If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Psychiatry, Massachusetts General Hospital, Boston, MassachusettsDepartment of Psychiatry, Harvard Medical School, Boston, MassachusettsPsychiatric and Neurodevelopmental Genetics Unit, Center for Genomic Medicine, Massachusetts General Hospital, Boston, Massachusetts
Psychiatric and Neurodevelopmental Genetics Unit, Center for Genomic Medicine, Massachusetts General Hospital, Boston, MassachusettsStanley Center for Psychiatric Research, The Broad Institute of Harvard and MIT, Cambridge, Massachusetts
Department of Psychiatry, Massachusetts General Hospital, Boston, MassachusettsDepartment of Psychiatry, Harvard Medical School, Boston, MassachusettsPsychiatric and Neurodevelopmental Genetics Unit, Center for Genomic Medicine, Massachusetts General Hospital, Boston, MassachusettsCenter for Genomic Medicine, Massachusetts General Hospital, Boston, Massachusetts
Physical activity is associated with mental health benefits in youth. Here, we used causal inference and triangulation with 2 levels of biology to substantiate relationships between sports participation and dimensional psychopathology in youths.
Methods
Baseline data from the Adolescent Brain Cognitive Development (ABCD) Study, which recruited children from 9 to 10 years of age across the United States, were included in multilevel regression models to assess relationships between lifetime participation in team sports (TS), individual sports, and nonsports activities and Child Behavior Checklist (CBCL) scores. We calculated polygenic risk scores for 8 psychiatric disorders to assess interactions with sports exposure on CBCL scores among European descendants. Following rigorous quality control, FreeSurfer-extracted brain magnetic resonance imaging structural data were examined for mediation of CBCL–activities relationships.
Results
Among those with complete data (N = 10,411), causal estimates using inverse probability weighting associated lifetime TS exposure with a 1.05-point reduction in CBCL total (95% CI, −1.54 to −0.56, p < .0001) a relationship that was specific to TS and strengthened with more years of exposure. Associations of attention-deficit/hyperactivity disorder polygenic loading with CBCL total weakened in European children with TS exposure (n = 4041; beta = −0.93, SE = 0.38, p = .013). Furthermore, TS participation and lower CBCL each associated with increased subcortical volumes (n = 8197). Subcortical volume mediated 5.5% of TS effects on CBCL total.
Conclusions
Our findings support prior associations of TS participation with lower psychopathology in youths through additional studies that demonstrate specificity, dose response, and coherence across 2 levels of biology. Longitudinal studies that further clarify causal relationships may justify interventional studies of TS for high-risk youth.
). Childhood psychiatric symptoms are known to adversely affect adult functioning, even if they do not develop into full-fledged diagnoses or persist into adulthood (
Physical activity is a potentially modifiable factor that can influence mental health outcomes in children. A recent meta-analysis of pediatric and adolescent cohorts reported small but significant associations between sedentary behavior and poor mental health (
Role of physical activity and sedentary behavior in the mental health of preschoolers, children and adolescents: A systematic review and meta-analysis.
). Conversely, the positive impact of physical activity on mental health, especially on depressive symptoms and anxiety, has been suggested in cross-sectional and longitudinal studies of children and adolescents (
). Higher frequency, greater intensity, longer duration, and earlier initiation of sports involvement have also been associated with positive mental health outcomes. Echoing results from large adult cohorts (
), school-based studies have shown that team sports (TS) associate more positively with mental health than do individual sports (IS), especially among girls (
). A small number of randomized controlled trials evaluating physical activity interventions for children with attention-deficit/hyperactivity disorder (ADHD) and autism spectrum disorder (ASD) have also demonstrated benefits for executive function, motor skills, and inattention (
Chronic physical activity for attention deficit hyperactivity disorder and/or autism spectrum disorder in children: A meta-analysis of randomized controlled trials.
However, the strength of available evidence linking physical activity to mental health benefits in youth is limited by the small number of randomized controlled trials and the likelihood of residual confounding with measured or unmeasured variables (e.g., socioeconomic variables). Large data sets from diverse and well-characterized populations, such as the ongoing Adolescent Brain Cognitive Development (ABCD) Study, enable isolation of specific sports exposures versus other activities and control of potential confounders. Hoffmann et al. (
) associated TS exposure in the ABCD Study with significant reduction in a range of psychopathology, after controlling for numerous potential demographic confounders. However, a firmer case for causality can be made through triangulation with underlying biological mechanisms. For example, in an analysis limited to hippocampal volumes, Gorham et al. (
) found that variance volume mediated the relationship between TS and depression symptoms in the ABCD cohort. Broader exploration of mechanisms that could transmit potential neuroprotective effects of sports exposure against psychopathology, coupled with additional analyses that probe causality, could further justify interventional studies.
