Exercise training for neural recovery in a restricted sample of pediatric brain tumor survivors: a controlled clinical trial with crossover of training versus no training

Abstract

Background.

Exercise promotes repair processes in the mouse brain and improves cognition in both mice and humans. It is not known whether these benefits translate to human brain injury, particularly the significant injury observed in children treated for brain tumors.

Methods.

We conducted a clinical trial with crossover of exercise training versus no training in a restricted sample of children treated with radiation for brain tumors. The primary outcome was change in brain structure using MRI measures of white matter (ie, fractional anisotropy [FA]) and hippocampal volume [mm³]). The secondary outcome was change in reaction time (RT)/accuracy across tests of attention, processing speed, and short-term memory. Linear mixed modeling was used to test the effects of time, training, training setting, and carryover.

Results.

Twenty-eight participants completed training in either a group (n=16) or a combined group/home (n=12) setting. Training resulted in increased white matter FA (Δ=0.05, P<.001). A carryover effect was observed for participants ~12 weeks after training (Δ=0.05, P<.001). Training effects were observed for hippocampal volume (Δ=130.98mm³; P=.001) and mean RT (Δ=-457.04ms, P=0.36) but only in the group setting. Related carryover effects for hippocampal volume (Δ=222.81mm³, P=.001), and RT (Δ=-814.90ms, P=.005) were also observed. Decreased RT was predicted by increased FA (R=-0.62, P=.01). There were no changes in accuracy.

Conclusions.

Exercise training is an effective means for promoting white matter and hippocampal recovery and improving reaction time in children treated with cranial radiation for brain tumors.

With advances in treatment, most children now survive pediatric brain tumors,¹ which presents a new problem. Survivors of childhood brain tumors are left with neurological and psychological disabilities that yield long-term challenges.^2,3 Although cognitive rehabilitation appears very promising for survivors of pediatric cancer,^4,5 there are no primary therapies for directly promoting neural recovery in long-term survivors. Exciting work in regenerative medicine offers new hope.

Animal models show that exercise can activate neural stem cells and other processes that induce the growth of neurons and glial cells.^6,7 In humans, higher fitness levels predict increased organization of white matter, larger hippocampal volumes, and better cognitive function in healthy children, adolescents, and adults,^8,9 adults with multiple sclerosis,¹⁰ and pediatric brain tumor survivors.¹¹ Clinical trials have demonstrated that aerobic exercise training can foster brain growth, slow cognitive decline in normal and pathological aging,^12–14 and enhance brain and cognitive function in healthy and obese children.^15,16

Given the promising results linking exercise with brain health, it is surprising that limited work has examined its therapeutic effects for brain recovery in pediatric brain tumor survivors. Although exercise training has been linked to improved cognitive performance in adults with traumatic brain injury, stroke, and multiple sclerosis, these trials have not assessed changes in the brain.¹⁷ Children treated with life-saving cranial radiation for a brain tumor are an excellent population to study interventions for neural recovery. They experience intellectual decline, difficulties in attention, slowed speed of thinking, and memory impairment.¹⁸ These deficits are long-term and are related to white matter damage and hippocampal atrophy.^19,20 Radiation inhibits hippocampal neurogenesis.^21,22 Notably, white matter tracts and the hippocampus—structures most affected by radiation—are niches for neural precursor cells.^23,24 We conducted a clinical trial of exercise training in survivors of pediatric brain tumors treated with cranial radiation. Our primary objective was to test whether exercise training leads to changes in brain structure. Our secondary objective was to test whether exercise leads to improvements across tests of attention, processing speed, and short-term memory. We hypothesized that children would show increased organization of white matter, larger hippocampal volume, increased accuracy, and/or decreased reaction time (RT) across cognitive tests following exercise training as compared to no training. If successful, our trial would provide evidence that exercise training has potential for fostering neural recovery in childhood brain tumor survivors.

Importance of the Study

Considerable literature exists demonstrating that exercise training fosters brain health and slows cognitive decline in normal and pathological aging. There has been limited translation of these trials for neural recovery following brain injury or in pediatric populations. Our trial is a critical step in this translation for children with significant injury from their treatment for a brain tumor. Exercise is readily accessible and, as such, provides a low burden and inexpensive therapy for neurorecovery. Moreover, there is greater opportunity to generalize the benefits of a physical activity approach to the community than many other medical and rehabilitation programs.

