Biological

Psychiatry Archival Report

Multivariate Patterns of Brain-Behavior-

Environment Associations in the Adolescent
Brain and Cognitive Development Study
Amirhossein Modabbernia, Delfina Janiri, Gaelle E. Doucet, Abraham Reichenberg, and
Sophia Frangou

ABSTRACT
BACKGROUND: Adolescence is a critical developmental stage. A key challenge is to characterize how variation in
adolescent brain organization relates to psychosocial and environmental influences.
METHODS: We used canonical correlation analysis to discover distinct patterns of covariation between measures of
brain organization (brain morphometry, intracortical myelination, white matter integrity, and resting-state functional
connectivity) and individual, psychosocial, and environmental factors in a nationally representative U.S. sample of
9623 individuals (aged 9–10 years, 49% female) participating in the Adolescent Brain and Cognitive Development
(ABCD) study.
RESULTS: These analyses identified 14 reliable modes of brain-behavior-environment covariation (canonical
rdiscovery = .21 to .49, canonical rtest = .10 to .39, pfalse discovery rate corrected , .0001). Across modes, neighborhood
environment, parental characteristics, quality of family life, perinatal history, cardiometabolic health, cognition, and
psychopathology had the most consistent and replicable associations with multiple measures of brain
organization; positive and negative exposures converged to form patterns of psychosocial advantage or adversity.
These showed modality-general, respectively positive or negative, associations with brain structure and function
with little evidence of regional specificity. Nested within these cross-modal patterns were more specific
associations between prefrontal measures of morphometry, intracortical myelination, and functional connectivity
with affective psychopathology, cognition, and family environment.
CONCLUSIONS: We identified clusters of exposures that showed consistent modality-general associations with
global measures of brain organization. These findings underscore the importance of understanding the complex
and intertwined influences on brain organization and mental function during development and have the potential to
inform public health policies aiming toward interventions to improve mental well-being.
https://doi.org/10.1016/j.biopsych.2020.08.014

Adolescence is a period of significant brain reorganization that
can be assessed using a range of neuroimaging measures
(1–8). Morphometric changes involve increase in brain volume,
expansion of the cortical surface area, and reduction in cortical
thickness (1,3). Maturational changes in intracortical myelina-
tion can be captured using the gray matter (GM)/white matter
(WM) contrast, while metrics derived from diffusion magnetic
resonance imaging (MRI) inform about similar changes in WM
microstructure (8). Intracortical myelination shows a prolonged
developmental trajectory, primarily within prefrontal and other
association regions, that facilitates the dynamic experience-
dependent configuration of cognitive systems (2,6,7). Devel-
opmental increases in the diameter and myelination of axons
and in the fiber packing density of WM tracts are reflected in
reductions in axial and radial water diffusivity and increases in
fractional anisotropy (FA) (9). Functional brain changes pri-
marily involve increased connectivity of the cognitive brain
networks linked to improvements in the adaptive control of
cognition and behavior (5).
Brain organization throughout the lifespan has been asso-
ciated with multiple genetic, molecular, behavioral, and envi-
ronmental factors (10–18). Large-scale studies have begun to
address the complexity of the brain-behavior-environment
associations. In the Human Connectome Project (www.
humanconnectomeproject.org), a U.S. study of more than
1000 healthy adults aged 22–37 years, positive personal at-
tributes and environmental factors were positively correlated
with each other and with brain structure and connectivity; the
reverse was observed for negative factors (19–21). A study on
the first 5000 participants of the UK Biobank (www.ukbiobank.
ac.uk), a population-based study involving adults aged 40–69
years, reported multiple distinct modes of population covari-
ation that reflected associations between multimodal mea-
sures of brain organization and life factors including

510 ª 2020 Published by Elsevier Inc on behalf of Society of Biological Psychiatry.

Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal ISSN: 0006-3223

Brain-Behavior-Environment Associations in Adolescents
physical characteristics and socioeconomic and cognitive
variables (22).
Understanding the role of positive and negative life factors
in shaping brain organization in early adolescence is particu-
larly relevant to mental health, as the first manifestation of most
mental disorders can be traced to this period (23).
Neuroscience-based public health policies to improve mental
health outcomes critically depend on evidence from large-
scale datasets. The Adolescent Brain and Cognitive Develop-
ment (ABCD) study (https://abcdstudy.org) (24–27) is such a
large dataset that provides high-quality neuroimaging data and
detailed environmental and behavioral information from a na-
tionally representative population-based sample of 9–10-year-
olds (N = 11,875) living in the United States. A recent study by
Alnaes et al. (18) found 3 major global modes of brain-behavior
associations in 7577 individuals from the ABCD sample by
pooling all neuroimaging data together. However, each imag-
ing modality provides information about different aspects of
brain organization (1–9), and pooled analyses risk losing this
level of granularity. In the present study, we tested 1) whether
patterns of covariation between a wide array of brain metrics
with a broad range of personal characteristics and environ-
mental exposures are robust and reproducible, 2) whether
different exposures show modality-specific or modality-
general associations with adolescent brain organization, and
3) whether covariation between exposures and brain metrics is
spatially diffuse or shows evidence of regional specificity. To
achieve this, we used multivariate analyses to identify patterns
of covariation between multimodal neuroimaging metrics and a
comprehensive array of nonimaging measures pertaining to
perinatal and developmental history, physical and mental well-
being, cognitive ability, parental characteristics, family func-
tioning, neighborhood environment, and socioeconomic
circumstances.
METHODS AND MATERIALS
Sample
The ABCD study recruited a nationally representative cohort of
11,875 participants aged 9–10 years at 22 U.S. sites using
multistage probability sampling (Supplemental Methods,
Section 1). The analyses presented here used data pre-
processed by the ABCD study and downloaded on July 2019
as part of the ABCD Study Curated Annual Release 2.0.1
(https://data-archive.nimh.nih.gov/abcd). We selected ABCD
participants based on the availability of nonimaging measures
(NIMs) and of high-quality neuroimaging data (Supplemental
Methods, Section 4). The World Health Organization defines
adolescence as the period beginning at the 10th birthday;
because most ABCD participants were close to or above this
age, we refer to all participants as adolescents.
Nonimaging Measures
In addition to age, sex, and race, we examined a total of 72
NIMs pertaining to perinatal events (e.g., parental age at child’s
birth, planned pregnancy, and maternal medical conditions
during pregnancy; n = 11), early development (i.e., duration of
breastfeeding, motor development, and verbal development;
n = 3), anthropometric measures (i.e., pubertal stage, height,
Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal
Biological
Psychiatry
weight, waist circumference; n = 4), psychopathology (i.e.,
8 subscales of the Child Behavior Checklist, prodromal
symptoms, subsyndromal mania, and sleep problems; n = 11),
cognition (i.e., fluid and crystallized cognition; n = 2), psy-
chological traits (e.g., negative urgency, sensation seeking,
and behavioral inhibition; n = 10), childhood medical and
neurological conditions (n = 4), school engagement and envi-
ronment (n = 3), life events and lifestyle (e.g., trauma, screen
time, and physical activity; n = 6), parental characteristics (e.g.,
education, psychopathology, and marital status; n = 5), family
environment (e.g., finances, family conflict, and quality of
parental engagement; n = 7), and neighborhood environment
(e.g., area deprivation, risk of lead exposure, and crime; n = 6)
(25–27). Detailed definition of the measures and the corre-
sponding assessment instruments are provided in
Supplemental Methods and Table S1.
Neuroimaging Measures
The availability of high-quality neuroimaging data differed for
each modality. To maximize sample size and identify modality-
specific signatures of association, we analyzed each modality
separately using morphometric data from 9623 participants,
GM/WM contrast measures from 9215 participants, diffusion
MRI (dMRI) measures from 7637 participants, and resting-state
network (RSN) connectivity measures from 5742 participants
(Table 1). Structural MRI (T1 and T2 weighted), dMRI, and
resting-state functional MRI data were collected, processed,
and analyzed according to previously published and stan-
dardized ABCD protocols (24,26,28) (details in Supplemental
Methods, Section 4). Definitions of all the neuroimaging mea-
sures are provided in Tables S2–S5.
Statistical Analyses
A detailed step-by-step description of the statistical analyses
is presented in the Supplemental Methods, Section 5. Using
an identical pipeline (Figure S1), we conducted multivariate
analyses between the NIMs and neuroimaging measures from
each modality separately (i.e., brain morphometry derived from
T1-weighted images, myelination as inferred from GM/WM
contrast in T1-weighted images, WM integrity derived from
dMRI and RSN connectivity) (Supplemental Methods, Section
5.1). The code for the analysis is provided in the
Supplement. Briefly, each pipeline consisted of the following
steps: 1) division of the original datasets into a training sub-
sample (85% of the original) and a test subsample (15% of the
original); 2) missing value imputation using bagImpute in the
Caret package (version 6.0–84); 3) residualization for potential
confounders, which were age, age2, sex, age 3 sex, age2 3
sex, ethnicity, handedness, and scanner site for both imaging
measures and NIMs, and head motion for the imaging data; 4)
principal component analysis (PCA) of the residualized values
implemented using the pca function from the mixOmics
package (version 6.8.0); 5) canonical correlation analysis (CCA)
