Informed consent was obtained according to the guidelines of the ethical committees of the participating centers.
The improved prognosis of childhood cancer makes monitoring of functional outcome important. The purpose of this study was to evaluate behavioral and educational functioning in survivors of childhood acute lymphoblastic leukemia (ALL) or a Wilms tumor. In this study, children with ALL received central nervous system directed chemotherapy without cranial irradiation.
In a multicenter study, behavioral functioning and school performance was examined in 199 children age 4 to 18. Sixty-four children were at least 1 year from finishing treatment with chemotherapy for ALL (n = 28) or a Wilms tumor (n = 36). They were compared with siblings (n = 37) and with a control group of healthy schoolchildren (n = 98).
A moderately increased risk of behavioral and educational problems was found in children with ALL but not in children with Wilms tumor. School performance was poorer in children with ALL attending primary school compared with same-age peers; however, the rate of utilization of special education services was low. Teacher-rated behavior and mathematics performance was correlated with attention function in children with ALL. An excess of problem behavior and underperformance at school was found in the ALL high-risk group compared with the standard-risk group. No differences were found between siblings and controls.
Long-term toxicity and functional outcome have become important in monitoring survivors of childhood cancer because survival rates have improved markedly over the past decades. Overall 5-year survival rates are now around 70%. Acute lymphoblastic leukemia (ALL) and Wilms tumor are among the malignant diseases with the best prognosis.1
Prophylactic central nervous system (CNS)-directed therapy is an essential part of treatment for ALL. From previous studies, neurocognitive sequelae of cranial irradiation have become evident,2, 3 and a relation between cognitive impairment and behavioral and academic outcome in survivors of ALL has been reported.4 In current ALL treatment protocols, cranial irradiation has largely been replaced by CNS-directed chemotherapy. Although the long-term neurocognitive and psychosocial effects of CNS-directed chemotherapy in survivors of ALL seem to be less severe than the sequelae of cranial irradiation,2, 5, 6 evidence is emerging that chemotherapy-only is also associated with long-term neuropsychological sequelae.7–9 However, knowledge of the possible association of neuropsychological side effects with behavioral and educational problems is limited. In children with a Wilms tumor who receive non-CNS chemotherapy, no evidence has been reported for long-term cognitive deficits.8–11
Numerous studies have been conducted on psychosocial and academic outcome in survivors of childhood cancer. An increase in the rate of behavioral problems, measured using standardized checklists, has been reported in childhood cancer survivors,12–14 but these reports are contrasted by studies that suggest that survivors do not differ importantly from controls and/or are within normal limits.15–17 Educational attainment in survivors of non-CNS cancer has been reported to be similar to the general population.18 The reports often concern heterogeneous samples and do not specify outcome by diagnosis or treatment group.
Although mean psychological adjustment in pediatric cancer survivors may be near normal levels, more subtle or specific areas can be adversely affected in long-term survivors.19 Populations at particular risk are children with brain tumors and children who received CNS-directed treatment.20–22 Because children with ALL receive CNS-directed treatment, they are likely a group at risk for behavioral and educational sequelae.
The purpose of this study was to examine the behavioral and educational outcome in children after chemotherapy for childhood ALL or Wilms tumor. Siblings of the survivors were also assessed. The results may help to improve care for survivors and their families by providing insight on the need for intervention and remediation of possible learning problems. Awareness of long-term risks may also be important for the design of future treatment protocols.
MATERIALS AND METHODS
Participants and Treatment Protocols
We studied children between 4.5 and 18 years of age in complete remission from ALL or a Wilms tumor who had finished treatment at least 1 year before. The children with ALL were recruited from the VU University Medical Center. Children were excluded if they suffered preexisting developmental problems, had presented with CNS leukemia, received cranial irradiation, or suffered a recurrence, resulting in a total of 42 eligible survivors of ALL. From both the VU University Medical Center and the Erasmus MC Sophia Children's Hospital, 42 children with a Wilms tumor were invited. These children were also at least 1 year from the end of treatment and were matched as closely as possible for age and gender to the children with ALL. The children were invited to perform an extensive neuropsychological assessment and their parents or caretakers were asked to return a questionnaire and to pass on questionnaires to the teacher of the child. Thirty-six children with ALL and 39 children with a Wilms tumor agreed to participate. Final analyses include those children for whom questionnaires were returned by parents or caretakers: 28 children with ALL and 36 children with a Wilms tumor.
