Differential trajectories of neurocognitive functioning in females versus males following treatment for pediatric brain tumors

Abstract

Background

Female and male trajectories of cerebellar and lobar brain structures are sexually dimorphic, making sex a potential candidate moderator of neurocognitive late effects from radiation treatment. We sought to evaluate longitudinal neurocognitive functioning in male versus female children treated for posterior fossa brain tumors.

Methods

Fifty-one female and 63 male survivors of posterior fossa tumors completed neuropsychological testing at 2 timepoints. We included patients treated with surgical resection, chemotherapy, and radiation therapy. Multilevel mixed modeling was used to predict IQ score as a function of patient sex following treatment (~2 or ~4 years post treatment). Effect sizes were used as a measure of clinical significance.

Results

Multilevel models resulted in a significant sex by time interaction (F = 6.69, P = 0.011). Females’ cognitive scores were considerably higher compared with males at 4 years posttreatment. Females demonstrated an average improvement of 7.61 standard score IQ points compared with a decline of 2.97 points for males at 4 years follow-up. Effect sizes for female IQ compared with male IQ at 4 years posttreatment were between 0.8 and 0.9.

Conclusion

Trajectories of neurocognitive functioning following posterior fossa tumor treatment differed between female and male children. Sexual dimorphism in radiation late effects may alter treatment decisions in children. Research into sex-specific neuroprotective mechanisms underlying neurocognitive development following pediatric brain tumor treatments is warranted.

Brain and associated central nervous system (CNS) tumors are the most common cancer sites diagnosed in children from birth to age 14, and include an age-adjusted incidence rate of 5.54 per 100 000.¹ It has been estimated that nearly half of pediatric brain tumors arise in the posterior fossa, with male:female incidence rates of embryonal tumors being approximately 1.5:1.^1,2 Treatment for pediatric brain tumors often involves surgical resection as well as chemotherapy and cranial radiation. Long-term survival rates for children with brain tumors have improved considerably over the last 30 years,² though many patients experience long-term debilitating physical, cognitive, and psychological impairments following treatment.³ A central challenge in treatment involves the decision to administer cranial radiation to young children (eg, ≤5 y) which may reduce the likelihood of disease progression, although it has been associated with devastating impacts to neurocognitive development.^4–6

The majority of research on neurocognitive outcomes in children with pediatric brain tumors has contained significantly higher representations of male children, despite only slightly increased incidence in males compared with females.¹ As an example, many seminal studies evaluating neurocognitive outcomes following radiation therapy included male:female ratios that prohibited the statistical evaluation of sex-specific outcomes (28 male and 16 female,⁷ 17 male and 5 female, 18 male and 6 female,⁸ 25 male and 9 female,⁹ 25 male and 14 female,⁴ 20 male and 10 female,¹⁰ 26 male and 10 female,¹¹ 24 male and 7 female,¹² 34 male and 16 female¹³). Thus, differential impacts to the developing female versus male brain have not received adequate scrutiny. In one study, however, analysis of sex differences indicated that higher baseline intellectual functioning was protective for male patients compared with female patients based on 7-year random coefficient predictive modeling.¹³ The absence of rigorous investigations of sex differences is likely due to small patient samples, which hinder statistical power as well as a lack of a clear etiological reason for doing so.

However, we propose that there is ample empirical reason to hypothesize that radiation may exert diverse effects on developing male versus female brains. For example, by the ages of 7–11, peak cerebellar volume is reached in females but not in males.¹⁴ More specifically, the cerebellar development follows an inverted U-shape, peaking at 11.3 years for females and 15.6 years for males.¹⁵ Differences in cerebellar development are of particular importance given that the posterior fossa (cerebellum, fourth ventricle, brainstem) is the most common site for pediatric brain tumors.¹ There is also evidence of sexual dimorphism in lobar structural brain development. In the largest pediatric neuroimaging study to date, the National Institutes of Health observed significant differences in male versus female cerebral brain development, with female cerebral volume maturing 4 years earlier than males (10.5 for females and 14.5 for males).¹⁶ Given these developmental sex differences in early brain maturation, sex may play a role in our understanding of the mechanisms for radiation late effects. Furthermore, because a primary biomarker for radiation-induced neurocognitive late effects involves abnormal rates of structural white and gray matter development,^17,18 it is possible that sex may moderate neurocognitive outcomes following radiation. Understanding sexually dimorphic pathways of neurocognitive late effects in pediatric brain tumors may translate into sex differences in treatment, thus reducing cognitive side effects and the likelihood of recurrence.