Methods

Study Design and Participants

We conducted a controlled clinical trial with crossover (Fig. 1) at two children’s hospitals (ClinicalTrials.gov, NCT01944761). Within each block, participants were assigned to start 12 weeks of either (a) no training or (b) exercise training. Full crossover then occurred whereby participants completed a second period of 12 weeks in the opposite training condition. Allocation to each training order was conducted quasi-randomly based on the order in which a participant was recruited for each training block. Measures of brain imaging, cognition, and fitness were obtained at study entry (baseline), at ~12 weeks immediately prior to crossover (period 1 assessment), and at ~24 weeks upon completion of the study (period 2 assessment).

Fig. 1.

Consort Diagram. Screening, group assignment, and completion of assessments and training. Eligible participants were identified via database review. Measures of brain imaging, cognition, and fitness were obtained upon study entry (Baseline), at ~ 12 weeks when half of the participants had completed training (Period 1 assessment), and at ~24 weeks following crossover and the remaining participants completed training (Period 2 assessment). The 4 participants who consented but did not complete training were not included in the analyses.

Participants were screened for eligibility using existing registries at both sites. Eligible participants were between 6 and 17 years of age, declared English as their native language or had at least 2 years of schooling in English at the time of their baseline assessment, were diagnosed with a hemispheric or posterior fossa tumor and treated with cranial or craniospinal radiation, and were between 1 and 10 years from conclusion of treatment. Participants were not eligible if they were being treated for recurrent disease or had severe neurological/motor dysfunction that precluded safe participation in an exercise program. The institutional review boards at each site approved the study protocol. Either written informed consent or assent and parental consent (where applicable) was obtained. Please see the Supplemental Materials online for details regarding participant identification, recruitment, consent, and allocation.

Procedures

The exercise training program was provided in either a group (n=16) or a combined group/home (n=12) setting. Participants were allocated to the group or combined training setting in a quasi-random manner based on the order in which they were recruited to each training block. The group setting consisted of three 90-minute sessions per week for 12 consecutive weeks. The combined setting consisted of two 90-minute group sessions and two 30-minute individual home-based sessions per week. All sessions included aerobic activities designed to increase participants’ heart rate. Heart rate was measured using heart-rate monitors. The goal of all training was to increase and maintain each child’s heart rate for at least 30 minutes per session at a minimum of 80% of the achieved peak heart rate during baseline fitness. During no training, participants were instructed to continue their normal routine of physical activity. A complete description of the training program is provided in the Supplemental Materials online.

Outcomes

The primary outcome was MRI measurements of brain structure, including white matter architecture and hippocampal volume, using diffusion tensor imaging and anatomical T1 sequences, respectively. Fractional anisotropy (FA) maps were calculated from diffusion data and provided an index of white matter architecture based on water molecule displacement. The left and right hippocampi were manually traced on the T1-weighted image.^19,25 Sixteen scans were randomly selected and retraced by a second rater for interrater reliability. Both raters were blind to the group status of each participant. Raw hippocampal volumes were calculated and corrected for intracranial volume using a regression-based technique.²⁶

The Cambridge Neuropsychological Test Automated Battery (CANTAB) is a well-validated computerized tool that has been used with multiple pediatric populations and as a secondary outcome.²⁷ We employed subtests measuring attention (Rapid Visual Information Processing, Match to Sample Visual Search), processing speed (Simple Reaction Time, Choice Reaction Time), and short-term memory (Delayed Matching to Sample). Participants’ accuracy and RT were averaged across all subtests.

We used the 6-Minute Walk Test (6MWT), which measures the distance covered during a 6-minute period as a proxy for physical fitness, and the Bruininks-OseretskyTest of Motor Performance (2nd Edition) (BOT-2) to measure gross motor function (see the Supplemental Materials online for a detailed description of all outcome measures).