Here, we leveraged baseline (age 9–10 years) data from the ABCD Study to further validate the relationship between sports exposure and mental health, including both statistical methods to assess causal inference (dose effects and inverse probability weighting) and triangulation with 2 levels of biology (whole-brain magnetic resonance imaging [MRI] analysis and polygenic risk scores [PRSs]). Furthermore, because there are known sex differences in the expression of dimensional psychopathology in youth (
), timing of pubertal development, and activity participation rates, additional analyses considered whether associations between activities, Child Behavior Checklist (CBCL) scores, and underlying biological mediators differed by sex.
Methods and Materials
Study Design and Population
The ABCD Study enrolled 11,875 children of ages 9 to 10 across 22 U.S. sites. Institutional review board approval for the study is described in Auchter et al. (
). All parents provided written informed consent and all youths provided assent. Enrolled participants match the racial and ethnic composition of the United States and were enriched for siblings (including twins and other multiples). A full description of study procedures can be found elsewhere (
). Here, we used cross-sectional baseline data obtained directly from youth participants (brain MRI and genomics) and primary caregivers (demographics, developmental history, symptom inventories, sports, and activities) (Table S1).
Dimensional Psychopathology
The CBCL measures youth dimensional psychopathology during the preceding 6 months (
). The CBCL is a widely used survey that measures 113 items across 8 domains (aggressive behavior, anxious/depressed, attention problems, rule-breaking behavior, somatic complaints, social problems, thought problems, and withdrawn/depressed) as well as broad-band indices of internalizing, externalizing, and total psychopathology. We used CBCL total as the primary outcome because it captures diffuse psychopathology.
Sports and Other Activities
Youth participants’ physical activity was assessed with the Sports and Activities Involvement Questionnaire. We categorized these activities into 3 types: TS, IS, and nonsports activities (NSA) (see Supplemental Methods). While IS could entail team membership, they were categorized as individual due to their stronger reliance on individual-level results. These 3 sets of measurements were dichotomized to create a lifetime participation variable (yes/no). TS participation was also calculated as a continuous variable for number of activities × years participated (e.g., 2 years of baseball and 1 year of soccer totals 3 activity years).
Polygenic Risk Score
Genome-wide genotype data from individuals of European ancestry were used for PRS analyses. For every sibling pair, one was randomly selected for inclusion; see Supplemental Methods for additional quality control (QC)–based exclusions. With PLINK version 1.9 (
), single nucleotide polymorphisms were filtered according to minor allele frequency, data missingness, deviation from Hardy–Weinberg equilibrium, and linkage disequilibrium. Shapeit version 2 (
). In PLINK, the resultant posterior effect sizes were applied to individual-level genotype data to generate PRSs for each subject. PRSs were calculated for 8 specific psychiatric disorders—ADHD, ASD, anorexia nervosa, bipolar disorder, major depressive disorder (MDD), obsessive-compulsive disorder, schizophrenia, and Tourette syndrome—using summary statistics from the Psychiatric Genomics Consortium (
International Obsessive Compulsive Disorder Foundation Genetics Collaborative (IOCDF-GC) and OCD Collaborative Genetics Association Studies (OCGAS) Revealing the complex genetic architecture of obsessive-compulsive disorder using meta-analysis.
Cross-Disorder Group of the Psychiatric Genomics Consortium Genomic relationships, novel loci, and pleiotropic mechanisms across eight psychiatric disorders.
Brain T1-weighted images were obtained at 3T on Siemens MAGNETOM Prisma, GE Healthcare Discovery MR750, and Philips Achieva scanners. Detailed scanning methods are described elsewhere (
). For this analysis, we downloaded minimally processed volumes from the National Institute of Mental Health Data Archive, which we subjected to strict QC through inspection of each scan, as described in Supplemental Methods. To correct low frequency intensity nonuniformity, we used the N4 bias field correction algorithm (
). Cortical reconstruction and volumetric segmentation were performed with the FreeSurfer image analysis suite, version 7 (http://surfer.nmr.mgh.harvard.edu/). We extracted cortical region of interest (ROI) values (surface area, thickness, and volume), subcortical volumes, and related global values (mean surface area, mean cortical thickness; and total cortical, subcortical, and gray matter volumes) with the Desikan-Killiany cortical atlas (
). We took a midpoint of total combined family income categories (see Supplemental Methods) and divided by its corresponding federal poverty level, according to household sizes.