In this study, we used multilevel mixed modeling to evaluate neurocognitive development in a large and balanced sample (N = 114) of male and female patients with posterior fossa brain tumors. The purpose of this study was to determine if male and female children’s trajectories of cognitive development, as evidenced by scores on measures of intellectual functioning (IQ), differed at approximately 2 and 4 years after receiving treatment for a brain tumor, after controlling for appropriate confounding variables.

Materials and Methods

Patients and Procedures

This retrospective study protocol was completed following approval from the institutional review board at a major regional pediatric medical center in the Northwest region of the United States. Patient data regarding age, tumor location, and treatment history including type and amount (chemotherapy, radiation therapy, and surgical intervention procedure) were obtained via local Brain Tumor Registry. We identified 224 patients from this registry who were treated for posterior fossa brain tumor in 1999–2014. We excluded patients due to having missing neuropsychological data, missing demographic data, or brain tumors outside of the posterior fossa (ie, supratentorial ependymoma), which resulted in a total sample size of 114 children. All patients (n = 114) were evaluated via routine neurocognitive assessments completed as part of their clinical surveillance due to a history of pediatric brain tumor. Because many of our patients completed multiple assessments, we obtained the first and second neuropsychological assessment that they underwent following treatment. One third of the patient’s data (38 IQ measures) were randomly selected for double entry to establish reliability and to evaluate potential manual data entry errors (Cohen’s kappa = 0.92).¹⁹ Patients consisted of 51 females and 63 males treated for posterior fossa brain tumors. The sample included 70 patients with medulloblastoma (61.4%), 20 with ependymoma (17.5%), and 24 with low grade astrocytoma (21.1%). Mean age at diagnosis was 7.10 years (SD = 5.06) for females and 7.45 (SD = 4.75) for males. Mean age at the time of their initial neurocognitive evaluation was 9.60 years (SD = 4.84) for females and 9.17 (SD = 4.17) for males. Mean age at follow-up neurocognitive evaluation was 10.55 (SD = 4.15) for females and 10.77 (SD = 3.60) for males. The interval between testing time 1 and testing time 2 was 1.71 years (SD = 0.969) for females and 2.26 (SD = 2.09) for males. The interval between diagnosis and testing time 1 was 2.68 years (3.42) for females and 2.45 (SD = 3.61) for males. Regarding treatment history, patients with medulloblastoma received surgical resection and radiation therapy (90% craniospinal, 5% no radiation, and 5% focal radiation), patients with ependymoma received surgical resection and radiation therapy (100% focal radiation), and patients with low grade astrocytoma received surgery alone. See Table 1 for participant demographics and treatment information.

Measures

Wechsler Abbreviated Scale of Intelligence (WASI)

²⁰Overall cognitive ability (ie, intellectual functioning) was assessed via standard administration of the WASI. The WASI was developed as a short form of the Wechsler Intelligence Scale for Children, Fourth Edition (WISC-IV)²¹ and the Wechsler Adult Intelligence Scale, Third Edition (WAIS-III)²² for ages 6–89 years of age. For the purposes of this study we included the Full Scale Intelligence Quotient (FSIQ), Verbal Comprehension Index (VCI), and Perceptual Reasoning Index (PRI). FSIQ is used as a robust composite of both VCI and PRI, though it was important to include VCI and PRI as there are often significant discrepancies between VCI and PRI, which invalidates the interpretation of the FSIQ.

Statistical Analyses

Categorical group comparisons were evaluated using chi-square analysis. Continuous group comparisons were evaluated with ANOVA as well as the Levene’s Test to evaluate homogeneity of variance (Table 1).

