Evolving therapies, neurocognitive outcomes, and functional independence in adult survivors of childhood glioma

Abstract

Background

Treatment of childhood glioma has evolved to reduce radiotherapy exposure with the goal of limiting late toxicity. However, the associations between treatment changes and neurocognition, and the contribution of neurocognition and chronic health conditions to attainment of adult independence, remain unknown.

Methods

Adult survivors of childhood glioma diagnosed in 1970-1999 in the Childhood Cancer Survivor Study (n = 1284; median [minimum-maximum] 30 [18-51] years of age at assessment; 22 [15-34] years from diagnosis) self-reported neurocognitive impairment and chronic health conditions. Multivariable models evaluated associations between changes in treatment exposures (surgery only, chemotherapy [with or without surgery], cranial radiation [with or without chemotherapy and/or surgery]), and neurocognitive impairment. Latent class analysis with 5 indicators (employment, independent living, assistance with routine and/or personal care needs, driver’s license, marital or partner status) identified classes of functional independence. Path analysis tested associations among treatment exposures, neurocognitive impairment, chronic health conditions, and functional independence. Statistical tests were 2-sided.

Results

Cranial radiation exposure decreased over time (51%, 1970s; 46%, 1980s; 27%, 1990s]. However, compared with siblings, survivors with any treatment exposure were at elevated risk for neurocognitive impairment, including surgery only (eg, memory: relative risk = 2.22; task efficiency: relative risk = 1.88; both P < .001). Three classes of functional independence were identified: independent (58%), moderately independent (20%), and nonindependent (22%). Cranial radiation was associated with nonindependence through impaired task efficiency (β = 0.06), sensorimotor (β = 0.06), and endocrine (β = 0.10) chronic health conditions and through the associations between these conditions and task efficiency (each β = 0.04). Sensorimotor and endocrine chronic health conditions were associated with nonindependence through memory.

Conclusion

Most long-term glioma survivors achieve adult independence. However, functional nonindependence is associated with treatment-related neurocognitive impairment and chronic health conditions.

Gliomas are the most common central nervous system (CNS) neoplasm in children, the majority of which are low-grade tumors (1). Despite excellent long-term survival for pediatric low-grade glioma (2,3), survivors are at risk of treatment-related morbidities, including neurocognitive deficits, reduced quality of life, chronic health conditions, and poor social attainment (4-9). Together, these morbidities substantially reduce the opportunity for survivors to live as functionally independent adults.

Frontline therapies for pediatric low-grade gliomas have evolved to reduce long-term morbidities while maintaining or improving survival (10). Surgical resection is often sufficient to achieve cure, but invasive surgeries are not without risk for subsequent neurocognitive, sensorimotor, neurologic, and endocrine dysfunctions (11-14). For low-grade gliomas where gross total resection cannot be achieved, or for recurrent low-grade gliomas, carboplatin-based chemotherapy regimens are increasingly used rather than cranial radiation therapy (RT) (2,15,16) because of reduced neurotoxicity (17). Conversely, cranial RT, which is commonly combined with surgery to treat high-grade gliomas, has been reduced and delayed over time for low-grade gliomas (18) given associated risk of physical and neurocognitive morbidities (7,8,18). Recent data demonstrated that the aforementioned treatment changes have reduced the incidence of chronic health conditions in survivors of childhood low-grade gliomas (19). However, the impact of evolving therapies on neurocognitive outcomes has yet to be investigated. Despite superior survival of low-grade gliomas compared with other pediatric CNS malignancies (ie, medulloblastoma), limited attention has been paid to long-term functional outcomes, including the attainment of independence in adulthood (20). Thus, the current study aimed to examine the impact of evolving cancer-directed therapies on neurocognitive functioning and assess the contribution of neurocognitive function and chronic health conditions to the attainment of functional independence in long-term survivors of childhood glioma.