Siblings of children with ALL and of children with a Wilms tumor were enlisted in a sibling group (n = 37). The control group consisted of healthy age-matched schoolchildren, among them schoolmates of the children with ALL or a Wilms tumor (n = 98). Informed consent was obtained according the guidelines of the ethical committees of both treatment centers.
The children with ALL had been treated according to the consecutive Dutch Childhood Leukemia Study Group (DCLSG) ALL treatment protocols 6, 7, 8, and 9,23–25 and were diagnosed between 1985 and 1999. Patients were stratified into risk groups based on tumor load, immunophenotype, and the presence or absence of extramedullar disease. Children with standard-risk ALL received standard treatment (n = 14), whereas children with high-risk ALL received intensified treatment (n = 14). Regarding potentially neurotoxic therapy, the main distinction between the 2 treatment groups was that intensified treatment entails a higher cumulative systemic methotrexate dose than standard treatment.8 The children with ALL received CNS-directed chemotherapy as part of their treatment, which, in addition to high-dose systemic methotrexate, consisted of intrathecal monotherapy with methotrexate, or triple therapy with methotrexate, hydrocortisone, and cytarabine. The total duration of ALL treatment was 18 to 24 months.
Children with a Wilms tumor were treated with preoperative chemotherapy, followed by nephrectomy, postoperative chemotherapy, and, when necessary, abdominal radiation therapy, according to consecutive SIOP protocols 9 and 93-01.26, 27 Risk group stratification was based on stage and tumor histology. Treatment continued for 1 to 9 months after diagnosis.
All children in this study performed a neuropsychological assessment, the results of which have been reported separately.8, 9 Children with ALL performed worse than controls on measures of sustained attention, attentional flexibility, and visuomotor control, with deficits restricted mostly to the children with high-risk ALL. Performance of children with a Wilms tumor and siblings did not differ from controls.
Questionnaires on Behavior and Educational Achievement
Child Behavior Checklist (CBCL).
The CBCL parent report is a widely used instrument that assesses social competence and behavioral problems of children 4 to 18 years of age.28 The CBCL has been used to evaluate children with cancer in several studies.12, 15, 17, 29 We used the Dutch, standardized version of the instrument.30 Socioeconomic status was measured as highest professional level of parents on a 6-point scale.31 We used the problem scale of the CBCL, which consists of 118 items. Three global scores were calculated: a total problem score and an internalizing and externalizing score. Eight subscale scores were computed: “withdrawn,” “somatic complaints,” “anxious/depressed,” “social problems,” “thought problems,” “attention problems,” and “aggressive behavior.” Raw scores were converted to T-scores to control for age and gender, with a mean score of 50 and a standard deviation (SD) of 10. Higher T-scores reflect poorer ratings.
Conners' Teacher Rating Scale (CTRS).
The CTRS is a commonly used questionnaire to assess teachers' perception of children's behavior in the classroom.32 It has been shown to be a valid and reliable tool.32, 33 The Dutch, validated, translation of the 32-item version was used in our study.34 Scores on three global scales were calculated: total score and internalizing and externalizing behavior. A 5-factor structure allowed analysis of the following subscales: acting-out, antisocial behavior, hyperkinesias, fear/withdrawn behavior, and social isolation. Higher scores correspond to poorer ratings.
The rate of children who needed special education services and of children who repeated a grade, currently or in the past, was assessed. A school performance index (SPI) was used for analysis of educational performance of children in primary school, concerning children 6 to 12 years of age. Teachers rated performance on 8 school topics (writing skills, mathematics, mental calculation, technical reading, comprehensive reading, spelling, written vocabulary, and oral vocabulary). Performance on each topic was scored on a 4-point scale (1 to 4, ranging from: “belongs to the 25% weakest pupils” to “belongs to 25% best pupils” in the grade). We identified three subscales using principal components analysis with varimax rotation in the control group: language skills, mathematics, and writing skills. Lower SPI scores reflect a poorer school performance.