Statistical Analysis

We used linear mixed modeling to examine the effects of time, training, training setting, and training carryover effects on (1) white matter FA, (2) hippocampal volume, (3) mean accuracy, and (4) mean RT. We calculated difference scores for each outcome measure with the following contrasts: Period 1 assessment minus Baseline, and Period 2 assessment minus Baseline. In the case of FA, difference maps were created first, and mean FA difference scores were then extracted from areas of significant change from immediately before and after exercise training across all participants identified from a longitudinal Tract-Based Spatial Statistics (TBSS) analysis.^28,29 These areas were then used as regions of interest to extract mean FA difference scores for linear mixed model analyses. Time (First period, Second period), Training (No Training/Training), Training Setting (Group/Combined), and Carryover (No Carryover/Carryover) were included as categorical fixed effects variables, and Participant was included as a random effects variable. We controlled for baseline measurement/performance and also assessed fixed effects adjusting separately for sex, handedness, age at baseline assessment, age at diagnosis, and time since diagnosis. We further conducted sensitivity analyses where we replaced missing data for our primary outcomes using multiple imputation, imputing on sex, baseline neurological exam findings, baseline IQ, tumor location, extent of resection, number of surgeries, frequency of hydrocephalus requiring cerebrospinal fluid (CSF) diversion, frequency of mutism, treatment with chemotherapy, and type of radiation field. Results are presented with 95% confidence intervals (CI). Finally, correlation analyses were conducted to examine relations between fitness (6MWT) and outcome measures, and hierarchical regression modeling was used to assess the relationship between our primary and secondary outcome measures. A 2-sided P value of < .05 indicated statistical significance. A description of the modeled variables and specific technical terms can be found in the Supplemental Materials online.

If exercise promotes brain recovery, it would then be expected that the brain structure of our participants should more closely resemble the healthy brain following training. We compared FA and hippocampal volume for our trial participants to a cohort of healthy children from an existing image bank who did not participate in exercise training and were only scanned at a single time point. We compared our participants’ MRI scans prior to and after exercise training to the relevant scan of the age-matched children. TBSS was used to examine group differences in FA, and an ANOVA was used to test for differences in hippocampal volume. By using the data from this cohort, we could normalize the participant scans to the healthy pediatric brain.

Results

The trial was conducted from February 2011 to December 2014 at Hospital for Sick Children and McMaster Children’s Hospital (see Fig. 1 for Consort Diagram). During this time, 166 patients were screened: Of 102 eligible patients, 32 participants consented to the trial, and 28 participants (see Table 1 for Participant Characteristics) completed exercise training. Participants did not differ from eligible brain tumor survivors who did not participate (n=51 for whom data were available) on tumor type, tumor location, frequency of hydrocephalus requiring CSF diversion, frequency of mutism, and type of radiation field. Fewer participants in the program were treated with chemotherapy (P=.03), and it appeared they were younger at diagnosis (P=.06) and that fewer had gross total resections (P=.07) compared with eligible brain tumor survivors who did not participate (Supplementary Table 1 online). Participants in the group training showed a greater frequency of cerebellar signs upon neurological exam prior to the program than those in the combined training setting (P=.03). No differences were observed between these groups in any other medical/demographic variables or baseline measures (Supplementary Table 2 online).

Table 1.