Following procedures as outlined by Cnaan, Laird, and Slasor²³ and Hock,²⁴ multilevel mixed modeling using restricted maximum likelihood was used to predict IQ scores as a function of patient sex (coded 1 = female, −1 = male), whether the person had had radiation treatment (coded 1 = radiation, −2 = no radiation), and time (coded 1 = 4 years posttreatment, and −1 = 2 years posttreatment). These variables were treated as categorical, and all main effects and interactions between them were included as fixed effects in the model. Age at treatment was also included in the fixed effects as a covariate. Nonindependence of observations due to repeated measures (time and IQ were within participant variables) was modeled using compound symmetry, which assumes equal residual variances across type and type of IQ as well as equal covariances between and across these variables. Due to potential cross-level interactions and multicollinearity between age at treatment and age at assessment, patient age (grand mean centered) was included as a covariate in the model. Patients were treated as the upper-level sampling unit and assessment occasion was treated as the lower-level unit.²⁵

Statistical Power

We calculated the statistical power of the sex by time interaction effect using G*Power.²⁶ The effect size for the sex by time interaction (based on the raw means and standard deviations), calculated by comparing the difference between the 2- and 4-year IQ average scores for boys and girls, and taking the repeated measures correlation of 0.806 into account was d = 0.87. Power for this test was greater than 0.9, which is considered high according to Cohen’s criteria.²⁷

Results

Demographics and Treatment Variables

Categorical variables

Chi-square analysis revealed no differences in male and female rates of receiving radiation treatment (χ² = 2.55, P = 0.110) or rates of having medulloblastoma, ependymoma, or low grade astrocytoma (χ² = 1.044, P = 0.593). We also found no differences in male or female rates of insurance type (private, Medicaid, or unknown; χ² = 1.89, P = 0.389), race/ethnicity (χ² = 7.97, P = 0.253), chemotherapy type (none, standard, or high dose) (χ² = 1.299, P = 0.522), use of proton or photon radiation (χ² = 3.83, P = 0.148), history of hydrocephalus (χ² = 1.138, P = 0.286), history of shunt (χ² = 0.010, P = 0.922), history of seizures (χ² = 0.305, P = 0.580), or history of posterior fossa syndrome (χ² = 0.001, P = 0.972).

Continuous variables

There were no statistically significant group differences between males and females in age at diagnosis (F = 0.140, P = 0.709), age at radiation therapy (F = 0.346, P = 0.558), age at first neurocognitive evaluation (F = 0.249, P = 0.619), age at follow-up neurocognitive evaluation (F = 0.064, P = 0.801), interval between first neurocognitive evaluation and follow-up neurocognitive evaluation (F = 2.16, P = 0.146), testing interval between age at diagnosis and age at time of first evaluation (F = 0.138, P = 0.489), or total amount of radiation therapy (F = 0.685, P = 0.410). The Levene’s test for homogeneity of variance was evaluated in each ANOVA, and no significant differences were observed suggesting relatively equal variance across comparison predictors (Table 1).

Neurocognitive functioning at follow-up

Multilevel mixed models revealed a significant sex by time interaction effect, including age at treatment as a covariate (F = 6.69, P = 0.011) as well as the main effect of sex (F = 9.678, P = 0.002; Figure 1). All other main effects and interactions had F-values <1.0 and P-values >0.36. Thus there was no evidence of differences as a function of the type of IQ test used or whether the individual had radiation treatment. The compound symmetry correlation was 0.807, indicating considerable stability in scores across time and test type. The effect sizes at 4 years were considered very large, ranging from 0.8 to 0.9 using Cohen’s criteria.²⁷ Importantly, when the sample was divided into those not treated with radiation, male and female IQ distributions were not different (eg, effect size of 0.16). When the sample was divided into those treated with radiation, males’ average IQ was considerably lower, resulting in an effect size of 0.86 (see Figure 2). Furthermore, when viewing the data clinically, mean full scale IQ improved by 7.61 standard score points for females and declined by 2.97 standard score points for males, on average, 4 years following treatment. This translates to a difference of 11.68 full scale IQ points at 4 years posttreatment for females compared with males. Despite these differences in standard scores at follow-up, both female and male survivors’ mean scores were in the age-appropriate range compared with their same-aged peers (Figure 3).

This translates to a difference of 11.68 full scale IQ points at 4 years post treatment for females compared to males. Despite these differences in standard scores at follow-up, both female and male survivors’ mean scores were in the age-appropriate range compared to their same-aged peers (see Table 2).

