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, Hospital de Clínicas de Porto Alegre, Universidade Federal do Rio Grande do Sul, Porto Alegre, Rio Grande do Sul, BrazilDepartment of Medicine, Jacobs School of Medicine and Biomedical Sciences, University at Buffalo, Buffalo, New York
Institute of Developmental Psychiatry for Children and Adolescents, Hospital de Clínicas de Porto Alegre, Universidade Federal do Rio Grande do Sul, Porto Alegre, Rio Grande do Sul, BrazilADHD Outpatient and Developmental Psychiatry Programs, Hospital de Clínicas de Porto Alegre, Porto Alegre, Rio Grande do Sul, Brazil
Institute of Psychiatry, Psychology and Neuroscience, Department of Psychological Medicine, King’s College London, London, United KingdomNational Institute for Health Research Maudsley Biomedical Research Centre, South London and Maudsley National Health Service Foundation Trust, King’s College London, London, United Kingdom
There have been significant challenges in understanding functional brain connectivity associated with adolescent depression, including the need for a more comprehensive approach to defining risk, the lack of representation of participants from low- and middle-income countries, and the need for network-based approaches to model connectivity. The current study aimed to address these challenges by examining resting-state functional connectivity of frontolimbic circuitry associated with the risk and presence of depression in adolescents in Brazil.
Adolescents in Brazil ages 14 to 16 years were classified into low-risk, high-risk, and depressed groups using a clinical assessment and composite risk score that integrates 11 sociodemographic risk variables. After excluding participants with excessive head movement, resting-state functional magnetic resonance imaging data of 126 adolescents were analyzed. We compared group differences in frontolimbic network connectivity using region of interest–to–region of interest, graph theory, and seed-based connectivity analyses. Associations between self-reported depressive symptoms and brain connectivity were also explored.
Adolescents with depression showed greater dorsal anterior cingulate cortex (ACC) connectivity with the orbitofrontal cortex compared with the 2 risk groups and greater dorsal ACC global efficiency than the low-risk group. Adolescents with depression also showed reduced local efficiency and a lower clustering coefficient of the subgenual ACC compared with the 2 risk groups. The high-risk group also showed a lower subgenual ACC clustering coefficient relative to the low-risk group.
These findings highlight altered connectivity and topology of the ACC within frontolimbic circuitry as potential neural correlates and risk factors of developing depression in adolescents in Brazil. This study broadens our understanding of the neural connectivity associated with adolescent depression in a global context.
). Characterizing neurobiological risk factors and correlates of depression in adolescents could lead to improvements in preventing and treating adolescent depression. Prior research has identified several patterns of resting-state functional connectivity (rsFC) associated with the risk or presence of depression (
). However, a number of barriers to a comprehensive understanding of rsFC in adolescent depression remain, which if addressed could clarify our understanding of the neural correlates of risk and presence of depression in adolescents.
First, previous studies examining rsFC in relation to adolescent depression have used either a case-control approach (depressed vs. control) or a risk approach (high risk vs. low risk). Research is needed that combines these approaches to compare low-risk (LR) adolescents, high-risk (HR) adolescents, and adolescents with major depressive disorder (MDD) to distinguish between potential neural risk factors and neural correlates of depression. Second, studies with a risk approach have generally relied on parental history of depression as the risk factor (
), and some of them may have a high risk for depression based on other factors (e.g., social isolation). Moreover, parental history of depression can be accurately reported only by interviewing parents. Examining risk factors that can be directly reported by the adolescent could have future novel applications for identifying and preventing depression risk in a range of settings. A more comprehensive and accessible risk assessment is needed. Third, the majority of rsFC studies on adolescent depression have used a seed-based approach, restricting our knowledge to the connectivity of particular brain regions. An alternative and more comprehensive approach is a network-based approach (
) have applied network-based approaches targeting the whole brain or the reward-related network. More studies with a network-based approach targeting the frontolimbic network, a network important for adolescent depression and affective development, are needed. Finally, only 38% of rsFC studies on adolescent depression were conducted in low- and middle-income countries (LMIC), and the majority of studies in LMIC (i.e., 13 out of 15) were from China (
), studies in LMIC are urgently needed to address adolescent depression research disparities across the globe.
