability to complete tests. ¶ Wide Range Assessment of Memory and Learning.

A paired samples t test was conducted to establish whether differences existed between measures of verbal and nonverbal intelligence. The difference between verbal and nonverbal IQ means was not statistically significant [t(12) = 0.696, P = n.s.] nor was the observed difference of sufficient magnitude to be considered clinically significant. Differences between mean Verbal and Visual Memory Indexes also did not reach statistical significance [t(13) = 2.036, P = n.s.].

A stepwise multiple regression analysis was performed to determine how much of the variability in intelligence was accounted for by the variables, age at diagnosis, and time since treatment. A significant proportion of the variability in Full Scale IQ (61%) was accounted for by age at diagnosis [r2 = 0.61, F(1,11) = 17.15, P < 0.01]. Sixty-seven percent of the variability in Verbal IQ was accounted for by age at diagnosis [r2 = 0.67, F(1,12) = 24.01, P < 0.01], whereas 45% of the variability in Performance IQ was accounted for by age at diagnosis [r2 = 0.45, F(1,11) = 9.31, P < 0.05]. Time since treatment did not account for a significant amount of the remaining variability in intelligence scores. When a stepwise multiple regression was performed on memory indexes, neither time since treatment nor age at diagnosis contributed significantly to the variability in scores.

Because age at diagnosis accounted for a significant amount of the variability in intelligence scores, the children were divided into two groups: those under 6 years of age (n = 6) and those 6 years of age and older (n = 9) at diagnosis. An independent sample t test determined that the estimates of Verbal, Performance, and Full Scale IQs for children less than 6 years of age were significantly lower than those of children who were older than 6 years at time of diagnosis (Table 3).

Table 3. IQ and memory index scores by age at diagnosis

		<6 yr (n = 6)		≥6 yr (n = 9)
		Mean	Standard Deviation	Mean	Standard Deviation
	IQ
	Verbal	77.33*	7.99	96.13	7.18
	Performance	73.4†	6.23	94.63	14.98
	Full Scale	73.2*	4.15	94.5	11.08
	WRAML
	Verbal	68.83	7.25	83.56	19.38
	Visual	76.4	13.35	93.44	24.01
	General	76.75	2.06	95	23.22

* P < 0.01. † P < 0.05.

The means for General, Verbal, and Visual Memory were also calculated for the two age groups. Although the difference between these means is sufficiently great to be of clinical relevance, statistical significance was not demonstrated.

Given the small sample size used in this review, posthoc power calculations were performed to determine the likelihood that a true difference between group means had been detected. The results indicated that there was sufficient power to detect differences in intelligence means, but not memory indexes.

A closer inspection of the data suggested that the power of the test may have been reduced by the magnitude of the variability in the memory scores for the group of children 6 years of age and older (see Table 3).

Discussion 

The use of radiotherapy has been linked to declines in cognitive functioning in children treated for posterior fossa tumors. Although IQ deficits have been documented in several studies, the literature on the effects of radiotherapy on memory functioning is limited. Variability in the assessment tools used to measure memory functioning has made it difficult to compare across studies. Standardized measures of both verbal and visual memory functioning have only been available for use with children since 1990.

In the present sample, the means derived from six measures of outcome all differed significantly from those of a normative population. This finding of impairment in children treated for posterior fossa tumors was consistent with that reported in the literature with respect to IQ. Full Scale, Verbal, and Performance IQ ratings were all consistently lower than normal, and fell within the Low Average range. Resection of cerebellar tumors has been linked to changes in expressive language skills [21], [22]. Cerebellar mutism may occur as a postoperative complication, particularly if the vermis is severed. None of the children in this review experienced mutism in the early stages of recovery; we did not therefore expect an adverse effect on Verbal IQ in our sample. Before this review we had expected a difference between IQ ratings, with the Performance IQ rating being lower. The absence of a difference between Verbal and Performance IQ was somewhat surprising. Considering that all tumors arose in the cerebellum, one would expect lower scores on the speeded Performance IQ subtests which place greater demands on the motor system. One plausible explanation was that the motor functioning of the children in this series was more intact than that reported in other studies. However, it was not possible to address this interpretation within this retrospective review. Prospective studies that include measures of manual dexterity, motor speed, and hand-eye coordination would be necessary to answer this question.

The means for the Verbal, Visual, and General Memory Indexes were also statistically lower than population means, with a marked impairment observed on measures of verbal memory. These findings of impairment in verbal memory are consistent with those of Packer et al. [23] and Mulhern et al. [24]. Few studies have included measures of visual memory, and comparisons across studies are complicated by differences in the type of information reported and the measures used. Using the Rey Auditory Verbal Learning Test [25], the Logical Memory subtest from the Wechsler Memory Scale [26], and the Benton Visual Retention Test [27], Johnson et al. [13] reported on the proportion of children obtaining impaired memory scores. Their results suggested a greater incidence of impairment in visual memory than in verbal memory.