Group differences of the parent and teacher questionnaires were compared using univariate analysis of variance (ANOVA). Planned simple contrasts were used to compare the scores of children with ALL, children with a Wilms tumor, and siblings with the healthy controls. For the analyses of the CBCL results, T-scores were used. For the CTRS and the SPI, raw scores were used. Because CTRS scores were influenced by gender, this variable was used as a covariate in the analyses. Associations between ratings on the various questionnaires were explored using correlation analyses.
Odds ratios (ORs) were computed, using logistic regression, to assess the likelihood of a score outside the normal range in each group compared with controls for each questionnaire. The upper limit of the normal range for the CBCL global scores, is 1 SD above the norm, corresponding to a T-score of 60.30 A T-score of 64 or higher is considered to be in the “clinical range,” indicating a need for clinical care. For the subscales, a T-score of 67 is the upper limit for the normal range. As similar cut-off norms were not available for the employed version of the CTRS, 1 SD above the mean of the control group in our study was taken as the upper limit of normal for the CTRS scales, computed separately for boys and girls. For the SPI, total and subscale scores of 1 SD below the mean of the control group were taken as the lower limit of the normal range.
Within the ALL group, the association of the previously obtained z-scores on neuropsychological tests of sustained attention, attentional flexibility, and visuomotor control with behavioral and educational measures was examined using correlation analysis. To assess the association between treatment intensity and rate of problem behavior and substandard school performance, the frequency of scores outside the normal range was contrasted with the controls for the ALL standard-risk group and in the ALL high-risk group separately. This was done for all outcome measures of the checklists using Fisher exact tests. P was set at 0.05 for all analyses.
In total, questionnaires were obtained for 199 children. Of the 84 survivors who were invited to participate in the study, 75 (89%) agreed: 36 (86%) children with ALL and 39 (93%) children with a Wilms tumor. Questionnaires were returned by parents or caretakers of 64 of the 75 (85%) survivors who participated in the study: 28 (78%) children with ALL and 36 (92%) children with a Wilms tumor. There were no differences regarding demographic, disease, or treatment variables between participants in the study and nonparticipants, or between the children for whom questionnaires were returned and for the children for whom they were not. The return rate of questionnaires for siblings and the healthy school children was 80% and 89%, respectively. Characteristics of the study children are shown in Table 1. There were no significant differences between groups on demographic variables, age at diagnosis, or time since diagnosis. The ALL group included 14 children with standard-risk ALL and 14 children with high-risk ALL. Four children (with high-risk ALL) received intrathecal monotherapy with methotrexate and 24 received triple intrathecal therapy (14 children with standard-risk ALL and 10 children with high-risk ALL).
The CBCL total problem score differed between groups (Table 2; F[3,195] = 3.26, P = .023). Children with ALL had a higher problem score than controls (CE = 5.94, P = .005), whereas the contrasts of children with a Wilms tumor and siblings, respectively, with controls were not significant. Also, more children with ALL had a problem score outside the normal range vs. controls (OR, 3.3; 95% confidence interval [CI] 1.1–9.9; P = .033). Furthermore, more children with ALL had a problem score in the clinical range: 5 children with ALL (17.9%) vs. four controls (4.1%; OR, 5.1; 95% CI, 1.27–20.54; P = .022).
Table 2. Child Behavior Checklist (CBCL) Scores by Group
There was a trend for a group effect on the internalizing scale (F[3,195] = 2.34, P = .074). A significant contrast with a higher internalizing score for children with ALL than for controls was found (CE = 5.47; P = .014). More children with ALL had a T-score outside the normal range on the internalizing scale compared with controls (OR, 5.12; 95% CI, 1.91–13.69; P = .001). Seven children with ALL (25.0%) vs. four controls (4.1%) had internalizing scores in the clinical range (OR, 7.83; 95% CI, 2.10–29.22; P = .002). On the externalizing scale, no group differences were detected.
Significant group differences were detected in 4 of 8 subscales of the CBCL, with significantly higher problem scores for children with ALL vs. controls: “withdrawn” (F[3,195] = 3.27, P = .022; CE for children with ALL vs. controls = 3.61, P = .003), “anxious/depressed” (F[3,195] = 3.01, P = .031; CE = 3.50, P = .004), “social problems” (F[3,195] = 3.30, P = .022; CE = 3.08, P = .005) and “attention problems” (F[3,195] = 4.59, P = .004, CE = 3.57, P = .0003). There were no significant differences between children with a Wilms tumor or siblings and controls.