on the resulting principal components implemented using the
CCA package (version 1.2) in R; 6) calculation of canonical
structural loadings, i.e., the correlation coefficient between
each variate and its corresponding variables, to quantify the
contribution of each variable to each mode; 7) calculation of p
values for the CCA models based on 10,000 randomly
511
Biological
Psychiatry
Table 1. Demographic Characteristics of Each Sample
sMRI (n = 9623) Gray/White Matter Contrast (n = 9215)
Age, Months, Mean (SD) 119 (7.5)
Sex, Female, n (%) 4766 (49%)
Race, White, n (%) 5043 (52%)
Race, Hispanic, n (%) 1402 (15%)
Race, Black, n (%) 1991 (21%)
Race, Other, n (%) 1187 (12%)
The criteria for the selection of the study samples are described in the Supplemental Methods.
DWI, diffusion-weighted imaging; MRI, magnetic resonance imaging; rs-fMRI, resting-state functional MRI; sMRI, structural MRI.
permuted datasets (Supplemental Methods, Section 5.5); and
8) testing of the generalizability of the CCA modes generated
from the training subsample to the test subsample. The
threshold for statistical significance, following false discovery
rate (FDR) (29) correction for multiple testing, was set at pFDR ,
.05. Significant modes that explained at least 1% of the
covariance were reported.
To ensure reliability and reproducibility, in addition to the
permutation testing mentioned above, we undertook further
analyses described in detail in the Supplemental Methods,
Section 5.6. These included testing the stability of the PCA-
CCA results 1) to sample composition using bootstrapping
with replacement and random resampling, 2) to site using the
leave-one-out method, 3) to imputation by repeating the ana-
lyses using only nonimputed data, 4) to family relatedness, and
5) to sex.
RESULTS
Brain Morphometry
We considered data from 9623 ABCD participants with high-
quality global and regional measures of cortical thickness
and surface area and subcortical volumes (Figure S2;
Tables S2 and S7). Five modes, labeled morphometry modes
(MMs) 1 to 5, were statistically significant at pFDR , .0001 and
replicable in the discovery (disc) (n = 8144; rdisc = .26 to .46)
and test (n = 1479; rtest = .11 to .39) subsamples (Figure 1;
Figures S3–S5). In MM1, which explained the largest amount
of covariance, the covariation pattern of the NIMs appeared
aligned with traditional views of positive and negative life
factors (Figure 1B). Specifically, the highest positive loadings
comprised measures reflecting better cognition, better phys-
ical development (e.g., higher birth weight, current height, and
weight), and enriched social environment (e.g., higher family
income and parental education, planned pregnancy, greater
parental warmth and engagement). Negative life factors with
the highest loading comprised indices of personal (e.g., birth
complications, higher psychopathology) and social (e.g.,
neighborhood deprivation and family conflict) disadvantage.
The direction of the morphometric loadings in MM1 were
consistent with the expected developmental changes (1,3):
volumetric and cortical surface area measures showed positive
loadings, and cortical thickness measures showed negative
loadings (Figure 1C). The largest negative loadings were
observed for the cortical thickness of the cingulate and of the
medial, lateral, and ventral prefrontal regions; the largest
positive loadings were noted for total brain volume and global
512 Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal
Brain-Behavior-Environment Associations in Adolescents
DWI (n = 7637) rs-fMRI (n = 5742)
119 (7.5) 119 (7.5) 120 (7.5)
4542 (49%) 3836 (49%) 2981 (52%)
4872 (53%) 4258 (55%) 3173 (55%)
1339 (15%) 1013 (13%) 723 (13%)
1883 (20%) 1587 (20%) 1145 (20%)
1121 (12%) 911 (12%) 701 (12%)
surface area. Of the remaining modes, MM5 mainly reflected
associations with psychopathology and behavioral traits. The
remaining modes (MM2, MM3, and MM4), explained residual
variance relating to 1) positive covariation between cortical
thickness and measures of cardiometabolic fitness (e.g., lower
weight and waist circumference) in addition to better socio-
economic status, better cognition, and lower psychopathology
in MM2 (Figure 1; Figures S3–S5), 2) positive covariation of
twin birth and high socioeconomic status and global
morphometric measures in MM3 (Figure 1; Figures S3-S5), and
3) negative covariation between subsyndromal mood symp-
toms, higher cognitive and school engagement, and higher
socioeconomic status with cortical thickness in MM4 (Figure 1;
Figures S3–S5).
GM/WM Contrast
We considered data from 9215 ABCD participants (Figure S2;
Tables S3 and S8) with high-quality GM/WM contrast mea-
sures. Three modes, labeled contrast modes (CMs) 1 to 3,
were statistically significant at pFDR , .0001 in the discovery
subsample (n = 7795; rdisc = .21 to .37) and replicable in the
test subsample (n = 1420; rtest = .10 to .20) (Figure 2;
Figures S6 and S7). CM1 was dominated by negative load-
ings involving weight, height, and waist circumference and
positive loadings on GM/WM contrast measures of multiple
regions throughout the brain (Figure 2; Figure S7). In CM2,
higher positive loadings in the GM/WM contrast in orbito-
frontal regions were most evident and associated with posi-
tive loadings for birth weight and negative loadings for
obstetrical events (prematurity and twin birth) (Figure 2;
Figure S7). CM3 captured the association between height,
better cognition, lower psychopathology, and higher physical
activity: GM/WM contrast measures showed positive load-
ings in orbitofrontal and entorhinal regions and negative
loadings in the visual and transverse temporal regions
(Figure 2; Figure S7).
WM Integrity
We considered data from 7637 ABCD participants (Figure S2;
Tables S4 and S9) with high-quality dMRI measures. Four
modes, labeled WM modes (WMMs) 1 to 4, were statistically
significant at pFDR , .0001 and replicable in the discovery (n =
6480; rdisc = .25 to .44) and test (n = 1157; rtest = .13 to .34)
(Figure 3; Figures S8-S9) subsamples. In WMM1, the highest
NIM loadings were negative and involved anthropometric
measures (most prominently weight, waist circumference,
height, and pubertal stage); the highest positive loadings
 Biological
Brain-Behavior-Environment Associations in Adolescents Psychiatry