Puberty Scale
We used the body hair scale from the Pubertal Development Scale and Menstrual Cycle Survey History from both parents and youths as indicators of puberty stage. Their responses were recorded in 4 categories (see Supplemental Methods for additional details).
Statistical Analysis
TS and Psychopathology
First, we assessed associations between lifetime activity participation and dimensional psychopathology. Then, we tested a TS-focused model with CBCL total score as the primary outcome (model 1). Exploratory analyses separately tested associations between TS and individual and broad-band CBCL scores, using the false discovery rate (FDR) to correct for 10 comparisons. Multilevel regression models were adjusted for age, sex, race, ethnicity, INR, and puberty scale as fixed effects and family identifier (ID) and site ID as random effects. Sensitivity analyses assessed specificity of TS association among 8 activity participation patterns using 3-way interaction among 3 types of activities (TS × IS × NSA) with CBCL total, as well as dose dependency of TS participation, based on activity years. Additionally, we used inverse probability weighting (IPW) methods to estimate causal effects of TS participation on psychiatric symptoms. With IPW, we made a pseudopopulation by weighting individual data, using the probability of TS exposure. The same set of variables was used to calculate stabilized inverse probability weights of TS participation. IPW enabled us to minimize confounding by measured covariates, so we could compare the weighted averages of CBCL scores between TS+ and TS− groups.
Genetics
Next, we assessed whether lifetime TS participation could offset genetic risk for psychopathology among participants of European ancestry. We first compared the mean PRS of each of the 8 psychiatric disorders (see above) by activity participation. In exploratory analyses, we regressed these 8 PRSs on CBCL total, using FDR to correct for 8 comparisons, adjusting for age, sex, INR, and puberty scales; their interactions with PRS and TS; and 5 principal components as fixed effects and site ID as a random effect (model 2). For those PRSs which were significantly associated with CBCL total, we then included TS and its interaction with PRS into the regression models (model 3). Sensitivity analyses added IS and NSA as well as their respective cross-product interactions with PRS. In instances when we identified a significant interaction term (i.e., statistical interaction between the PRS and TS on CBCL total), we also ran separate regression models to examine possible interactions on each of the CBCL syndrome-specific and broad-band scales.
Imaging
We tested whether brain volume features mediated the effects of TS participation on dimensional psychopathology, using an iterative set of analyses that focused first on global and then regional MRI data. First, we examined associations between 6 global MRI measurements (mean cortical thickness, total surface area, cortical volume, subcortical gray matter volume, cerebellum gray matter volume, and total gray matter volume) and CBCL total (model 4) and associations between TS participation and the same 6 global MRI measurements (model 5), using multilevel linear regression models with FDR correction for 6 comparisons. For global MRI measures that associated with both CBCL total and TS exposure, we conducted mediation analyses (
) to determine whether brain volumes mediated effects of TS exposure on CBCL total score (model 6). Mediation analysis decomposed the total effect into 2 components, the causal mediation effects and the direct effects. Furthermore, to achieve greater regional specificity, global MRI measurements that associated with both CBCL total and TS exposure were subject to ROI-based analyses, using the same approach as models 4 and 5. ROIs showing nominally significant relationships in both models were then subject to mediation analyses, as in model 6. For all analyses, we included age, sex, race, ethnicity, INR, number of surface holes, puberty scale, and estimated total intracranial volume (only for ROI values) as fixed effects and family ID, site ID, and MRI scanner type as random effects.
Sex Differences
Finally, to explore potential sex differences, we included an additional interaction term (sex × independent variable of interest) in models 1, 3, 4, and 5 (i.e., model 1s, 3s, 4s, and 5s).
Statistical analyses were conducted with R version 4.1.2 (https://www.R-project.org/). We used the lme4 package (
) for multilevel regression model, geepack package (https://cran.r-project.org/package=geepack) for IPW analysis, and mediation package (https://cran.r-project.org/web/packages/mediation/vignettes/mediation.pdf) for mediation analysis. All continuous covariates except CBCL scores were z-standardized prior to analysis. We used two-sided p < .05 as the significance threshold for all analyses. The same threshold was used for the between model FDR correction in exploratory analyses.