Methods

Participants

The Childhood Cancer Survivor Study (CCSS) is a retrospective cohort with prospective survey-based follow-up of childhood cancer survivors who were diagnosed with common pediatric malignancies at age younger than 21 years, treated at 1 of 31 participating institutions between 1970 and 1999, and survived for 5 or more years (21). The study population included survivors who completed the follow-up 2 or follow-up 5 survey, which included a neurocognitive questionnaire completed by self- or proxy report. Among 14 498 eligible survivors, 1514 were diagnosed with childhood glioma. Because of the very low 5-year survival for high-grade gliomas, most tumors were presumed to be low-grade gliomas. The current study included survivors aged 18 years and older at time of survey completion with neurocognitive data. Survivors with genetic syndromes associated with neurocognitive impairment unrelated to primary cancer diagnosis were excluded (see Figure 1). Sibling controls were identified using the same criteria, except for childhood cancer diagnosis. The institutional review board at each center approved the study; all participants provided informed consent.

Treatment exposures

Treatment exposures were categorized in 3 mutually exclusive groups based on treatment received within 5 years of diagnosis: surgery only, chemotherapy (with or without surgery), or cranial RT (with or without chemotherapy and/or surgery). Dose of RT to the posterior fossa, temporal, frontal, and parieto-occipital lobe was quantified as the maximum dose when 50% or more of the region was irradiated (22) and categorized as none, more than 0-30 Gy, and more than 30 Gy. Treatment era was categorized by decade of diagnosis: 1970s, 1980s, or 1990s.

Primary outcome

Neurocognitive impairment was assessed using the CCSS Neurocognitive Questionnaire (CCSS-NCQ), a 25-item questionnaire validated in childhood cancer survivors (23). Neurocognitive problems were rated on a Likert scale ranging from 0 (never a problem) to 2 (often a problem) in 4 domains: task efficiency, emotional regulation, organization, and memory. Impairment was defined as scores in the 90th or higher percentile of sibling norms (6,24). Additional details regarding the CCSS-NCQ are provided in the Supplementary Methods (available online).

Secondary outcomes and covariates

Chronic health conditions and their age at onset were collected through surveys, and only conditions with onset before neurocognitive data collection were included. Chronic health conditions were classified by organ system and graded in severity using the National Cancer Institute’s Common Terminology Criteria for Adverse Events v4.0 (25). Conditions with potential impact on functional independence (24,26) were a priori selected: sensory-motor, hearing, vision, stroke, seizure and/or epilepsy, endocrine, pulmonary, and cardiac. For each organ system, the maximum grade was used to dichotomize chronic health conditions as grade 0-1 (none to mild) vs 2-4 (moderate to life threatening).

Functional independence was assessed using 5 indicators (20): 1) employment (full time: working full time, caring for home or family, student; part time: working part time; unemployed: not currently working, unemployed and looking for work, unable to work, retired), 2) marital status (history of marriage: married or living as married, no longer married or living as married, widowed, divorced, separated; never married: single, never married), 3) independent living (yes: lives with spouse or partner or alone; no: lives with parents or other relatives), 4) current driver’s license (yes or no), and 5) assistance with personal care and/or routine needs (such as eating, dressing, or everyday chores; yes or no).

Demographic (sex, age at survey completion, and race and ethnicity) and clinical (age at diagnosis, relapse or subsequent neoplasm after 5 years, and neurofibromatosis 1) covariates were used.

Statistical analyses

The χ² tests examined the distribution of treatment exposures and neurocognitive impairment by treatment era. Multivariable modified Poisson regression models evaluated risk of neurocognitive impairment associated with treatment exposures, region-specific dose of RT, and moderate to life-threatening chronic health conditions, adjusting for demographic and clinical covariates as appropriate. Generalized linear models compared risk of impairment in survivors and siblings, adjusting for demographic covariates.

Latent class analysis restricted to survivors aged 25 years and older (to permit time for attainment of independence) used the 5 indicators of functional independence. Models were fit with 1-5 classes, and multiple-fit indices were evaluated to select the optimal class number (27), with emphasis on adjusted Bayesian information criterion, entropy, and clinical meaning. The χ² tests examined the distribution of chronic health conditions across the classes of independence and the distribution of the classes of independence by treatment era.