Figure 1. Significant morphometry modes. (A) Variable loadings for morphometry measures for the 5 significant morphometry modes. (B) Variable loadings
for nonimaging measures for the 5 significant morphometry modes. (C) Loadings of the morphometry measures on regional cortical thickness and cortical
surface area for the 5 significant modes. For simplicity, only the right hemisphere is shown. For a more detailed presentation see Figures S4 and S5. A, cortical
surface area; CBCL, Childhood Behavior Checklist; T, cortical thickness; V, subcortical volumes.

involved the axial water diffusivity of the corpus callosum and
the major WM tract fibers (Figure 3; Figures S8-S9). In WMM2,
NIMs representing positive perinatal factors (higher birth
weight, longer duration of breastfeeding), better cognition, and
enriched social environment (family income and parental ed-
ucation and involvement) covaried with one another; similar
covariation was observed among negative early-life events
(maternal medical problems and substance use, birth compli-
cations, and delayed development), psychopathology (partic-
ularly social, prodromal, attention, and externalizing
symptoms), and social disadvantage (greater area deprivation
and severe financial difficulties). In WMM2, the highest positive
and negative loadings respectively involved the FA and radial
water diffusivity of large WM tracts, particularly those con-
necting prefrontal to other brain regions (Figure 3; Figures S8
and S9). WMM3 accounted for further variance attributable
to perinatal events (with positive loadings for twin birth and
prematurity and negative loading for birth weight) and their
association with positive axial water diffusivity and FA, but
negative radial water diffusivity loadings for corticostriatal and
corticospinal tracts (Figure 3; Figures S8 and S9). In WMM4,
NIMs showed a positive-negative covariation pattern similar to
WMM2, but with opposite signs on height and FA in large WM
tracts (particularly the uncinate and cingulate bundles)
(Figure 3; Figures S8 and S9).
Resting-State Network Connectivity
We considered data from 5742 ABCD participants (Figure S2;
Tables S5 and S10) with high-quality RSN connectivity
measures. Two modes (Figure 4; Figures S10 and S11),
labeled functional modes (FMs) 1 and 2, were statistically
significant at pFDR , .0001 in the discovery subsample (n =
4854; rdisc = .28 and .29) and replicable in the test subsample
(n = 888; rtest = .13 for both). In FM1, NIMs recapitulated the
pattern of covariation observed in the other modalities: it
showed high positive loadings for better cognition, positive
early-life experiences, and family environment and negative
loadings for area deprivation and psychopathology (particu-
larly externalizing psychopathology and psychotic-like

Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal 513

Biological
Psychiatry Brain-Behavior-Environment Associations in Adolescents

Figure 2. Significant gray/white matter contrast modes. (A) Variable loadings for gray/white matter contrast measures for the 3 significant gray/white matter
contrast modes. (B) Variable loadings for nonimaging measures for the 3 significant gray/white matter contrast modes. (C) Loadings of the gray/white matter
contrast measures depicted on the brain for the 3 significant modes. For a more detailed presentation see Figures S6 and S7. CBCL, Childhood Behavior
Checklist.

symptoms). In FM1, the largest positive loadings were
observed for the connectivity between the cingulo-opercular
and cingulo-parietal network, between the auditory and
salience networks, and between the retrosplenial-temporal
and visual networks. In the same mode, the largest negative
loadings were noted for the internetwork connectivity of the
sensorimotor networks, between the default mode and
retrosplenial-temporal networks, and between the dorsal and
ventral attention networks (Figure 4; Figures S10 and S11).
FM2 highlighted the positive covariation of socioeconomic
status and birth events, and negative covariation of screen
use and anthropometric variables (height, weight, and waist
circumferences), with positive loadings on intra- and inter-
network connectivity of the default mode and salience net-
works and negative loadings for the internetwork connectivity
of the default mode network with the cingulo-opercular and

514 Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal

 Biological
Brain-Behavior-Environment Associations in Adolescents Psychiatry

Figure 3. Significant white matter integrity modes. (A) Variable loadings for white matter integrity measures for the 4 significant white matter integrity modes.
(B) Variable loadings for nonimaging measures for the 4 significant white matter integrity modes. For a more detailed presentation see Figures S8 and S9.
CBCL, Childhood Behavior Checklist.

dorsal attention networks, and of the ventral attention
network with the visual network (Figure 4; Figures S10 and
S11).
Reliability and Reproducibility
The results reported above were robust to sample composi-
tion, imputation, site, relatedness, and sex, as shown in
Figures S12–S18 and Tables S11–S13.
DISCUSSION
Multivariate data-driven analyses identified 14 patterns of
brain-behavior-environment covariation in a large U.S.
population–based cohort of young adolescents. The findings
were robust to sample size and composition and
independent of site. In line with previous studies
(18,19,21,22,30), life factors typically considered positive
were associated with brain measures that reflect advanta-
geous brain development, while the reverse was the case for
factors considered negative. We demonstrated that positive
and negative exposures do not occur in isolation but cluster
together. Furthermore, we characterized a set of exposures
that are associated with brain development in a global and
modality-general manner that could serve as targets for
universal primary prevention.
A particular strength of the study is that the rich ABCD
dataset enabled us to expand prior literature by quantifying
the associations between multiple brain measures and a wide
range of NIMs in an integrated analysis framework. Positive
exposures tended to covary with one another, as did negative

Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal 515

Biological
Psychiatry Brain-Behavior-Environment Associations in Adolescents

Figure 4. Significant resting-state functional connectivity modes. (A) Variable loadings for resting-state functional connectivity measures for the 2 significant
resting-state functional connectivity modes. (B) Variable loadings for nonimaging measures for the 2 significant resting-state functional connectivity modes. (C)
Top 10 between-network connectivity loadings for the 2 significant resting-state functional connectivity modes. For a more detailed presentation see
Figures S10 and S11. Aud, auditory network; CBCL, Childhood Behavior Checklist; CO, cingulo-opercular network; CP, cingulo-parietal network; DA, dorsal
attention network; DM, default mode network; FP, frontoparietal; RST, retrosplenial-temporal network; Sal, salience network; SMH, sensorimotor hand; SMM,
sensorimotor mouth; VA, ventral attention network; Vis, visual network.