Results
Demographics
Among 11,875 ABCD Study participants, 9638 participants had complete data for all covariates, 4014 met the criteria for PRS calculation, and 8197 cleared MRI QC rating criteria, without missing data (Figure S1). Among 9638 youths, 6201 (64.3%) had lifetime TS participation, 6668 (69.2%) had IS exposure, and 5577 (57.9%) had NSA exposure. Demographics of the study population are shown in Table 1 and Table S2.
Table 1Demographics by Lifetime Team Sports Participation (n = 9638)
Among the 3 types of activities, only TS was significantly and negatively associated with CBCL total T score (Table S3a), echoing results of Hoffmann et al. (
), but now using linear rather than negative binomial regression. In model 1, we observed a significant negative association between lifetime TS and CBCL total T score (beta = −0.93 [Cohen’s d = 0.083], 95% CI, −1.41 to −0.44, p = .00016) (Table S3b). Among the CBCL subscales in exploratory analyses, anxious/depressed, withdrawn/depressed, social problems, thought problems, attention problems, and internalizing scores were negatively associated with TS status (Table S4).
A series of novel sensitivity analyses consistently supported the findings in model 1. The negative association between lifetime TS and CBCL total persisted when the 8 activity participation patterns (TS × IS × NSA) were added into the model (Figure S2; Table S3c). To assess dose response, we calculated the history of TS participation (number of sports type × years of participation). A gradual decrease in CBCL total score was observed as TS exposure activity year increased (beta = −0.271, 95% CI, −0.335 to −0.208, p < .0001) (Figure 1). With IPW, we estimated causal effects of TS on CBCL total score. Lifetime TS exposure was estimated to reduce CBCL total T score by 1.05 points (95% CI, −1.54 to −0.56, p < .0001).
Figure 1Team sports (TS) participation history and Child Behavior Checklist (CBCL) total. Estimated marginal means of CBCL total score by activity years category (0, 1–3, 4–6, 7–9, 10+ activity years) (n = 9638) (p < .0001). Error bars indicate 95% CIs. Covariates included age, sex, race, ethnicity, income-to-needs ratio, puberty scales (fixed effects), family identifier, site identifier (random effects).
The 8 psychiatric disorder PRS values were unassociated with activity participation, except lower mean ASD PRS in TS participants and lower ADHD PRS in IS participants (Table S5). Consistent with a recent report, which controlled for other indices of socioeconomic status but not for INR (
), PRSs for ADHD, MDD, and ASD positively associated with CBCL total score in the exploratory analyses, while no such association was reported for other PRSs after FDR correction for multiple comparisons (model 2) (Table S6a).
Including a cross-product interaction term (TS × PRS) in model 3 for ADHD, MDD, and ASD PRSs indicated that only ADHD PRS had a nominally significant interaction with TS in association with CBCL total score. While ADHD PRS positively associated with CBCL total across TS participation status, the slope for TS+ youths was smaller than that of TS− youths (beta = −0.93, SE = 0.38, p = .013) (Figure 2; Table 2). Notably, this interaction survives correction for 3 PRSs that showed a main effect on CBCL total, but not for all 8. Another sensitivity analysis that also included IS and NSA and their interactions with PRS retained interaction of TS and PRS on CBCL total and showed an oppositely signed interaction of NSA and PRS on CBCL total (Table S7).
Figure 2Interactive effect of team sports exposure and attention-deficit/hyperactivity disorder (ADHD) polygenic risk scores (PRSs) on Child Behavior Checklist (CBCL) total. Red line: association between ADHD PRS (z-standardized) and CBCL among children with no lifetime team sports participation. Green line: association between ADHD PRS and CBCL among children with team sports participation. Covariates included age, sex, income-to-needs ratio, puberty scales, and 5 principal components as fixed effects and site identifier as a random effect (n = 4014).
Among CBCL subscales, ADHD PRS associated positively with all the CBCL subscales, except for anxious/depressed and withdrawn/depressed, after FDR correction (Table S6b); however, ADHD PRS did not significantly interact with TS on any of the CBCL subscales (Table S8).
