Two hundred fifty patients, younger than age 21 years at diagnosis, were enrolled onto the therapeutic trial between 1985 and 1992. Detailed descriptions of patient characteristics, disease parameters, and methods of treatment have been previously reported.9–16
Children 36 months of age and older at study entry were randomly assigned to one of two treatment arms, both of which included radiotherapy 54 Gy in 1.8-Gy fractions directed to the tumor volume with 2-cm margins. The control regimen, regimen A, consisted of weekly vincristine during involved-field irradiation followed by eight cycles of lomustine, vincristine, and prednisone administered every 6 weeks. The experimental regimen, regimen B, involved two cycles of the eight-drugs-in-1-day (8-in-1) chemotherapy regimen (lomustine, vincristine, hydroxyurea, procarbazine, cisplatin, cytarabine, dacarbazine, and methylprednisolone), followed by irradiation given at the same volume and dose and then by eight cycles of 8-in-1 maintenance chemotherapy. Patients younger than 36 months at diagnosis were nonrandomly assigned to the 8-in-1 treatment regimen and did not receive radiotherapy. Patients with primary spinal cord high-grade gliomas age 36 months and older were nonrandomly assigned to receive 8-in-1 with 36 Gy of irradiation to the craniospinal axis.
All surviving patients who completed multimodality therapy on CCG-945 were eligible to participate in the COG L991 outcomes study, including patients who had experienced relapse and required additional treatment. Patients provided signed informed consent before enrollment on institutional review board–approved protocols in compliance with the Declaration of Helsinki.
Patients were administered a battery of standardized tests either by a licensed psychologist or, in some cases, by a psychology trainee under the supervision of a licensed psychologist at their respective local institutions. Age-stratified normative standards for the neuropsychological, social-emotional, behavioral, and QoL assessment instruments were used (Appendix
, online only).
To assess for potential selection bias, χ2 analyses were performed to compare demographic and clinical characteristics of the 54 patients enrolled onto this study with those of 25 surviving patients who were enrolled onto CCG-945 but not COG L991. There was no indication of selection bias based on age, sex, race, extent of resection, and tumor location. Individual test performances were compared with the normative test means for each measure using one-sample t tests. One-way analysis of variance tests were conducted to determine the differential impact of the following risk factors: age at diagnosis, race, sex, tumor pathology, extent of resection, tumor location, treatment regimen, relapse status, and months of follow-up. For predictors with more than two levels, when a main effect was found, a Tukey post hoc test was conducted to identify significant differences between the groups. Two-sided analysis was used in all comparisons, and a P < .05 was considered statistically significant for all instruments except for the Delis-Kaplan Executive Function Test (DKEFS), the California Verbal Learning Test (CVLT), and the Rey Complex Figure Test, where an α level of P < .01 was established to minimize the chance of a type I error, given the multiple comparisons. For the Neurobehavioral Function Inventory (NFI), matched-pair t tests were conducted to examine differences in family and patient ratings on the six subscale scores.
Fifty-four survivors were enrolled between 2001 and 2005 at 25 institutions across North America, representing 81% of the available survivors (54 of 79 survivors). Thirteen survivors declined participation, and 12 could not be located by the treating institutions. The study sample included 29 males and 25 females with a median age of 8.8 years (range, 0.2 to 19.5 years) at diagnosis; median length of follow-up was 15.1 years (range, 9.5 to 19.2 years); and median age at study evaluation was 23.6 years (range, 11.3 to 36 years; ).
Patient Demographics and Clinical Characteristics
Mean intellectual functioning was measured to be within the low-average range for Full Scale IQ (FSIQ), Verbal IQ (VIQ), and Performance IQ (PIQ) on the Wechsler Abbreviated Scale of Intelligence ().
Neuropsychological Functioning, Behavior, and Quality-of-Life Test Results
Predictor variables were initially determined using VIQ, PIQ, and FSIQ as outcome measures. A sex main effect was found for VIQ (F = 5.42, P = .02). Female patients obtained lower VIQ (mean score, 74.9) compared with male patients (mean score, 87.9). In addition, age at treatment was significantly associated with both FSIQ (F = 4.87, P = .03) and PIQ (F = 5.28, P = .026). Patients younger than 3 years at treatment obtained lower FSIQ scores (mean score, 70.0) compared with older patients (mean score, 86.1). Younger patients also reported lower PIQ (mean score, 71.4) than patients older than age 3 years at the time of treatment (mean score, 86.7). Tumor location main effects were associated with FSIQ (F = 3.25, P = .03) and VIQ (F = 3.96, P = .01). Tukey post hoc tests indicated significantly higher FSIQ and VIQ in patients with spinal cord tumors (mean scores, 100.3 and 103.3, respectively) than those treated for tumors in the posterior fossa (mean scores, 64.9 and 66.1, respectively).