516 Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal


Brain-Behavior-Environment Associations in Adolescents
exposures (Figures 1B, 2B, 3B, and 4B). Although their rela-
tive contribution varied across modes, the direction of asso-
ciation was usually consistent with their hypothesized effects:
positive exposures were positively associated with measures
reflecting favorable brain development, and the reverse was
the case for negative exposures (19,21,22). Of note, both
positive and negative factors showed spatially diffuse pat-
terns of covariation with brain measures. We were able to
demonstrate this by retaining global measures of brain
thickness and volume in our models as opposed to removing
their variance by residualizing regional measures. Together,
these observations suggest that attributions of specificity in
correlations between single NIMs and brain measures may
need to be revisited, and future studies would benefit from
focusing on the wider context in which brain development
occurs.
Factors most consistently associated with neuroimaging
measures across modalities involved socioeconomic circum-
stances, the quality of parental engagement, perinatal events,
cognitive ability, and psychopathology. In modes where psy-
chopathology loadings were present, they typically covaried
with lower cognition and lower quality of family and neigh-
borhood environments. We suggest that these commonly co-
occurring exposures, and their effect on the brain, should be
considered in models of brain psychopathology. Within
modes, the different measures of psychopathology also co-
varied with one another, in line with arguments for a common
latent dimension for psychopathology (31). The association
between general psychopathology and lower cognitive abilities
has been documented in other population-based cohorts,
notably the Dunedin Multidisciplinary Health and Development
Study (23).
However, two modes went against this general pattern. In
MM4, we observed covariation of affective and sleep prob-
lems, better academic ability, higher socioeconomic status,
and greater cortical thinning. These findings resonate with
epidemiological evidence linking higher cognitive and aca-
demic performance in adolescence with the emergence of
bipolar morbidity in adulthood (32,33). MM5 captured the
remaining covariance relating to dysregulation in attention
and impulse control and social adversity. This pattern
also has been identified in epidemiological studies in
connection to childhood disorders such as attention-deficit/
hyperactivity disorder and disruptive mood dysregulation
disorder (34), and mood and substance use disorders in
adulthood (35).
We confirmed the known links between higher cognitive
ability and brain development across modalities (36,37).
Although we highlight here the associations of cognitive ability
with RSN connectivity and GM/WM contrast because of their
high loadings in FM1 and CM3, we emphasize that these
occurred within the context of advantageous socioeconomic
and family conditions. In FM1, better cognitive ability was
associated with increased connectivity between networks that
support higher-order cognitive functions associated with
salience processing (cingulo-opercular and salience networks)
and executive control (cingulo-parietal network). The higher
integration of these networks is consistent with their evolving
role in coordinating multiple mental operations in the maturing
brain (5). In CM3, higher GM/WM contrast intensity in
Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal
Biological
Psychiatry
prefrontal regions was also associated with better cognition.
Higher GM/WM contrast intensity is thought to represent lower
levels of intracortical myelin (38,39). The lower myelin con-
centration in the prefrontal cortices, together with their pro-
longed maturation pattern, is thought to facilitate flexible,
experience-dependent configuration of prefrontally linked
cognitive systems (7,40).
Anthropometric measures made major contributions to
multiple modes across modalities, confirming body-brain
associations during development (30,41). We did not
remove the variance of these variables from the sample,
because it was important to demonstrate the strength of
their associations with brain organization, which are greater
than those of many other factors. This perspective is
important when making inferences about the strength of the
associations of different factors with brain measures that
often tend to overemphasize factors of much smaller
magnitude. We highlight the loadings of weight and waist
circumference on morphometry (MM2, MM3), axial diffu-
sivity (WMM1), and intracortical myelin (CM1). These mea-
sures, which are considered indices of “cardiometabolic
fitness” (42), appear to influence brain organization even at
the early stages of adolescence. We also note the persis-
tent association between brain morphometry (MM1 and
MM2) and WM integrity (WMM2) with perinatal events,
particularly prematurity, birth weight, and twin births. These
findings add to the substantial body of evidence linking
perinatal events to health outcomes throughout the lifespan
(43,44).
It is important to consider key study limitations that point
to avenues for future research. Covariation is not evidence of
causation, although we assume that many of the reported
associations are underpinned by biological, and likely caus-
ative, mechanisms. The modes identified here are based on
cross-sectional data; however, as the ABCD study has a
longitudinal design, it would be possible to track the evolu-
tion of brain-behavior-environment covariation when follow-
up data become available. Conversely, brain development
is influenced by multiple factors in operation before the
enrollment age of the ABCD participants. It is conceivable
that some of the results reported here are conditional on or
influenced by very early exposures that are not fully captured
here. We used a dimensional approach to model psycho-
pathology as this approach is considered psychometrically
robust, accommodates comorbidity, and circumvents un-
certainty about the biological validity of diagnostic cate-
gories (45). Comparisons between categorical and
dimensional models of psychopathology with regard to their
brain-environment associations are worth addressing in
future studies. We did not model specific domains of
cognition that may characterize individual cognitive profiles
more precisely (46). Nevertheless, global measures of
cognition show significant covariation with domain-specific
measures, which supports the usefulness of global mea-
sures in informing public health strategies. The macroscale
measures examined here are widely used in developmental
neuroscience but do not capture brain-behavior-environment
associations that may operate at a microstructural level. Prior
research on brain-behavior-environment associations has
been criticized for overestimating their effect size and for
517
Biological
Psychiatry
paying insufficient attention to their reproducibility (47,48). In
response, we undertook extensive reliability testing and we