Magnetic Resonance Imaging
CBCL total score associated significantly and inversely with all 6 global volume measurements except cortical thickness (model 4) (Table S9a). Among global MRI indices, only subcortical gray matter volume was significantly associated with TS after FDR correction (beta = 0.061, SE = 0.209, p = .0035) (model 5) (Table S10a). These associations were robust to sensitivity analysis with IS and NSA included in the model (Tables S11a and S12a). Subsequent mediation analyses indicated a significant mediating effect of total subcortical volume on the relationship between TS exposure and CBCL total score, with total subcortical volume accounting for 5.5% of the variance in this relationship (model 6) (Figure 3).
Figure 3Mediation of team sports exposure effects on Child Behavior Checklist (CBCL) total via subcortical volume. Estimated total effect of team sports on CBCL total was −0.792 points. Five and a half percent of the total effect (beta = −0.002) was mediated by subcortical volume and direct effect of team sports was −0.748 points (n = 8197). Covariates included age, sex, race, ethnicity, income-to-needs ratio, number of surface holes, puberty scale, and estimated total intracranial volume as fixed effects and family identifier, site identifier, and magnetic resonance imaging scanner type as random effects.
In post hoc analyses of subcortical ROI, nominally significant associations emerged between CBCL total and 9 ROIs, although none reached FDR significance (model 5) (Table S9b). For TS, among subcortical ROIs, volume in 8 total regions associated with TS exposure, reaching FDR significance in 2 of them (left thalamus and left pallidum) (model 4) (Table S10b). We also tested mediation effects of subcortical ROIs that met inclusion criteria (nominally significant association both with TS and with CBCL). The left cerebellum cortex, left hippocampus [consistent with Gorham et al. (
)], and right accumbens area demonstrated significant mediation effects on CBCL total by TS (Table 3).
Table 3Mediation Effects by ROI on Estimated Causal Effects of Team Sports on CBCL Total Score (Model 6) (n = 8197)
Region
Estimate
Lower 95% CI
Upper 95% CI
p Value
Subcortical Volume
Proportion mediated: 0.0545 (95% CI, 0.016–0.19)
ACME
−0.044
−0.08
−0.01
.010
ADE
−0.748
−1.28
−0.27
.004
Total effect
−0.792
−1.32
−0.31
.002
Left Cerebellum Cortex
Proportion mediated: 0.0378 (95% CI, 0.004–0.14)
ACME
−0.031
−0.06
−0.00
.034
ADE
−0.753
−1.30
−0.23
.004
Total effect
−0.783
−1.33
−0.27
<.001
Left Hippocampus
Proportion mediated: 0.0379 (95% CI, 0.006–0.14)
ACME
−0.031
−0.06
−0.00
.024
ADE
−0.756
−1.28
−0.23
.002
Total effect
−0.787
−1.31
−0.26
<.001
Right Accumbens Area
Proportion mediated: 0.0381 (95% CI, 0.006–0.14)
ACME
−0.030
−0.06
−0.00
.016
ADE
−0.751
−1.30
−0.20
.010
Total effect
−0.781
−1.32
−0.23
.006
ROI values are z-standardized. ROIs that are nominally significant in both Tables S6 and S7 are listed. Left GC-ML-DG was the only left hippocampal subregion that showed nominally significant association between both TS—ROI and ROI—Child Behavior Checklist. Covariates: age, sex, race, ethnicity, income-to-needs ratio, puberty scales, estimated intracranial volume, surface holes (fixed effects), family identifier, site identifier, scanner type (random effects). Total effect is the sum of a mediation (indirect) effect and a direct effect.
ACME, average causal mediation effect; ADE, average direct effect; ROI, region of interest; TS, team sports.
There was a significant sex-dependent difference in TS and CBCL association represented with interaction term (model 1s) (beta = 1.10, 95% CI, 0.178 to 2.02, p = .019). TS participation associated with lower CBCL total among boys (beta = −1.57, SE = 0.38, p < .0001) but not among girls (beta = −0.46, SE = 0.327, p = .16) (Figure 4).
Figure 4Effects of team sports participation and Child Behavior Checklist (CBCL) total by sex. Box plots for estimated mean CBCL total scores are displayed (n = 9638). Covariates included age, race, ethnicity, and income-to-needs ratio as fixed effects and family identifier and site identifier as random effects.
Sex did not interact with ADHD PRS to influence CBCL total (model 3s) (beta = 1.27, SE = 0.22, p = .16), and there was no significant sex × TS × PRS interaction with CBCL total (beta = −0.85, SE = 0.74, p = .26).