Teachers reported a higher overall problem score for children with ALL compared with controls (Table 3; CE = 4.25, P = .047). Children with ALL were also more likely than controls to have a total problem score outside the normal range (OR, 4.1; 95% CI, 1.2–13.5; P = .020). The contrast between children with ALL and controls was significant on the externalizing scale (CE = 3.93, P = .039), and more children with ALL scored outside the normal range (OR, 3.6; 95% CI, 1.1–11.6; P = .032). On the subscales, only 1 difference was found, contrasting ALL patients with healthy controls for the factor “hyperkinesia” (CE = 1.40, P = .042). The contrasts between children with a Wilms tumor and siblings with controls were not significant.
Table 3. Conners' Teacher Rating Scale (CTRS) Scores by Group
Two children with ALL received special education, as did 1 sibling. No children with a Wilms tumor and no controls were in special education programs. Eight children with ALL (30.8%) had repeated a grade, as had 6 children with a Wilms tumor (17.1%), 4 siblings (11.1%), and 9 controls (9.4%). The difference in the rate of having repeated a grade was significant only between children with ALL and controls (OR, 4.30; 95% CI, 1.46–12.64; P = .008).
The SPI differed between groups (Table 4; F[3,90] = 7.57, P <.001). The total score was lower for children with ALL compared with controls (CE = −6.92, P <.001), whereas the children with a Wilms tumor and siblings did not differ significantly from controls. The SPI of children with ALL was also more likely to be outside the normal range than controls (OR, 22.5; 95% CI, 4.1–122.0, P<.001). The effect of group was significant for language skills (F[3,90] = 7.37, P<.001) and mathematics (F[3,90] = 3.97, P = .010), with significantly lower scores for children with ALL compared with controls (CE = −0.93, P<.001, and CE = −0.96, P = .001, respectively).
Table 4. School Performance Index (SPI) Scores by Group
Associations between Neuropsychological Test Results, Behavioral, and Educational Ratings in the ALL Group
Parent-teacher agreement for behavioral ratings was low for the total problem scores and for the internalizing and externalizing scores on the CBCL and CTRS. However, we did find a moderate correlation between the parent rating on the subscale of attention on the CBCL and the total score on the CTRS (r = .52, P = .016). The ALL group showed a high negative correlation between the attention subscale score on the CBCL and mathematics ability on the SPI (r = −0.76, P = .007).
Concerning the neuropsychological measures, poorer attentional flexibility, was associated with higher problem ratings on the CTRS total scale (r = .67, P = .033), internalizing scale (r = .62, P = .033) and acting out subscale (r = .60, P = .041). Poorer sustained attention was associated with higher problem ratings on the CTRS: total scale (r = .59, P = .020), externalizing (r = .66, P = .007), acting out (r = .67, P = .006) and hyperkinesia (r = .65, P = .009). Poorer sustained attention was also associated with a worse score for mathematics performance on the SPI (r = −0.77, P = .006).
Effect of Treatment Intensity in the ALL Group
Compared with controls, a higher frequency of scores outside the normal range was found in the high-risk ALL group compared with the standard-risk ALL group (Fig. 1). We did not find differences between children with ALL treated with intrathecal monotherapy with methotrexate compared with those treated with triple therapy.
In this study, behavioral and educational functioning were examined in children at least 1 year after finishing chemotherapy for ALL or a Wilms tumor. The results indicate an increased risk of behavioral and educational problems in children with ALL compared with same-age peers. In contrast, we found no indications of behavioral or educational limitations in children who survived a Wilms tumor or in siblings of the survivors.
The inclusion of several comparison groups permitted us to relate the results of the children with ALL who received CNS-directed treatment to children with a Wilms tumor, who experienced cancer but received no CNS-directed treatment, and to a sibling control group, who did not experience cancer, but did experience the emotional distress of a family member with the disease. This makes the hypothesis that neurotoxic effects of CNS-directed chemotherapy, rather than the experience of childhood cancer or the associated emotional distress, is a factor in the emergence of the behavioral and educational problems in children with ALL more convincing. The inclusion of a large group of normal controls offered the advantage of not having to rely solely on population norms for comparison. The multi-informant perspective obtained by using parent as well as teacher reports adds to the value of our findings.