Age at diagnosis accounted for a significant amount of variability in intelligence scores, although time since treatment did not. The mean IQ scores for children less than 6 years of age at diagnosis was significantly lower than that of children who were 6 years of age or older at diagnosis. These findings are consistent with previous reports [for example 9,11], which demonstrated that age at diagnosis is an important factor in the prediction of later cognitive status in children treated for posterior fossa tumors. The difference between the means in the current series of children was of sufficient magnitude to have significant implications for educational programming.

Neither age at diagnosis nor time since treatment accounted for a significant amount of the variability observed in the memory scores of our series. Heideman et al. [10] have suggested that clinically apparent late effects of radiotherapy on cognitive functioning may not fully emerge until three or more years after treatment. Because some of the children (40%) in this series were tested less than 2 years after they received radiotherapy, this may account for the failure to demonstrate an effect of time since treatment on IQ and memory functioning. These results may therefore underestimate the long-term adverse effects of radiotherapy on cognitive functioning in this population.

The current review did not allow for an examination of delayed recall, or working memory which is thought to be mediated by frontal lobe structures and important for executive functioning. At the time this series was assessed, there were no standardized and norm-referenced tests available for use with children to examine these skills. To date, there is only one measure of memory functioning normed on children that provides measures of delayed recall for information presented audibly or visually. The Children’s Memory Scale [28] has only been in use since 1997.

The difference between the performance on measures of memory functioning for children less than 6 years at diagnosis and those over 6 years at diagnosis was not statistically significant. The memory index scores of the group of children 6 years and older were highly variable, with some children in this group performing just as poorly as those in the younger group. This variability diminished the likelihood of detecting a true difference between the means within groups in such a small sample. From a clinical perspective, the mean memory index scores of children less than 6 years of age at diagnosis fell consistently below average. The mean verbal memory index for this group of children was more than 2 S.D. below the normative mean, and the mean visual memory index was more than 1 S.D. below the normative mean (see Table 3).

Dennis et al. [2] suggested that evidence of increased impairment in children less than 6 years of age may be confounded with longer intervals between diagnosis and assessment. This confound could lead to an overestimation of the impact of age at diagnosis on outcome. This possibility is a concern in retrospective analyses in which the interval between treatment and assessment is unplanned.

Reports on the role of associated hydrocephalus and the effects of shunt history on cognitive outcome have been unclear and conflicting [2]. It was not possible to examine these variables in greater detail with the current sample size.

The strength of this review is in the inclusion of a standardized norm-referenced test of memory functioning, designed specifically for use with children. A limitation to this retrospective review is the absence of an appropriate control group. A prospective study designed to examine this question should include a comparison of children with posterior fossa tumors requiring radiotherapy to children with posterior fossa tumors who undergo surgery but do not receive cranial irradiation.

The underlying risk for sequelae may not be the same for children treated by resection alone. Variables that may differentiate children who have undergone cranial irradiation from those who receive no radiotherapy include the rate of rise of intracranial pressure as a result of cerebral swelling and the frequency of postoperative complications—variables that may or may not influence a child’s performance on cognitive measures. In addition, until recently, the risks of late effects were considered to be minimal for children treated only with surgery, and as such, they were not routinely referred for neuropsychologic assessment as a component of clinical care.

In the absence of pretreatment baseline measures it was not possible in this retrospective review to be sure that the cognitive impairments observed represented a real decline in functioning posttreatment. The gathering of data before treatment can be complicated by the child’s physical state. Children may be too ill to undergo testing immediately before surgery, or the time interval between surgery and radiotherapy may be too short. Alternatively, the question of deterioration in functioning could be addressed by conducting serial assessments at six months, two years, and five years after treatment on all individuals. If evidence of a deterioration in memory functioning may not become apparent for at least three years after treatment, serial assessments would facilitate investigation of possible late-onset memory problems. Therefore the children in the current review could demonstrate greater impairment at a later point in time.

A larger sample would facilitate grouping of children by tumor type (i.e., medulloblastoma vs cerebellar astrocytoma) and allow for an examination of the effects of whole brain vs focal irradiation. However, a systematic prospective study of this nature would be complicated by factors such as changes in the methods of administering radiation therapy, tumor recurrence, or residual tumor growth.

In summary, this report confirms previous findings of intellectual deterioration in children who received cranial irradiation in the course of treatment for posterior fossa tumors. From a clinical perspective, children less than 6 years of age at diagnosis experienced greater difficulty on measures of memory functioning.