None of the associations between global MRI indices and CBCL total were significantly different by sex (model 4s) (Table S13). There was no sex-dependent association between TS and global MRI indices (model 5s) (Table S14) or interaction between TS and MRI indices on CBCL (Table S15).
Discussion
Using cross-sectional data from 9- to 10-year-old participants in the ABCD Study, we examined the association between sports participation and dimensional psychopathology through novel validation analyses. First, using IPW, we demonstrated that the association of TS with lower CBCL scores is robust to potential confounders, such as socioeconomic status, study site, and puberty. Furthermore, associations of TS with psychopathology were strongest among individuals with more lifetime TS exposures, consistent with a dose effect. We then turned to analyses that examined TS-CBCL effects on 2 levels of brain biology. TS participation was associated with reduced effects of polygenic risk on psychopathology; and this relationship was mediated by volume differences in subcortical regions that themselves were associated with increased risk. These analyses of specificity, dose response, and coherence across 2 levels of biology provide convergent validation for the relationship between TS exposure and reduced psychopathology in youths.
These results provide biological context for beneficial effects of TS participation on children’s overall mental health. Meta-analyses of prospective studies indicated that physical activity has mental health benefits, regarding the prevention of adult depression and schizophrenia (
) reported the incident rate ratio of TS by CBCL psychopathology from year 1 cross-sectional data. Our results also specifically link TS (compared with IS or NSA) more strongly to mental health, using mixed-effect linear models to estimate effect size. This specific result could reflect synergistic social, psychological, and physical benefits of TS. Previous studies indicated that TS may provide benefits beyond individualized sports through positive social interactions with peers that provide social support and acceptance, more self-esteem, and less body dissatisfaction (
). A systematic review has found limited evidence of benefits in mental health for organized NSAs (e.g., arts or music) among children and adolescents (
), which is consistent with our findings. These results may explain the heterogeneity of previous reports on effects of physical activity and warrant further examination of the mechanism behind the benefits of TS, including its physical and social components. Even though the effect size for the association between TS and CBCL is small (Cohen’s d = 0.083), the population-level impact of TS-based intervention could be substantial, with fairly low equipment and other costs (
To our knowledge, this study is the first to implicate polygenic loading in the relationship between TS and psychopathology. Though interactions with specific polymorphisms such as within BDNF have been reported (
); more robust, genome-wide approaches have not yet been deployed to study whether TS participation reduces psychopathology symptoms among children with high polygenic risk. Consistent with prior work that did not include socioeconomic status as a covariate (
), we observed the association between PRS and CBCL scores for ADHD, ASD, and MDD PRSs; however, TS only mitigated the relationship between ADHD PRS and psychopathology. Among CBCL subscales, this interaction was most pronounced for thought problems, although interaction of TS and ADHD PRS only reached nominal significance. Notably, the effect size of this interaction was small, and while its statistical significance survives correction for 3 comparisons (i.e., ADHD, ASD, and MDD × TS), it would not have survived had we not constrained the interaction analysis based on PRSs showing a main association with CBCL total. The fact that the summary statistics from Psychiatric Genomics Consortium used in this analysis (
International Obsessive Compulsive Disorder Foundation Genetics Collaborative (IOCDF-GC) and OCD Collaborative Genetics Association Studies (OCGAS) Revealing the complex genetic architecture of obsessive-compulsive disorder using meta-analysis.
Cross-Disorder Group of the Psychiatric Genomics Consortium Genomic relationships, novel loci, and pleiotropic mechanisms across eight psychiatric disorders.
) were derived predominantly from adult participants may influence their impact on psychopathology and their relevance to gene-environment interactions in childhood. However, given the likelihood that polygenic risk will continue to influence emergent psychiatric symptoms later in adolescence, longitudinal studies of youth TS participation on mental health outcomes against the background of polygenic risk are warranted.
Several previous MRI studies have associated physical activity in youths with increases in global brain volumes—whole brain (
A whole brain volumetric approach in overweight/obese children: Examining the association with different physical fitness components and academic performance. The ActiveBrains project.
). While the cellular biology underlying these associations remains unknown, increased cytogenesis and neuroplasticity have been associated with exercise, through a variety of specific mechanisms, in animal models (
). Two such possibilities include known effects of both physical activity and enriched environment on hippocampal neurogenesis, with potential downstream effects in other subcortical and cortical regions (
). Future translational studies may clarify region- and development-specific effects on sports participation on brain biology. Moreover, we found that these subcortical volumes mediate the relationship between TS and CBCL, extending the previously reported analysis that focused on hippocampal volumes (
). These results are consistent with the idea that TS participation alters brain development in regions that are relevant to psychopathology risk, although fully prospective studies would better enable causal inference.