report only canonical modes that were reliable and general-
izable. We ensured the reproducibility of the analyses using
multiple techniques, including permutation, resampling sta-
bility, and replicability testing by splitting the sample into
independent subsets. It is conceivable that our analyses may
have missed certain brain-behavior-environment associa-
tions that may only exist within distinct but small and yet
unknown subpopulations.
In summary, the richness of the imaging and nonimaging
data available in the ABCD study and the large size of the
cohort enables us to examine brain-behavior-environment
associations in young adolescents in greater depth than
has been previously possible. We identified 14 modes that
represent patterns of covariation between different aspects
of brain organization and behavioral and environmental fac-
tors. We show that many of these factors tend to occur along
patterns of increased advantage or adversity. Modeling life
factors in isolation is therefore likely to miss the importance
of the wider context in which they occur. Adolescence rep-
resents a window of opportunity when interventions targeting
vulnerable individuals may succeed in shifting developmental
trajectories to positive mental health outcomes. In this
context, the present study identified a set of exposures
relating to neighborhood environment, parental characteris-
tics, quality of family life, perinatal history, cardiometabolic
health, cognition, and psychopathology that had the most
consistent association with brain organization and could be
considered as priority targets for universal interventions to
improve mental well-being in youths. Importantly, many of
these exposures are modifiable and amenable to in-
terventions that fall within the purview of health and educa-
tion services. We note that the present results provide further
support recent initiatives in school-based interventions
(49,50) and move the field forward toward neuroscience-
informed public policy for the promotion of mental well-
being.
ACKNOWLEDGMENTS AND DISCLOSURES
This work was supported by the National Institute of Mental Health (Grant
Nos. R01MH113619 and R01 MH116147 [to SF]), National Institute on Aging
(Grant No. R03-AG064001 [to GED]), National Institute of General Medical
Sciences (Grant No. P20GM130447 [to GED]), National Institute of Child and
Human Development (Grant No. R01-HD098883 [to AR]), and National
Institute of Environmental Health Sciences (Grant No. R01-ES026904 [to
AR]).
This work was supported in part through the computational resources
and staff expertise provided by Scientific Computing at the Icahn School of
Medicine at Mount Sinai.
Data used in the preparation of this article were obtained from the
Adolescent Brain Cognitive Development (ABCD) study (https://abcdstudy.
org), held in the National Institute of Mental Health Data Archive. ABCD is
a multisite longitudinal study designed to recruit more than 10,000 children
aged 9–10 years and follow them over 10 years into early adulthood. The
ABCD study is supported by the National Institutes of Health and additional
federal partners under Award Nos. U01DA041022, U01DA041028,
U01DA041048, U01DA041089, U01DA041106, U01DA041117,
U01DA041120, U01DA041134, U01DA041148, U01DA041156,
U01DA041174, U24DA041123, U24DA041147, U01DA041093, and
U01DA041025. A full list of supporters is available at https://abcdstudy.org/
518 Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal
Brain-Behavior-Environment Associations in Adolescents
federal-partners.html. A listing of participating sites and a complete listing of
the study investigators can be found at https://abcdstudy.org/scientists/
workgroups/. ABCD consortium investigators designed and implemented
the study and/or provided data but did not necessarily participate in analysis
or writing of the present report, which reflects the views of the authors and
not necessarily the opinions or views of the National Institutes of Health or
ABCD consortium investigators. The ABCD data repository grows and
changes over time.
Details regarding data access can be found at https://nda.nih.gov/abcd/
request-access. The analysis code is available at https://github.com/
AmirhosseinModabbernia/ABCD.
The authors report no biomedical financial interests or potential conflicts
of interest.
ARTICLE INFORMATION
From the Department of Psychiatry (AM, AR, SF), Department of Environ-
mental Medicine and Public Health (AR), Seaver Center for Autism Research
and Treatment (AR), and Mindich Child Health and Development Institute
(AR), Icahn School of Medicine at Mount Sinai, New York, New York; Boys
Town National Research Hospital (GED), Omaha, Nebraska; Department of
Psychiatry and Neurology (DJ), Sapienza University of Rome, Rome, Italy;
and Djavad Mowafaghian Centre for Brain Health (SF), Department of Psy-
chiatry, University of British Columbia, Vancouver, British Columbia,
Canada.
Address correspondence to Sophia Frangou, M.D., Ph.D., at sophia.
frangou@mssm.edu.
Received May 10, 2020; revised Aug 17, 2020; accepted Aug 19, 2020.
Supplementary material cited in this article is available online at https://
doi.org/10.1016/j.biopsych.2020.08.014.
REFERENCES
1. Giedd JN, Blumenthal J, Jeffries NO, Castellanos FX, Liu H,
Zijdenbos A, et al. (1999): Brain development during childhood
and adolescence: A longitudinal MRI study. Nat Neurosci 2:861–
863.
2. Grydeland H, Walhovd KB, Tamnes CK, Westlye LT, Fjell AM (2013):
Intracortical myelin links with performance variability across the
human lifespan: Results from T1- and T2-weighted MRI
myelin mapping and diffusion tensor imaging. J Neurosci 33:18618–
18630.
3. Tamnes CK, Herting MM, Goddings AL, Meuwese R, Blakemore SJ,
Dahl RE, et al. (2017): Development of the cerebral cortex across
adolescence: A multisample study of inter-related longitudinal
changes in cortical volume, surface area, and thickness. J Neurosci
37:3402–3412.
4. Lewis JD, Evans AC, Tohka J (2018): T1 white/gray contrast as a
predictor of chronological age, and an index of cognitive performance.
NeuroImage 173:341–350.
5. Gu S, Satterthwaite TD, Medaglia JD, Yang M, Gur RE, Gur RC, et al.
(2015): Emergence of system roles in normative neurodevelopment.
Proc Natl Acad Sci U S A 112:13681–13686.
6. Norbom LB, Doan NT, Alnaes D, Kaufmann T, Moberget T, Rokicki J,
et al. (2019): Probing brain developmental patterns of myelination and
associations with psychopathology in youths using gray/white matter
contrast. Biol Psychiatry 85:389–398.
7. Huntenburg JM, Bazin PL, Goulas A, Tardif CL, Villringer A,
Margulies DS (2017): A systematic relationship between functional
connectivity and intracortical myelin in the human cerebral cortex.
Cereb Cortex 27:981–997.
8. Lewis JD, Fonov VS, Collins DL, Evans AC, Tohka J, Brain Develop-
ment Cooperative Group, Pediatric Imaging, Neurocognition, and
Genetics Study (2019): Cortical and subcortical T1 white/gray contrast,
chronological age, and cognitive performance. Neuroimage 196:276–
288.
9. Lebel C, Beaulieu C (2011): Longitudinal development of human brain
wiring continues from childhood into adulthood. J Neurosci 31:10937–
10947.