Verbal Learning and Memory
Verbal learning and memory was within the low-average range on the CLVT. Short Delay Recall remained in the low-average range but declined to the borderline range after a 20-minute long delay. Recognition Memory improved to the low-average range.
Tumor location main effects were found for the following three scales of the CVLT: Overall Learning (F = 3.88, P = .015; trend), Short Delay Free Recall (F = 3.95, P = .014; trend), and Short Delay Cued Recall (F = 4.74, P = .006). Tukey post hoc tests indicated significant differences in Overall Learning between spinal cord (mean score, 54.0) and posterior fossa (mean score, 25.6) patient subgroups; in Short Delay Free Recall between spinal cord (mean score, 0.5) and midline (mean score, −2.4) patient subgroups; and in Short Delay Cued Recall between spinal cord (mean score, 0.1) and midline patient subgroups (mean score, −3.0).
Visual-Spatial and Visual Memory
Mean performance on the Rey Complex Figure Test, a combined measure of visual-spatial constructional planning and motor integration, was within the impaired range indicating severely affected skills. When subsequently asked to redraw the design from memory, the study sample demonstrated a borderline ability in both short- and long-delay recall conditions, as well as on the recognition trial.
Performances on the executive functioning subtests from the DKEFS varied from the low-average to borderline ranges. Tumor location main effects were determined for the following subtests of the DKEFS: Color-Word Interference Test (Word Reading, F = 5.7, P = .0024; Inhibition, F = 5.99, P = .0019; Inhibition/Switching, F = 7.27, P < .001), and Trail Making Test (Number Sequencing, F = 0.43, P = .013, trend; Number-Letter Switching, F = 4.56, P = .0079). Significant differences were noted between spinal cord and posterior fossa patient subgroups on all five subtests and between spinal cord and cerebral hemisphere patient subgroups in all three Color-Word Interference subtests. There was a significant overall tumor location main effect for the correct number of sorts in the Sorting Test (F = 4.72, P = .0072) and an indication of trend in Sort Descriptions (F = 4.13, P = .014) and in Color Naming of the Color-Word Interference Test (F = 4.27, P = .0103). However, for all three measures, Tukey tests comparing all four patient subgroup means did not yield significant findings for any of the paired means.
Psychomotor Processing Speed
Psychomotor processing speed on the Symbol Digit Modalities Test was found to be the most severely affected higher cortical function as documented by the impaired study sample mean. Tumor location main effects were observed for psychomotor processing speed on the Symbol Digit Modalities Test (F = 4.30, P = .0097). Tukey post hoc analysis indicated a significant difference in processing speed between patients with spinal cord tumors (mean score, −0.74) and both patient groups with posterior fossa and midline tumors (mean scores, −3.51 and −3.55, respectively).
QoL is a value compromised of psychological and social functioning, along with any physical discomfort and disability. Despite documented functional variability, responses on the Short Form-36 Health Survey (version 2.0) and Child Health Questionnaire indicated overall physical and psychological QoL to be solidly within the average range.
Combined analysis using the overall QoL summary scores from the Child Health Questionnaire and Short Form-36 indicated that sex was the single best predictor of physical health (P = .05). Female patients (mean score, 42.8) reported poorer overall physical QoL when compared with male patients (mean score, 49.7). There were no predictor variables associated with the overall psychological QoL scale.
Social-Emotional and Behavior Functioning
Both patients and their caregivers (parents, 89%; spouses or siblings, 11%) completed the NFI to assess social-emotional and behavioral functioning. The consistency between reports from patients and these secondary data sources was calculated for each subscale by differences in matched-pair T scores, with higher scores indicating greater disparities (). There were no significant differences between family and patient mean scores for the Motor, Somatic, Depression, and Aggression subscales. The Communication subscale revealed a significant difference between family reports (mean score, 58.87) compared with patient endorsements (mean score, 51.68) of problems (t = 4.16, P < .01). The Attention/Memory subscale demonstrated a significant difference between family perceptions (mean score, 50.71) compared with patient reports (mean score, 46.19) of difficulties (t = 2.78, P < .01).
Comparisons Between Patient and Caregiver Reports of the Neuropsychological Functioning Inventory
Analyses to identify potential predictors of lowered social-emotional and behavioral functioning were conducted using the six subscales of the NFI as outcome measures. Univariate analysis revealed that sex was significantly associated with scores on the Depression subscale. Family members of female patients (mean score, 45.7) perceived more frequent problems involving depression than family members of male patients (mean score, 43.7; F = 7.24, P = .01).