Characteristics of study participants

	No Training First (n=12	Training First (n=16	P value
Sex (male)	7 (58.3%)	9 (56.3%)	.63
Handness (right)	10	15	.53
Age at diagnosis (y)
Mean	6.33	5.61	.35
Standard deviation	1.56	2.61
Range	2.92–8.08	1.92–9.33
Age at baseline assessment (y)
Mean	12	11.19	.36
Standard deviation	3	2.98
Range	8.08–16.92	7.67–16.92
Time from diagnosis to baseline assessment (y)
Mean	5.88	5.53	.77
Standard deviation	3.41	2.38
Range	1.50–10.42	1.08–8.58
Bruininks-Osteretsky Test of Motor Performance - 2 at Baseline
Mean Body Coordination T Score	35.08	27.94	.07
Standard deviation	10.08	9.22
Mean Strength and Agility T Score	33.5	28.69	.15
Standard deviation	7.26	9.46
Most Recent Neurological Exam at Baseline a
Cerebellar Signs (ataxia, dysmetria, dysdiadochkinesia)	66.70%	50.00%	.37
Hemiparesis	16.70%	12.50%	.75
Cranial nerve deficit	0%	12.50%	.2
Nystagmus	0%	25.00%	.06
Most Recent Intellectual Exam at Baseline b
Mean FSIQ (SS)	84.16	83.38	.92
Standard deviation	21.98	21.26
Scanner Type (3T)	7	11
Tumor Type			.5
Anaplastic astrocytoma	0	1
Ependymoma	1	1
Anaplastic ependymoma	1	3
Medulloblastoma	8	8
Pineoblastoma	0	1
Sarcoma	1	0
Germ cell	0	2
Astroblastoma	1	0
Tumor Location
Supratentorial	2	2	.38
Subtentorial	10	14
Gross Total Resection	5	7	.68
Number of Surgeries			.21
1 surgery	7	10
2 surgeries	4	4
3 surgeries	1	2
Radiation Type			.32
Focal (5400–5940 Gy)	4	5
Craniospinal (2340–3600 Gy) + Boost (1800–3240 Gy)	8	10
Periventricular (2100–3000 cGy)	0	1
Chemotherapy			.31
None	3	1
ACNS-0121 (caboplatin, cyclophosphamide, vincristine, etoposide)	0	6
A9961 (vincristine, lomustine, cisplatin)	2	2
COG9631 (etoposide, cisplatin, cyclophosphamide, vincristine)	1	0
COG99703 (thiotepa, carboplatin)	0	1
ICE (carboplatin, ifosfamide, etoposide)	1	0
SJMB96 & SJMB03 (vincristine, cisplatin, cyclophosphamide)	5	5
CARE (carboplatin, etoposide)	0	1
Hydrocephalus at diagnosis			.8
No hydrocephalus	3	5
Hydrocephalus with no treatment	4	6
Hydrocephalus requiring CSF diversion	5	5
Mutism following surgery c	3	4	.26
6-Minute Walk Test d
Period 1 – Baseline	-47±63	37±96	.12
Period 2- Baseline	-29±73	16±72

Abbreviations: CSF, cerebrospinal fluid; y, year(s).

^aNeurological exam conducted within a mean of 3.5 months (SD=3.5 mo) prior to baseline.

^bIntellectual Exam using Wechsler Scales conducted within a mean of 17 months (SD=10.9 mo) prior to baseline. Data were unavailable for 3 participants.

^cPatients were classified as having mutism if they had diminished speech output, linguistic difficulties, or dysarthria following surgery. Mutism is a transient dysfunction and had resolved in all participants by the time of baseline assessment.

^dOn the 6MWT, higher scores indicated higher functioning. Each participant was asked to walk as far as possible using a 25m hallway, without running, for 6 minutes, and distance was recorded in meters. Linear mixed modeling revealed a significant training effect for the 6MWT, P<.01

Prior to the trial, participants displayed impaired gross motor function, a high frequency of cerebellar motor deficits, and lowered Full Scale IQ (Table 1). There were no adverse effects of training, and adherence and retention rates were 84% and 100%, respectively (see Supplemental Materials online). Training significantly improved physical fitness across both training settings (Table 1, P=.01).

Independent of training, participants showed increased FA and hippocampal volume over time, with a significant slowing in growth from Period 1 to 2 for both measures (Ps=0.003 and 0.017, respectively). Training resulted in increases in FA (P< .001, Fig. 2) across the corpus callosum, cingulum, and superior longitudinal fasciculi bilaterally, and the right corticospinal tract and inferior frontal occipital fasciculus (Supplementary Table 3 online). A significant carryover effect was also observed (P< .001): Participants continued to show increases in FA even ~12 weeks after training had ended. These effects were observed across both training settings, and all effects remained significant after adjusting for all covariates and in post-hoc sensitivity analyses (Ps< .0.05). Younger age at diagnosis predicated greater increases in FA over the trial (R=-0.48, P=.012).^* The association between improved fitness measured with the 6MWT, and change in FA over the trial was not significant (R=.09, P > 0.10).

Fig. 2.