Brain-Behavior-Environment Associations in Adolescents
10. Tost H, Champagne FA, Meyer-Lindenberg A (2015): Environmental
influence in the brain, human welfare and mental health. Nat Neurosci
18:1421–1431.
11. Grasby KL, Jahanshad N, Painter JN, Colodro-Conde L, Bralten J,
Hibar DP, et al. (2020): The genetic architecture of the human cerebral
cortex. Science 367:eaay6690.
12. Hair NL, Hanson JL, Wolfe BL, Pollak SD (2015): Association of child
poverty, brain development, and academic achievement. JAMA
Pediatr 169:822–829.
13. Noble KG, Houston SM, Brito NH, Bartsch H, Kan E, Kuperman JM,
et al. (2015): Family income, parental education and brain structure in
children and adolescents. Nat Neurosci 18:773–778.
14. Natu VS, Gomez J, Barnett M, Jeska B, Kirilina E, Jaeger C, et al.
(2019): Apparent thinning of human visual cortex during childhood is
associated with myelination. Proc Natl Acad Sci U S A 116:20750–
20759.
15. Calderon-Garciduenas L, Engle R, Mora-Tiscareno A, Styner M,
Gomez-Garza G, Zhu H, et al. (2011): Exposure to severe urban air
pollution influences cognitive outcomes, brain volume and sys-
temic inflammation in clinically healthy children. Brain Cogn
77:345–355.
16. Velthorst E, Smith L, Bello G, Austin C, Gennings C, Modabbernia A,
et al. (2017): New research strategy for measuring pre- and postnatal
metal dysregulation in psychotic disorders. Schizophr Bull 43:1153–
1157.
17. Lees B, Aguinaldo L, Squeglia LM, Infante MA, Wade NE, Hernandez
Mejia M, et al. (2020): Parental family history of alcohol use disorder
and neural correlates of response inhibition in children from the
Adolescent Brain Cognitive Development (ABCD) study. Alcohol Clin
Exp Res 44:1234–1244.
18. Alnaes D, Kaufmann T, Marquand AF, Smith SM, Westlye LT (2020):
Patterns of sociocognitive stratification and perinatal risk in the child
brain. Proc Natl Acad Sci U S A 117:12419–12427.
19. Smith SM, Nichols TE, Vidaurre D, Winkler AM, Behrens TE,
Glasser MF, et al. (2015): A positive-negative mode of population
covariation links brain connectivity, demographics and behavior. Nat
Neurosci 18:1565–1567.
20. Moser DA, Doucet GE, Lee WH, Rasgon A, Krinsky H, Leibu E, et al.
(2018): Multivariate associations among behavioral, clinical, and
multimodal imaging phenotypes in patients with psychosis. JAMA
Psychiatry 75:386–395.
21. Moser DA, Doucet GE, Ing A, Dima D, Schumann G, Bilder RM, et al.
(2018): An integrated brain-behavior model for working memory. Mol
Psychiatry 23:1974–1980.
22. Miller KL, Alfaro-Almagro F, Bangerter NK, Thomas DL, Yacoub E,
Xu J, et al. (2016): Multimodal population brain imaging in the UK
Biobank prospective epidemiological study. Nat Neurosci 19:1523–
1536.
23. Caspi A, Houts RM, Belsky DW, Goldman-Mellor SJ, Harrington H,
Israel S, et al. (2014): The p factor: One general psychopathology
factor in the structure of psychiatric disorders? Clin Psychol Sci 2:119–
137.
24. Casey BJ, Cannonier T, Conley MI, Cohen AO, Barch DM,
Heitzeg MM, et al. (2018): The Adolescent Brain Cognitive Develop-
ment (ABCD) study: Imaging acquisition across 21 sites. Dev Cogn
Neurosci 32:43–54.
25. Barch DM, Albaugh MD, Avenevoli S, Chang L, Clark DB, Glantz MD,
et al. (2018): Demographic, physical and mental health assessments in
the Adolescent Brain and Cognitive Development study: Rationale and
description. Dev Cogn Neurosci 32:55–66.
26. Garavan H, Bartsch H, Conway K, Decastro A, Goldstein RZ,
Heeringa S, et al. (2018): Recruiting the ABCD sample: Design con-
siderations and procedures. Dev Cogn Neurosci 32:16–22.
27. Zucker RA, Gonzalez R, Feldstein Ewing SW, Paulus MP, Arroyo J,
Fuligni A, et al. (2018): Assessment of culture and environment in the
Adolescent Brain and Cognitive Development study: Rationale,
description of measures, and early data. Dev Cogn Neurosci 32:107–
120.
Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal
Biological
Psychiatry
28. Hagler DJ Jr, Hatton S, Cornejo MD, Makowski C, Fair DA, Dick AS,
et al. (2019): Image processing and analysis methods for the
Adolescent Brain Cognitive Development study. Neuroimage
202:116091.
29. Benjamini Y, Hochberg Y (1995): Controlling the false discovery rate: A
practical and powerful approach to multiple testing. J R Stat Soc
Series B Stat Methodol 57:289–300.
30. Modabbernia A, Reichenberg A, Ing A, Moser DA, Doucet GE,
Artiges E, et al. (2020): Linked patterns of biological and environmental
covariation with brain structure in adolescence: A population-based