Discussion

This study evaluated the longitudinal trajectories of neurocognitive functioning in a balanced sample of female and male patients with posterior fossa brain tumors. To our knowledge, this is the largest single site clinical study to examine longitudinal sex differences in neurocognitive functioning in pediatric brain tumors. We observed similar performance, at the mean level, between male and female children at their baseline evaluation 1.8 years posttreatment, though significant differences emerged at nearly 4 years posttreatment, with females’ cognitive scores improving, on average, compared with males, who declined. This pattern suggests the possibility of sexual dimorphic neurocognitive trajectories in children with posterior fossa brain tumors and warrants further study.

Our findings of relative improvement in female patients compared with males is contrary to prior studies. Ris et al²⁸ found significant declines (approximately 4 IQ points per year) in full scale IQ, verbal IQ, and nonverbal IQ in male and female patients treated for medulloblastoma/primitive neuroectodermal tumors treated with reduced dose craniospinal radiation and adjuvant chemotherapy. They observed a greater decline in verbal IQ for females (8 points per year) compared with males (3 points per year). The authors acknowledged that the uneven samples size (34 males and 9 females) impacted statistical power, likely increasing type I error. It is also worth noting that this study was completed at multiple institutions that used 5 different intelligence tests to derive IQ domain scores, including multiple reported off-protocol deviations in test selection. This further complicates the interpretability and reliability of these findings and further validates the inherent challenges of multi-institutional research in low-incidence diseases. In a similar study of 50 children treated with 35–40 Gy craniospinal radiation for medulloblastoma, female sex was found to be predictive of lower baseline IQ scores compared with males (96.62 for males and 87.50 for females, P = 0.0479).¹³ The median patient IQ in this sample of 51 patients was 88, which is well below the median IQ score in the current study, which was 95 for both initial and follow-up timepoints. One potential reason for the higher baseline IQs in our study was the inclusion of patients with low grade astrocytoma who received only surgical resection. Despite non-significant difference in IQ scores at initial (P = 0.296) and follow-up (P = 0.157), low grade astrocytoma patients had higher mean level full scale IQ scores (initial 98.16, follow-up 102.83) relative to patients with medulloblastoma (initial 91.23, follow-up 91.76) or ependymoma (initial 95.28, follow-up 100.70) who received radiation therapy. We also observed that male patients treated with radiation declined, whereas female patients treated with radiation improved at 4 years posttreatment. Regardless, mean level IQ improved for all tumor groups in the current study, suggesting that declines in IQ estimates may be independent of tumor type. Clinical investigations of cognitive trajectories following radiation for different brain tumor types, stratified by sex, will be challenging in terms of obtaining adequate sample size, though they may yield critical information about differential sex differences following treatment.

Sex Differences

Sex differences have tremendous implications for health and disease susceptibility, progression, mortality, as well as drug metabolism.^29,30 As an example in pediatric neuro-oncology, medulloblastoma, the most common malignant tumor type in children, includes 4 distinct molecular varieties; Wnt, sonic hedgehog, and Group 3 and Group 4.^31,32 Only Group 3 and Group 4 exhibit a 2:1 male to female incidence ratio compared with Wnt and sonic hedgehog, which have roughly identical male to female incidence. It is possible that sex differences underlying genetic tumor variants may become useful biomarkers or associated risk factors for cognitive late effects. Despite a growing literature that is identifying meaningful sex differences in cancer oncogenics³³ and therapeutics,³⁴ few studies have evaluated the treatment side effect profiles, including cognitive late effects, between males and females. The question of whether and how sex may moderate neurocognitive outcomes remains equivocal.

One potential reasons for females’ apparent resiliency relative to male peers may be due to sex differences in trajectories of structural brain development. In the typically developing population, males and females exhibit significant differences in their rates of cortical and subcortical development.^16,35 Despite total cerebral volume being approximately 7–10% larger in males compared with females at ages 7–11, the cerebellum reaches peak volumes for females at age 11.3 and males at age 15.6.^14,15 Thus, rates of cortical maturation may be a critical variable for understanding and predicting how radiation impacts developing brain tissue. Results are mixed regarding sex differences in gray matter volume or rates of cortical thickening in adolescents,^16,36,37 though there are important genetic differences that regulate lobar thickness compared with volume.³⁸ More research is needed regarding the potential mechanism behind neurocognitive late effects, particularly the relationship between radiation dose and cortical thickness and cortical volume. Very few investigations into the impact of radiation on cortical gray matter structures, relative to white matter, have been completed.³⁹ More comprehensive multi-method longitudinal imaging studies of radiation late effects using standardized cognitive assessments will be critical for obtaining a more complete model of the neurobiological mechanisms of radiation late effects.