Change in fractional anisotropy (FA) following exercise training. (A) Clusters of significant increase in FA following training in all participants. Cluster size was thresholded at P<.05, which is family-wise fully corrected for multiple comparisons across space. Clusters are displayed in red-yellow with study-specific White Matter skeleton shown in blue and superimposed on FMRIB FA template. Images are in radiological convention. Numbers represent Montreal Neurological Institute (MNI) Z-coordinates. Please see eTable 1 for specific white matter bundles in which training effects were observed. (B) Change in mean FA difference scores extracted from regions (shown above) for main effects of time, training, training setting, and training carryover. Scores higher/lower than zero indicate change; positive scores indicate an increase in FA. (C) FA difference scores for individual participants. Participants assigned to no training → training are shown in green, and those assigned to training → no training are shown in orange. The exclusion of a single outlier did not change the results.

Because effects were consistent across hemispheres, we only report findings for total hippocampal volume. Training resulted in increased hippocampal volume for participants in the group training setting only (P=.001; Fig. 3). A carryover effect was observed for these participants as they continued to show increases in hippocampal volume ~12 weeks after training had ended (P=.01). All effects remained significant after adjusting for covariates and in post-hoc sensitivity analyses (Ps< 0.05).^† Improved performance on the 6 MWT and increased hippocampal volume over the trial were marginally related within the group training setting (R=0.43, P=.10).

Fig. 3.

Change in hippocampal volume following exercise training. (A) Illustration of hippocampal segmentation in sagittal and coronal plane. Images in coronal plane are shown in radiological orientation: red is the right, and blue is the left hippocampus. (B) Change in total hippocampal volume for the interaction of training x training setting and carryover x training setting; I bars indicate 95% confidence intervals. Scores higher/lower than zero indicate change; positive scores indicate an increase in volume. (C) Change in hippocampal volume for individual participants. Participants assigned to no training → training are shown in green, and those assigned to training → no training are shown in orange. The exclusion of a single outlier did not change the results.

Significant effects of time, training, and carryover were observed for RT in the group training setting only. Over time, participants showed increased RT, with a significant slowing from Period 1 to Period 2 (P=.03). In contrast, training in the group setting was associated with decreased mean RT (P=.04, Fig. 4). A significant carryover effect was also observed (ie, participants continued to show decreased RT ~12 weeks after training had ended [P=0. 05]). All effects remained significant after adjusting for covariates (Ps< 0.05). Handedness was a significant covariate (P=.02); a greater change in RT was observed in left- (n=3, mean=-361ms) versus right-handed (n=25, mean=-37ms) participants.^†

Fig. 4.

Change in reaction time (RT) on the Cambridge Neuropsychological Test Automated Battery (CANTAB) following exercise training. We assessed performance across 5 subtests of the CANTAB including rapid visual information processing, matching-to-sample visual search, simple reaction time, choice reaction time, and delayed matching-to-sample. Mean RT was calculated for each participant across all 5 subtests and using only correct trials. (A) Change in RT for the interaction of training x training setting and carryover x training setting; I bars indicate 95% confidence intervals. Scores higher/lower than zero indicate change; negative scores indicate a decrease in reaction time. (B) Change in Reaction Time for individual participants. Participants assigned to no training → training are shown in green, and those assigned to training → no training are shown in orange. With the exclusion of a single outlier, the training X training setting interaction was marginally significant (P=.07), and the carryover X training setting interaction remained significant (P< .05).

Improved fitness predicted decreased RT over the trial for participants in the group setting (R=-0.51, P=.05). Decreased RT following exercise training was associated with increased FA (R=-0.62, P=.01) and hippocampal volume (R=-0.52, P=.04) in these participants. Notably, increased FA contributed uniquely in accounting for decreased RT after hippocampal volume change had been considered; change in hippocampal volume did not contribute significantly when considered after FA (Supplementary Table 4 online). There were no significant effects for accuracy.

Trial participants were matched to 28 healthy children from an existing imaging data bank on age (mean age=11.87 y), sex (male/female=16:12), scanner used for imaging (3T/1.5T=20:8), and handedness (R/L=24:4). Our participants clearly displayed white matter insult prior to exercise training as FA was significantly lower than healthy controls across multiple voxels (Fig. 5 top panel). When we considered only those white matter tracts previously identified as showing training effects, FA was lower in 86% of the voxels for participants prior to training versus. healthy children (Supplementary Table 5 online). When we conducted the same analyses using participants FA maps after exercise training; only 74% of voxels in the participants differed from the same healthy children’s scans (ie, a normalization of FA for 12% of all voxels) (Supplementary Table 5 online). The remaining 14% of voxels did not change as they were not different between the groups prior to exercise training. The bottom panel in Fig. 5 provides visual examples of this normalization within the corpus callosum.