Path analysis examined associations between treatment exposures and functional independence through chronic health conditions and neurocognitive impairment. Separate analyses were conducted for the moderately independent and nonindependent groups. To achieve parsimonious models, chronic health conditions were grouped based on potentially similar underlying mechanisms impacting neurocognition and functional independence: cardiovascular and pulmonary conditions, stroke and seizure and/or epilepsy, and hearing and vision conditions. An a priori causal model with direct and indirect pathways from treatment exposures to functional independence through chronic health conditions and/or neurocognitive impairment was hypothesized (Supplementary Figure 1, available online). Modified Poisson regressions examined the association of neurocognitive impairment and chronic health conditions with functional independence, adjusting for age at diagnosis, age at evaluation, sex, and race and ethnicity. Statistically significant predictors were included in the initial path models, one for each neurocognitive domain. Finally, each model was reduced by removing nonsignificant paths, iteratively one at the time until the best model-fitting criteria were achieved based on the comparative fit index and root mean square error of approximation. Because all analyses were based on a priori hypotheses, the results were not adjusted for multiple comparisons (28,29).

Statistical tests were 2-sided, and P values less than .05 were considered statistically significant. Analyses were conducted in SAS software v9.4 (SAS Institute, Cary, NC, USA) and Mplus 8.2 (30). More information about the statistical approach is available in the Supplementary Methods (available online).

Results

Sample characteristics

Among 1284 survivors of childhood glioma (48% male), the median (minimum-maximum) time from diagnosis was 22 (15-34) years, and the median age at survey completion was 30 (18-51) years. Age at diagnosis and age at assessment were correlated (r = 0.76), but not collinear. Of the survivors, 43% were treated with surgery only, 9% with chemotherapy (with or without surgery), and 38% with cranial RT [with or without chemotherapy and/or surgery; Table 1 (31)]. The remaining 10% could not be categorized into predefined treatment groups and were excluded from analyses including treatment exposures. Treatment with cranial RT was more common in the 1970s (51%) compared with the 1980s (46%) or 1990s (27%), and treatment with chemotherapy (0.4%, 5.0%, 14.8%, respectively) and surgery only (36%, 39%, 50%, respectively) increased during this period (Supplementary Table 1, available online, all P < .001; Supplementary Table 2, available online). Cranial RT in more recent eras was more likely to be delayed at least 1 year from diagnosis (1970s, 4%; 1980s, 11%; 1990s, 27%; P < .001). Among all study questionnaires, 82% were completed by survivors, and 56% were collected at follow-up 5.

Neurocognitive impairment

The prevalence of neurocognitive impairment differed by treatment exposure but less so by treatment era (Table 2). The impairment proportion was similar in survivors treated with immediate vs delayed (<1 vs ≥1 year from diagnosis) cranial RT (Table 2; all P > .05). Compared with siblings, survivors with any treatment exposure had elevated risk of impairment across most neurocognitive domains (Table 3). Among survivors, compared with treatment with surgery only, RT was associated with risk of impairment in each domain, with the highest risk for task efficiency and memory (both relative risk [RR] = 1.97; P < .001; Table 4). Similar risk of impairment was observed for RT compared with chemotherapy (Supplementary Table 3, available online). All effects remained statistically significant when adjusting for treatment era (Table 4; Supplementary Table 3, available online), with greater risk of neurocognitive impairment in more recent treatment decades.

Cranial RT to each brain region was associated with higher risk of neurocognitive impairment on all domains, with a dose response observed for most regions. For example, for the frontal lobe, RT greater than 0-30 Gy and RT greater than 30 Gy were associated with 28%-81% and 77%-146% excess risk of neurocognitive impairment, respectively (Supplementary Table 4, available online).

Neurocognitive impairment and chronic health conditions

All grade 2-4 chronic health conditions, except cardiac, were associated with risk of neurocognitive impairment (Table 5; Supplementary Table 5, available online). For example, sensorimotor conditions were associated with impaired task efficiency (RR = 1.48; P < .001), organization (RR = 1.32; P = .007), and memory (RR = 1.47; P < .001); endocrinopathies showed similar associations. Stroke (RR = 1.27; P = .03) and seizure and/or epilepsy (RR = 1.39; P < .001) were associated with memory impairment.