Lastly, decades of animal research indicate that testosterone and estradiol plasma levels directly impact neurodevelopment. For this reason, it is also possible that sex differences in underlying pubertal steroid levels may influence cognitive change following treatments via alterations in structural and functional brain development. For example, for males, testosterone is associated with myelinogenesis,⁴⁰ whereas estradiol levels have been associated with hippocampal cell proliferation,⁴¹ dendritic spine number,⁴² and synaptogenesis.⁴³ In particular for female patients, recent work has observed that level of progesterone, progestin, or progestin metabolites may modulate brain neuroplasticity via inhibiting the release of free radicals from microglia⁴⁴ or by increasing myelin production.⁴⁵ This is of particular relevance to pediatric brain tumor patients, as progesterone levels have been associated with improved cognitive performance following traumatic brain injury⁴⁶ and hippocampal regeneration.⁴⁷ Thus, it is possible that alterations in hypothalamic/pituitary function and pubertal sex hormones caused by radiation therapy⁴⁸ may induce structural and functional reorganization of the brain,⁴⁹ adding another potential candidate cause of radiation-induced neurocognitive change following treatment.

Females treated for pediatric brain tumors also experience high rates of precocious puberty due to alterations in steroid production from treatment; however, precocious puberty is also associated with significantly better neurocognitive functioning compared with same-aged females not experiencing precocious puberty.⁵⁰ Therefore, another possible explanation for our findings of improved cognitive performance for females following treatment relative to males is that treatment-induced alternations to sex hormones may inadvertently be protective against neurotoxic effects of radiation.

Limitations

The current study is not without important limitations. First, patients in this study were not selected at random, though referred via standard clinical practice at a single institution. Patients were referred for neuropsychological evaluations from providers in neurosurgery and neuro-oncology based on each provider’s judgment that the patient could physically participate in the evaluation and that the patient would benefit from the evaluation. Thus, patients who were deemed neurologically devastated, underwent palliative care, or presented without any cognitive, academic, learning, or social/emotional challenges may not have been evaluated. It is certainly possible that our sample represents a somewhat limited swath of the overall patient population, with both significantly cognitively or physically devastated or very healthy patients not being referred for evaluation. Another limitation is regarding treatment characteristics that may moderate outcomes. Because the study was retrospective, we were not able to identify all relevant a priori variables and potential moderators of cognitive outcome. For example, it is likely that the effects of proton versus photon therapy, differential radiation doses, surgical variables (eg, shunt characteristics, extent of recovery from posterior fossa syndrome) or different chemotherapy regimens may have moderated or mediated cognitive outcomes. By including a wide age range across multiple tumor types and treatment types, we do feel that this clinical sample represents well the average patient who is treated at our institution. Lastly, we are fortunate to experience no-show rates of less than 5% in our neuropsychology clinic, which may limit the bias in our study sample by not excluding some families that get lost to follow-up.

Summary

The current study evaluated neurocognitive functioning at 2 timepoints in a relatively large sample of children treated for posterior fossa brain tumors. We observed a relative decrease for males and a relative improvement for females 3.7 years following treatment, even after controlling for age at treatment and age at radiation treatment. The results suggest that sex may moderate neurocognitive functioning following treatment. Potential protective mechanisms for females include increased rates of subcortical and cerebellar maturation compared with males as well as the role of sex hormones that may buffer against neurotoxicity of radiation were discussed. We also describe variables that could be included in future work that may provide a more complete algorithm from which to study the emergence of cognitive late effects. Such studies may provide the basis for more targeted treatments and better long-term outcomes for patients.

Funding

This work received partial funding through Pediatric Pilot Funds, Seattle Children’s Hospital, Center for Child Health, Behavior, and Development and Pediatric Brain Tumor Funds, Seattle Children’s Hospital.

Conflict of interest statement.

The authors report no conflicts of interest.

Authorship statement:

Critical revision: JB, DB, SL, DW Manuscript writing: JB, DB, SL Acquisition of data: JB, MB, SS, SL Study conception and design: JB, DB, MB, SS, RE, JO, DW, SL, RG Statistical analysis and interpretation of data: JB, DB, SL

References