The finding that both TS and ADHD PRSs associated more strongly with dimensional psychopathology in boys than in girls extends previous literature on sex differences in how youth sports participation influences psychopathology. A prior study indicated a similar association among boys, with TS exposure predicting fewer internalizing problems and less depressed mood (
). Thus, considering our results and data from these previous studies, more work is needed to connect exercise, the mechanism underlying resilience, psychopathology, and sex-dependent outcomes.
Studies of psychopathology and MRI phenotypes in children can be confounded by a number of factors, including socioeconomic differences, puberty stage, and motion-related and other MRI artifacts. As an index of socioeconomic status, we adjusted for INR, a measure that incorporates U.S. poverty statistics and household size. In the ABCD cohort, INR captures effects of numerous individual economic and psychosocial factors that have been associated with neurodevelopmental differences (
Similar but distinct – Effects of different socioeconomic indicators on resting state functional connectivity: Findings from the Adolescent Brain Cognitive Development (ABCD) Study®.
). We additionally adjusted for puberty, given previously reported effects of pubertal status on psychopathology and brain morphology, especially subcortical volumes (
). For analyses with MRI outcomes, we filtered out low quality images by manual QC of all 11,263 T1 scans available after excluding scans needing clinical consultation or failing the FreeSurfer processing pipeline. Furthermore, we adjusted for the number of surface holes to eliminate bias due to image and segmentation quality (
There are several limitations to this study. First, sports experience was measured using data reported retrospectively by parents, making it subject to recall bias. Second, because of the cross-sectional nature of the data, the possibility of reverse causality (e.g., children with more psychopathology may be less interested in sports) cannot be ruled out, especially regarding psychopathology and MRI outcomes. While prospective data on TS participation and psychopathology are currently available (through year 2) in the ABCD cohort, year 2 MRI scans have not yet undergone the rigorous QC that was applied to baseline scans; furthermore, TS were substantially disrupted by the COVID-19 pandemic during this window, complicating longitudinal analyses. However, the fact that baseline findings persisted through IPW and other sensitivity analyses may somewhat alleviate this concern. Third, despite controlling for INR and puberty stage, unmeasured confounders likely influence the relationship between TS and psychopathology. Additionally, while the ABCD Study population mirrors the racial and ethnic composition of the U.S. population, there are selection biases with regard to socioeconomic status and geographical distribution, which could affect the access to TS activities (
). Finally, as with numerous population genetics studies, results from participants of European ancestry may not generalize to other populations, and more inclusive genome-wide association studies involving non-European individuals are needed.
Despite these limitations, the results of this study indicate the promise of TS participation as a modifiable, protective factor for psychopathology risk and point to potential intermediate biological mechanisms. It will be of interest to revisit effects of TS exposure longitudinally in ABCD Study participants as they approach late adolescence, a time of heightened risk for emergence of mental illness, in concert with other neurodevelopmental changes. This additional work may provide additional support for TS-based interventions for youth who are at increased biological risk for psychopathology.
Acknowledgments and Disclosures
This study was supported by the National Institutes of Health (Grant No. R01MH120402 [to JLR and AED], Grant No. R01MH124694 [to JLR], and Grant No. R01MH113930 [to ECD]) and the Massachusetts General Hospital Early Brain Development Initiative. This research was made possible, in part, by the computational hardware generously provided by the Massachusetts Life Sciences Center (https://www.masslifesciences.com/).
The authors report no biomedical financial interests or potential conflicts of interest.
Role of physical activity and sedentary behavior in the mental health of preschoolers, children and adolescents: A systematic review and meta-analysis.
Chronic physical activity for attention deficit hyperactivity disorder and/or autism spectrum disorder in children: A meta-analysis of randomized controlled trials.
A whole brain volumetric approach in overweight/obese children: Examining the association with different physical fitness components and academic performance. The ActiveBrains project.
Similar but distinct – Effects of different socioeconomic indicators on resting state functional connectivity: Findings from the Adolescent Brain Cognitive Development (ABCD) Study®.
00009-5&id=gr1.jpg){kind=link}
00009-5&id=gr2.jpg){kind=link}
00009-5&id=gr3.jpg){kind=link}
00009-5&id=gr4.jpg){kind=link}