Classes of functional independence

Among 1015 (88%) survivors aged 25 years and older (Supplementary Table 6, available online), 3 classes of functional independence were identified (Supplementary Table 7, available online): independent (58%); moderately independent, nonindependent living, and unmarried or partnered (20%); and nonindependent (22%) (Figure 2; Supplementary Tables 8 and 9, available online). There was higher prevalence of grade 2-4 chronic health conditions among nonindependent survivors, except for pulmonary conditions (all P < .001; Supplementary Table 8, available online). The proportion of survivors in each class of independence did not change across treatment eras (all P > .05; Supplementary Table 8, available online).

Pathways from treatment exposures to functional independence

Path analyses revealed several indirect associations between cranial RT and functional nonindependence through chronic health conditions and/or neurocognitive impairment (see Figure 3 for statistically significant paths; Supplementary Table 10, available online, includes values for statistically significant and nonsignificant paths). For example, cranial RT was indirectly associated with nonindependence through impaired task efficiency (beta = 0.06; P = .02) and from sensorimotor (beta = 0.04; P = .001) and endocrine (beta = 0.04; P = .005) conditions through impaired task efficiency. Cranial RT also was indirectly associated with nonindependence through sensorimotor (beta = 0.08; P = .001) and endocrine (beta = 0.11; P = .002) conditions and through the associations between these conditions and impaired memory (sensorimotor: beta = 0.02; P = .02; endocrine: beta = 0.03; P = .02). Cranial RT was indirectly associated with nonindependence through endocrinopathies (beta = 0.11; P < .001) and through the associations between endocrinopathies and impaired organization (beta = 0.03; P = .008), although a larger proportion of variance was explained by the direct association between cranial RT and nonindependence (beta = 0.29; P < .001).

The moderately independent class demonstrated no direct or indirect associations between treatment exposures and independence through neurocognitive impairment (Supplementary Figure 2, available online).

Discussion

In a cohort including more than 1200 survivors of pediatric glioma diagnosed across 3 decades, 58% of long-term survivors attained full independence as adults. Among the remainder, failure to attain full independence was associated, in part, with neurocognitive impairment and comorbid chronic health conditions. Neurocognitive impairment was reported by more than half of long-term survivors of pediatric glioma in our sample. Any treatment exposure placed survivors at risk (up to four-fold) for neurocognitive impairment compared with sibling counterparts. Even surgery alone, which is considered a sparing treatment, was associated with 40%-122% risk of neurocognitive impairment. These results highlight the pervasive neurocognitive morbidity in long-term survivors of pediatric glioma and its adverse impact on functional outcomes in adulthood.

Arguably, the most important outcome for long-term survivors of childhood CNS tumors is the ability to achieve independence in adulthood. Our study of glioma survivors identified similar patterns of functional independence as a past study of survivors treated for a range of pediatric CNS tumors (20). However, a larger proportion (60% vs 40%) of survivors in our study achieved independence, potentially reflecting the limited use of multimodal therapy for gliomas (unlike medulloblastoma, for example). Nonetheless, the similar proportion of nonindependent survivors across 3 treatment decades is concerning, indicating limited improvements in real-life functional outcomes despite medical advancements. Our study also demonstrated the potential impact of neurocognitive problems and chronic health conditions on functional independence. Impaired task efficiency, memory, and organization were on the pathway between cranial RT and nonindependence through sensorimotor and/or endocrine conditions. In childhood cancer survivors, chronic health conditions have been shown to mediate the association between cranial RT and neurocognition (24,26), and neurocognition mediates associations between chronic health conditions and social attainment (32). The current study expands past findings by considering multilevel indirect associations with functional independence. Our results highlight real-world implications of associations between treatment exposures, chronic health conditions, and neurocognitive function. Furthermore, the task efficiency model supports the “double hit” hypothesis (26) whereby chronic health conditions may exacerbate neurocognitive deficits beyond the initial insult of early neurotoxic treatment exposures.