longitudinal study [published online ahead of print May 22]. Mol
Psychiatry.
31. Caspi A, Moffitt TE (2018): All for one and one for all: Mental disorders
in one dimension. Am J Psychiatry 175:831–844.
32. MacCabe JH, Lambe MP, Cnattingius S, Sham PC, David AS,
Reichenberg A, et al. (2010): Excellent school performance at age 16
and risk of adult bipolar disorder: National cohort study. Br J Psychi-
atry 196:109–115.
33. Kendler KS, Ohlsson H, Mezuk B, Sundquist K, Sundquist J (2016):
A Swedish national prospective and co-relative study of school
achievement at age 16, and risk for schizophrenia, other nonaffective
psychosis, and bipolar illness. Schizophr Bull 42:77–86.
34. Bruno A, Celebre L, Torre G, Pandolfo G, Mento C, Cedro C, et al.
(2019): Focus on disruptive mood dysregulation disorder: A review of
the literature. Psychiatry Res 279:323–330.
35. Papachristou E, Oldehinkel AJ, Ormel J, Raven D, Hartman CA,
Frangou S, et al. (2017): The predictive value of childhood sub-
threshold manic symptoms for adolescent and adult psychiatric out-
comes. J Affect Disord 212:86–92.
36. Burgaleta M, Johnson W, Waber DP, Colom R, Karama S (2014):
Cognitive ability changes and dynamics of cortical thickness
development in healthy children and adolescents. Neuroimage
84:810–819.
37. Khundrakpam BS, Lewis JD, Reid A, Karama S, Zhao L, Chouinard-
Decorte F, et al. (2017): Imaging structural covariance in the devel-
opment of intelligence. Neuroimage 144:227–240.
38. Westlye LT, Walhovd KB, Dale AM, Bjornerud A, Due-Tonnessen P,
Engvig A, et al. (2010): Differentiating maturational and aging-related
changes of the cerebral cortex by use of thickness and signal in-
tensity. Neuroimage 52:172–185.
39. Eickhoff S, Walters NB, Schleicher A, Kril J, Egan GF, Zilles K, et al.
(2005): High-resolution MRI reflects myeloarchitecture and
cytoarchitecture of human cerebral cortex. Hum Brain Mapp 24:206–
215.
40. Whitaker KJ, Vertes PE, Romero-Garcia R, Vasa F, Moutoussis M,
Prabhu G, et al. (2016): Adolescence is associated with genomically
patterned consolidation of the hubs of the human brain connectome.
Proc Natl Acad Sci U S A 113:9105–9110.
41. Vuoksimaa E, Panizzon MS, Franz CE, Fennema-Notestine C,
Hagler DJ Jr, Lyons MJ, et al. (2018): Brain structure mediates the
association between height and cognitive ability. Brain Struct Funct
223:3487–3494.
42. Zaqout M, Michels N, Bammann K, Ahrens W, Sprengeler O, Molnar D,
et al. (2016): Influence of physical fitness on cardio-metabolic risk
factors in European children. The IDEFICS study. Int J Obes (Lond)
40:1119–1125.
43. Belbasis L, Savvidou MD, Kanu C, Evangelou E, Tzoulaki I (2016):
Birth weight in relation to health and disease in later life: An
umbrella review of systematic reviews and meta-analyses. BMC
Med 14:147.
44. Hollanders JJ, Schaefer N, van der Pal SM, Oosterlaan J, Rotteveel J,
Finken MJJ, et al. (2019): Long-term neurodevelopmental and func-
tional outcomes of infants born very preterm and/or with a very low
birth weight. Neonatology 115:310–319.
45. Coghill D, Sonuga-Barke EJ (2012): Annual research review: Cate-
gories versus dimensions in the classification and conceptualisation of
child and adolescent mental disorders—implications of recent empir-
ical study. J Child Psychol Psychiatry 53:469–489.
519
Biological
Psychiatry
46. Kan K-J, van der Maas HL, Levine SZ (2019): Extending psychometric
network analysis: Empirical evidence against g in favor of mutualism?
Intelligence 73:52–62.
47. Kharabian Masouleh S, Eickhoff SB, Hoffstaedter F, Genon S (2019):
Empirical examination of the replicability of associations between
brain structure and psychological variables. Elife 8:e43464.
48. Paulus MP, Thompson WK (2019): The challenges and opportunities of
small effects: The New normal in academic psychiatry. JAMA Psy-
chiatry 76:353–354.
520 Biological Psychiatry March 1, 2021; 89:510–520 www.sobp.org/journal
Brain-Behavior-Environment Associations in Adolescents
49. Hayes D, Moore A, Stapley E, Humphrey N, Mansfield R, Santos J,
et al. (2019): Promoting mental health and wellbeing in schools:
Examining mindfulness, relaxation and strategies for safety and well-
being in English primary and secondary schools: Study protocol for a
multi-school, cluster randomised controlled trial (INSPIRE). Trials
20:640.
50. O’Reilly M, Svirydzenka N, Adams S, Dogra N (2018): Review of mental
health promotion interventions in schools. Soc Psychiatry Psychiatr
Epidemiol 53:647–662.