Fig. 5.

Tract-Based Spatial Statistics (TBSS) comparison map of trial participants versus healthy children. The top panel shows voxelwise differences between participants prior to exercise training and matched healthy children. Skeletons (blue) were overlaid on mean fractional anisotropy (FA) in FMRIB (MNI) space. Axial images are shown in radiological space with corresponding Z coordinates. Clusters (red) show reduced FA in trial participants versus healthy children. One large cluster consisting of 81 ∙892 voxels is evident, indicating widespread white matter compromise in trial participants prior to exercise training. The bottom panel shows differences in corpus callosum FA versus healthy children as a function of exercise training. Red indicates voxels of reduced FA in trial participants relative to healthy children. Blue indicates the underlying TBSS skeleton. Images (a) and (b) show reduced FA in trial participants in the genu and splenium of the corpus callosum. Images (c) and (d) illustrate fewer voxels with reduced FA in those same regions after exercise. Coordinates are noted in MNI space.

We also observed normalization of hippocampal volume following exercise. Prior to exercise training, bilateral hippocampal volume was significantly smaller in participants versus healthy children (3528 vs 3862mm³, 7.2% less, P=.02). Following exercise training, bilateral hippocampal volume was not significantly different between participants and the healthy comparison sample (3637 vs 3862mm³, 5.8% less, P=.08).

Discussion

We present novel evidence that aerobic exercise in a group setting can foster increases in white matter architecture, hippocampal growth, and faster reaction time on cognitive tasks in children treated with radiation for a brain tumor, irrespective of typical brain maturation or practice effects on the cognitive tasks. Previous exercise trials have shown increased hippocampal volume in healthy older adults,^12,30 increased white matter structure in adults with schizophrenia,³¹ increased organization of brain activation in overweight but otherwise healthy children,³² and improved cognitive performance in adults with acquired brain injury.³³ To our knowledge, ours is the first trial to provide neuroimaging evidence of exercise-related recovery in multiple regions of the brain well after injury, supporting the effectiveness of interventions long past the original trauma: The mean time from diagnosis to training was 5.25 years for our sample. Our findings are particularly striking when considering that pediatric brain tumor survivors treated with radiation display increasingly perturbed white matter, smaller hippocampi, slowed processing speed, and delayed cognitive development over time.^18–20 Exercise training, especially in a group setting, appears to mitigate these effects. Indeed, our sample displayed white matter and hippocampal insult, poor gross motor ability, cerebellar motor dysfunction, and cognitive deficit prior to participation in the exercise training. It is incredibly encouraging that exercise training is effective in promoting recovery in children with such impairment.

Exercise training likely engages multiple and synergistic mechanisms conducive to neural and cognitive recovery. Improved cardiorespiratory fitness from aerobic exercise induces a cascade of beneficial neural processes.³⁴ Animal models show that exercise increases gliogenesis, angiogenesis, and neurogenesis in neural precursor niches as well as the production of neurotrophic molecules involved in the protection and promotion of cell survival, neurite outgrowth, and synaptic plasticity. By harnessing these endogenous processes, exercise promotes brain recovery in rodent models of hypoxia,³⁵ radiation,³⁶ and traumatic brain injury.³⁷ Consistent with this, we found that increased fitness levels were significantly and marginally related to decreased RT and increased hippocampal volume, respectively.