In our cohort, as expected, a smaller proportion of survivors received cranial RT in more recent decades, particularly in the first year of diagnosis, although the median RT dose was similar across decades. Concurrently, we observed expected increases in treatment with chemotherapy and surgery only over this time. Despite these therapeutic changes, we did not observe lower risk of neurocognitive impairment in more recent treatment decades or following delayed RT. Different internal standards used by respondents to evaluate and report neurocognitive problems (33) may explain this unexpected result, as self- or proxy-reported outcomes reflect day-to-day functional concerns rather than specific deficits measured through formal neurocognitive evaluation. Not surprisingly, cranial RT remained the strongest predictor of neurocognitive impairment among survivors in models with and without treatment era. These results suggest that changes in treatment for childhood gliomas between the 1970s and 1990s did not result in reduced prevalence of neurocognitive impairment. Importantly, these results may reflect selection bias, as a larger proportion of nonparticipants who died before completing the neurocognitive questionnaire were diagnosed in the 1970s (data not shown). In fact, improved survival among low-grade glioma survivors treated between 1970 and 1999 has recently been demonstrated (19). Although our results do not generalize to survivors treated in more recent eras, they underscore the known risk of any cranial RT exposure on long-term neurocognitive impairment, highlighting the importance of continued surveillance of neurocognitive functioning in contemporarily treated childhood glioma survivors.

In addition to unmodifiable (previously received) treatment exposures, the burden of chronic disease, which increases as survivors age (25), further contributes to the neurocognitive morbidity, thus identifying potentially modifiable health factors that may be amenable to intervention. In our cohort, moderate to life-threatening chronic health conditions affecting sensorimotor, endocrine, and neurologic systems, which are the most prevalent among glioma survivors (4), were associated with higher risk of neurocognitive problems. Sensorimotor conditions such as neuropathy may affect neurocognition through competing mental resources to cope with associated pain (34). Growth hormone deficiency may alter neurogenesis and brain maturation, resulting in structural brain abnormalities (35), whereas estrogen and testosterone deficiency may reduce neuroprotective actions triggered by these hormones (36). Stroke (37) and seizure (38), especially in their recurrent forms, may adversely affect neurocognition through neuroanatomical or microvascular damages. Importantly, treatment and management of chronic health conditions may positively affect neurocognitive functioning. For example, among brain tumor survivors, resolution of seizures has been associated with improved attention and memory and increased likelihood of marriage (32). Treatment for growth hormone deficiency may improve memory and attention (39,40) by stimulating growth hormone receptors in brain structures that underpin these functions (eg, hippocampus) (41-43).

As discussed above, our results may be subject to bias given increasing survival rates in this population (19) as our sample included twice the number of participating survivors in the 1980s and 1990s compared with the 1970s. Future research should investigate how novel chemotherapy agents (44,45) and radiation delivery approaches (46) may affect late outcomes. In particular, advancements in modern radiotherapy, including proton beam and intensity-modulated radiation, may result in better neurocognition for newly diagnosed patients (47). Ensuring equitable access to such therapies is imperative to prevent disparities in long-term outcomes (48). Although the use of questionnaires maximized data ascertainment from a large sample, the collection of all outcomes from a single source may generate common method variance, possibly biasing the observed associations (49). In addition, self-report measures are susceptible to recall bias and social desirability, limited by self-awareness, and may underestimate or overestimate the true burden of neurocognitive impairment (9) and chronic health conditions (50) in this population. The adequate psychometric properties of the CCSS-NCQ to assess daily functional concerns (23) support the validity of current findings, but discrepancies may arise from inclusion of self- and proxy report. Specifically, proxy-reported outcomes may bias results away from the null hypothesis because of the higher prevalence of neurocognitive problems reported by proxies (6). These discrepancies, however, may reflect actual greater difficulties of more disabled survivors (6). Furthermore, other cognitive-behavioral processes (eg, attention and depression) that share variance with the CCSS-NCQ constructs may influence neurocognitive outcomes. Direct assessment of neurocognitive function and health status of glioma survivors will be important to reduce the potential for shared method variance and enhance ascertainment of chronic health conditions to identify intervention targets and hypothesized pathophysiologic mechanisms. Unfortunately, tumor location is not available for the entire CCSS cohort, and thus its impact on outcomes could not be assessed. It is also possible that cranial RT was never received by survivors for whom delayed radiation was initially planned. Finally, path analyses do not establish causality (51), and the temporal associations among exposures and outcomes cannot be established from this cross-sectional study. However, our comprehensive a priori model builds on previous clinically meaningful mediation models in childhood cancer survivors (24,26,32) and provides some support for the proposed pathways. Given the similar level of functional independence observed across decades, future research should explore adaptive functioning [ie, the ability to adequately perform daily living tasks (52,53)] as a more sensitive predictor of functional independence that could be targeted in early interventions.