To overcome these barriers, the current study compared rsFC of 39 LR adolescents, 45 HR adolescents, and 42 adolescents with MDD recruited in Brazil, which is classified as a middle-income country according to the World Bank (
). To address the limitations of prior research that used parental history of depression to determine depression risk, adolescents in this study were classified using a clinical assessment and an empirically validated multivariable prognostic model (
) that integrated 11 sociodemographic variables (e.g., childhood maltreatment, social isolation), which has been shown to predict depression risk across a variety of countries, including Brazil, Nepal, the United Kingdom, and Nigeria (
). Moreover, to better understand aberrant functional neural architecture in high-risk and depressed adolescents, in addition to seed-based connectivity analysis, this study employed a network-based approach targeting frontolimbic circuitry implicated in both adolescent affective brain development (
) and found their altered connectivity with other brain regions in high-risk and depressed adolescents (e.g., decreased amygdala–medial PFC [mPFC]/ACC connectivity, increased ACC–ventromedial PFC connectivity). Interestingly, a study (
) using predictive modeling found that rsFC within the nodes of frontolimbic circuitry, but not whole-brain connectivity, predicted both current and future depressive symptoms measured after 18 months, highlighting its high relevance to adolescent depression.
The current study aimed to examine the strength and topology of frontolimbic network connectivity associated with the risk and presence of depression in Brazilian adolescents. Given that the above-mentioned study (
) examining rsFC of frontolimbic nodes in adolescents observed that positive dorsal ACC (dACC) connectivity with other regions exhibited the greatest contribution to predicting depressive symptoms, we hypothesized that adolescents with high risk for depression and depressed adolescents in our sample would show stronger (i.e., greater rsFC), more efficient (i.e., greater global efficiency, lower average path length), and a greater number of connections (i.e., greater degree) of the same dACC node compared with the LR adolescents. To test this hypothesis, we used region of interest (ROI)–to–ROI analysis and graph theory analysis. Owing to the nascent literature that has used network-based approaches for frontolimbic rsFC, we also explored group differences in connectivity strength and topological properties of all the regions within the frontolimbic network. We additionally conducted amygdala seed-based connectivity analysis to see whether the most well-established connectivity pattern in the depression literature [i.e., reduced amygdala–mPFC/ACC connectivity identified using a seed-based approach (
)] would be replicated in our sample of Brazilian adolescents. We hypothesized that the HR and MDD groups would show reduced amygdala–mPFC/ACC connectivity compared with the LR group. Besides examining the association of connectivity with risk group status and clinical depression, we also explored the association of frontolimbic connectivity with subjectively experienced depressive symptoms using a self-reported, continuous measure, which allowed us to examine associations with depressive symptoms spanning the subclinical through clinical range.
Methods and Materials
Participants were recruited for the IDEA-RiSCo (Identifying Depression Early in Adolescence Risk Stratified Cohort). Full details regarding the procedures of recruitment, screening, exclusion, clinical assessment, and questionnaires assessing sociodemographic variables are provided in the published protocol for the study (
For sampling adolescents who met the criteria of LR, HR, and MDD groups, 7720 adolescents 14 to 16 years of age were screened from June 2018 to November 2019 in Porto Alegre, Brazil. To stratify risk groups, we used a multivariable prognostic model, the IDEA Risk Score (IDEA-RS) (
). This model integrates 11 sociodemographic variables (i.e., skin color, biological sex, school failure, drug use, fight involvement, ran away from home, social isolation, childhood maltreatment, poor relationship with mother, father, and between parents) and generates the probability of presenting with a unipolar depressive episode in 3 years.
The LR group met the criteria of having a risk score equal to or below the 20th percentile of the Pelotas 1993 cohort, and the HR and MDD groups met the criteria of having a risk score equal to or above the 90th percentile of the Pelotas 1993 cohort. We required the MDD group to have high-risk sociodemographic profiles (an IDEA-RS equal to or above the 90th percentile) to attribute any neural differences between the HR and MDD groups only to the presence of depression.