The effects of exercise on hippocampal volume and reaction time were observed for the group but not the combined training setting. Notably, group training yielded more intense exercise than home-based training in our participants (see Supplemental Material online - Heart Rate). It may be that the higher intensity exercise in the group training setting underlies the hippocampal and cognitive effects. Group training may have provided a more enriched environment—including social stimulation and motivational support that resulted in more intense training—and this created optimal conditions for brain recovery. In rodents, running in a social environment leads to increases in hippocampal neurogenesis.³⁸

Cardiorespiratory fitness does not always predict brain structure or cognitive function in humans, however.³⁹ Consistent with prior work, we found that the relationship between fitness and white matter FA was not significant.⁹ Observed changes in white matter following training may instead reflect increased myelination elicited simply by neural activation underlying the complex motor behavior required to engage in physical exercise.⁴⁰ For example, we observed the majority of change in white matter architecture following training in the corpus callosum (Supplementary Table 3). Activity-dependent growth in callosal white matter has been documented following motor learning.⁴¹ Transcallosal pathways may be particularly sensitive to exercise training because physical activity requires a high degree of interhemispheric motor communication.⁴² Further, observed white matter growth in the cingulum may also reflect activity-dependent recovery: Engagement in exercise training requires attention, performance monitoring, and motor control (all functions that activate the cingulate gyrus).⁴³ The inclusion of nonexercise comparison groups focused on motor or other learning activities, and comparing training setting would help elucidate these issues in future studies. Likewise, in future studies it would be important to compare activities that involve minimal cognitive resources (ie, treadmill running; cycle ergometer) with activities that involve skill learning to determine if the training effects observed are related exclusively to increases in cardiorespiratory fitness or to activity-dependent changes associated with the recruitment of larger brain networks.

All white matter tracts in which we observed training-related change (including the corpus callosum, cingulum, superior longitudinal fasciculus, inferior-frontal-occipital fasciculus, and corticospinal tract) have been associated with processing speed in children.⁴⁴ It is compelling that the recovery in processing speed following training was mediated, at least in part, by white matter change: In group training, increased FA predicted decreased RT on tasks of attention, processing speed, and short-term memory. In typical development, white matter growth mediates improved processing speed with age.⁴⁴ Exercise training may recapitulate this normal development process. The hippocampus is a “neural hub,” ⁴⁵ and its growth may further enhance processing speed by facilitating communication across the brain. Future studies should include tasks sensitive to hippocampal functioning such as those focused on long-term and relational memory.⁴⁶

Critically, we found that training effects were not only maintained ~12 weeks after training had ended but that participants also continued to exhibit enhanced brain recovery. A notable observation was that decreases in RT were also observed 12 weeks after training had ended. However, it is not clear whether this reflects the training carryover effects or if participants continued to exercise on their own after the trial ended. A limitation of the current study is that we did not track participants’ level of exercise during the no-training period. It is also unclear whether such effects would be maintained beyond 12 weeks. It will be important to determine the longer-term impact of exercise as well as outcomes following longer and more frequent training periods. As is the case for many interventions focused on brain repair,⁴⁷ we acknowledge that our sample size was relatively small. We do note, however, that our sample size and/or participation rate are consistent with prior exercise trials in children.^15,16,32 Because of our relatively small sample size and that our sample of healthy children was only assessed at a single time point, our ability to determine whether the observed training effects reflects normal brain growth is limited. Further, we did identify selection bias in recruitment: we found that patients and families who needed to travel longer periods of time to participate in group sessions were underrepresented in our sample (see Supplemental Material online). In the development of future trials, it will be important to find ways to deliver exercise training programs that are more broadly accessible. In that regard, the use of technology (including accelerometers and physical activity monitors, exercise-related smartphone applications) and video exercise programs that could be performed independently and monitored remotely would reduce travel and time demands on patients and families and potentially increase participation rates.

Our findings support exercise training, especially in a group setting, as an effective, safe, and relatively low-cost therapy for fostering brain recovery following treatment with radiation for a brain tumor in childhood. Future studies are needed to clarify the effects of exercise in other pediatric and adult cancer populations and to determine the optimal type, intensity, and volume of exercise for brain recovery. Because exercise training is a relatively low-cost intervention and is straightforward to implement, it can be disseminated efficiently for pediatric care around the world, even in emerging economies and low-resource countries. With evidence that brain recovery and cognitive restoration are possible, there is new hope that we can mitigate the adverse long-term outcomes associated with pediatric brain tumors.

Supplementary material

Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/).

Funding

Canadian Institute of Health Research (202958); Canadian Cancer Society (2012–401423); Sunshine Kids Foundation; The W. Garfield Weston Foundation - Brain Canada Multi-Investigator Research Initiative (MIRI); and a postdoctoral fellowship from the Canadian Institute of Health Research (L.R.).