In summary, despite therapeutic changes, we did not observe reduced risk of neurocognitive impairment among glioma survivors treated in more recent decades. Chronic health conditions and neurocognitive impairment were associated with reduced ability to attain independence in adulthood. Survivors who develop neurocognitive problems and sensorimotor, endocrine, or neurologic conditions should be closely monitored, and early interventions may be warranted to facilitate attainment of social milestones and optimized independence.

Data availability

The Childhood Cancer Survivor Study (CCSS) is a US National Cancer Institute funded resource (U24 CA55727) to promote and facilitate research among long-term survivors of cancer diagnosed during childhood and adolescence. CCSS data are publicly available on dbGaP at https://www.ncbi.nlm.nih.gov/gap/ through its accession number phs001327.v2.p1. and on the St Jude Survivorship Portal within the St. Jude Cloud at https://survivorship.stjude.cloud/. In addition, utilization of the CCSS data that leverages the expertise of CCSS Statistical and Survivorship research and resources will be considered on a case-by-case basis. For this utilization, a research Application of Intent followed by an Analysis Concept Proposal must be submitted for evaluation by the CCSS Publications Committee. Users interested in utilizing this resource are encouraged to visit http://ccss.stjude.org. Full analytical data sets associated with CCSS publications since January 2023 are also available on the St. Jude Survivorship Portal at https://viz.stjude.cloud/community/cancer-survivorship-community∼4/publications.

Author contributions

Chiara Papini, PhD, MPH (Investigation; Visualization; Writing—original draft; Writing—review & editing), Sedigheh Mirzaei S., PhD (Formal analysis; Investigation; Writing—original draft; Writing—review & editing), Mengqi Xing, MR (Formal analysis; Investigation; Writing—original draft; Writing—review & editing), Ingrid Tonning Olsson, PhD (Conceptualization; Methodology; Writing—review & editing), Peter M.K. de Blank, MD, MSCE (Writing—review & editing), Katharine R Lange, MD (Writing—review & editing), Ralph Salloum, MD (Writing—review & editing), Deokumar Srivastava, PhD (Formal analysis; Investigation; Writing—original draft; Writing—review & editing), Wendy M Leisenring, ScD (Data curation; Methodology; Writing—review & editing), Rebecca R Howell, PhD (Data curation; Writing—review & editing), Kevin C Oeffinger, MD (Data curation; Writing—review & editing), Leslie L Robison, PhD (Funding acquisition; Project administration; Writing—review & editing), Gregory T Armstrong, MD, MSCE (Data curation; Funding acquisition; Methodology; Project administration; Writing—review & editing), Kevin R Krull, PhD (Conceptualization; Investigation; Methodology; Writing—original draft; Writing—review & editing), and Tara M Brinkman, PhD (Conceptualization; Investigation; Methodology; Writing—original draft; Writing—review & editing).

Funding

This work was supported by the National Cancer Institute (CA55727, GTA). Support to St Jude Children’s Research Hospital was also provided by the Cancer Center Support (CORE) grant (CA21765, C. Roberts) and the American Lebanese-Syrian Associated Charities (ALSAC).

Conflicts of interest

KCO and GTA, who are JNCI Associate Editors and co-authors on this paper, were not involved in the editorial review or decision to publish the manuscript. The authors have no conflict of interest to disclose.

Acknowledgements

The funders had no role in design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication.

A portion of these data were presented at the 20th International Symposium on Pediatric Neuro-Oncology, June 12-15, 2022, Hamburg, Germany.

References