The results of the CBCL indicate a higher incidence of parent-reported problem behavior for children with ALL compared with controls, mainly attributable to internalizing problems and attention problems. Although excess behavioral problems as measured with the CBCL have been reported previously in children with ALL, most of those studies concerned children treated with cranial irradiation.4 Studies on behavioral outcome in children with ALL treated with chemotherapy only are scarce. Mostly they have been included in larger heterogeneous samples, without specifying the results of this particular group. That behavioral problems are also reported by parents of children with ALL treated with chemotherapy only is merely suggested by the study of Anderson et al.,15 who found that total CBCL problem scores did not differ between survivors of ALL treated with cranial irradiation and a heterogeneous chemotherapy group including children with ALL. In adult survivors of childhood cancer, intensive chemotherapy has previously been reported to add to the risk for depressive symptomatology.35
Teachers also reported more problem behavior in the ALL group. Corresponding with the parent ratings, the teachers did not report excess behavioral problems for survivors of a Wilms tumor or for the siblings. Whereas parent ratings were high on the internalizing scale, teacher ratings were highest on the externalizing scale in children with ALL. The low parent-teacher agreement on these ratings corresponds with previous literature.36 We did find an association between parent-reported attention problems and teacher-reported behavior problems. Our results are in accordance with a recent study, also using the CTRS to assess behavior in the classroom, where methylphenidate was shown to have short-term benefit in improving hyperactive behavior and attention in a mixed group of childhood ALL and brain tumor survivors.37
Attentional problems appear to be an important factor in problem behavior and underperformance at school in children with ALL treated with chemotherapy only. It is a striking finding that poorer attentional function, as measured with neuropsychological tests, is associated with higher teacher-rated problem behavior and lower school performance scores in this group. This suggests that attentional dysfunction previously reported in children with ALL8 is not associated merely with subclinical cognitive problems, but forms a clinically significant finding with consequences for daily life.
In the children who were in primary education, weaker overall academic performance was found in survivors of ALL compared with controls. Previous reports of educational problems in children with ALL mainly concerned children who were treated with cranial irradiation.15, 38–42 We found that weaker mathematics performance was associated with poorer sustained attention, as measured by neuropsychological tests. This corroborates previous reports of an association between mathematics weaknesses and cognitive deficits in children with ALL.43–45 Poor school attendance during treatment may have contributed to the lower performance levels, but this does not likely play a major role, as over 70% of children with ALL of whom the SPI results were available had completed treatment before commencing primary school education.
Our findings of weaker school performance in the children with ALL must be seen in the light of the finding that the great majority of these children are able to function in normal schools. The rate of children who needed special education services in our study is low. It compares favorably with previously published rates of children with ALL treated with cranial irradiation,40, 41 and corroborates reports of lower utilization of special education services in children with ALL treated with chemotherapy only.21, 46
Although siblings have previously been reported to experience emotional distress and behavioral problems past the time of initial diagnosis and treatment of the cancer patient,47 we did not find evidence to support longer-term behavioral and educational problems in this group.
Regarding the role of treatment intensity within the ALL survivor group, our findings indicate that the children who received intensified treatment for high-risk ALL exhibited a higher frequency of behavioral problems, according to parents and teachers. Also, mathematics achievement was lower in this group of children. This is in accordance with our previous report of treatment intensity as a risk factor for attentional dysfunction in children with ALL treated with chemotherapy only.8
Our study contributes to the knowledge of long-term behavioral and educational outcome after chemotherapy in children with cancer. Whereas survivors of a Wilms tumor and siblings are functioning normally, subtle behavioral and educational sequelae are found in survivors of ALL. Even though most survivors of ALL are in normal schools, they are underperforming compared with their peers. From these observations emerges a consistent image, not of obvious disability, but of subtle yet significant problems of failing to quite realize innate potential regarding cognitive, educational, and behavioral functioning. In the follow-up of childhood cancer survivors, careful attention to possible behavioral and educational problems is indicated for children with ALL treated with chemotherapy only, so that appropriate assessment and intervention can be initiated when necessary. Our findings could be a basis for further research on remediation of attentional problems, attention training, and the efficacy of pharmacological interventions for behavioral and educational problems in survivors of childhood ALL.