Presence of a current MDD episode (in the MDD group) or absence of a current or past MDD episode (in the LR and HR groups) was determined with the Brazilian Portuguese version of the Schedule for Affective Disorders and Schizophrenia for School-Age Children–Present and Lifetime Version (
To make our sample more homogeneous, we included only participants without long-term or current use of psychotropic medications. Percentages of participants with lifetime comorbid diagnoses are reported in the Supplement.
After screening and clinical assessment, 150 participants who met inclusion criteria and did not meet exclusion criteria [see inclusion and exclusion criteria in the published protocol (
)] underwent MRI scanning from August 2018 to December 2019. Written informed assent and consent were obtained from adolescents and their caregivers, respectively, after the procedures had been fully explained. After exclusion of 24 participants with excessive head movement (i.e., greater than 20% volumes were censored), the sample size was 126 (LR group: n = 39; HR group: n = 45; MDD group: n = 42). This study was approved by the Brazilian National Ethics in Research Commission.
Data Acquisition, Preprocessing, and Denoising
All images were acquired on a 3T Ingenia (Philips Healthcare) MRI scanner. Structural MRI images were acquired before acquiring blood oxygen level–dependent functional MRI images for resting-state connectivity. Full details regarding the data acquisition parameters, preprocessing, and denoising are reported in the Supplement. The data were analyzed using CONN toolbox 18b. We confirmed that data from all participants had good signal coverage (i.e., signal coverage over 98% of voxels within each of our ROIs) (see Supplement for a specific method to inspect signal loss).
We estimated rsFC between all pairs of 47 ROIs (Figure 1; see Table S1 for the coordinates of the ROIs), which consisted of the frontolimbic circuitry or adolescent-depression network, as defined in a previous study (
). Pearson’s correlation coefficients were normalized through Fisher’s z-transformation.
An analysis of covariance (ANCOVA) was conducted to examine group differences (LR vs. HR vs. MDD) in dACC (center of mass [mm]: 7, 21, 32) connectivity with all other nodes, controlling for age, sex, and head movement (i.e., mean framewise displacement). False discovery rate (FDR) correction was applied to correct for the number of target nodes (i.e., 46). Next, we explored group differences in connectivity in any pair of 47 frontolimbic nodes using ANCOVA. We combined the connection-level threshold (uncorrected p < .001) and network-based statistics FDR-corrected p (pFDR) (by intensity) < .05 (two-sided). Any pattern of connectivity that showed a significant group difference was submitted to a pairwise comparison t test (Tukey corrected) to specify the pattern of group difference. With the same statistical threshold, we conducted multiple regression analyses with the independent variable of log-transformed MFQ-C (self-reported depressive symptoms) and covariates of age, sex, and head movement.
Graph Theory Analysis
To identify group differences in the topological properties of the frontolimbic circuit, which consisted of 47 nodes (Figure 1), we used CONN’s automated protocol to construct individuals’ graph theory measures. This protocol thresholded each participant’s 47 × 47 correlation matrix to generate an adjacency matrix. We adopted cost thresholding for constructing the adjacency matrix. To illustrate, if the cost value (i.e., K) is 0.15, only the pairs with the highest 15% of the correlation coefficient values have a value of 1, and all other pairs have a value of 0. Based on 100 simulations that generated 4 unique optimal cost values (i.e., 0.1198, 0.12997, 0.14015, 0.14986) (see Figure S1 for details), we defined our graph theory measures by averaging graph theory measures obtained using each of the 4 optimal values. We report results from both one-sided cost thresholding that considers only positive correlations when defining the highest K% connections and two-sided cost thresholding that considers both positive and negative correlations when defining the highest K% connections.
We first conducted an ANCOVA that examines the group difference of 6 graph theoretical measures of the dACC node: global efficiency, local efficiency, clustering coefficient, betweenness centrality, average path length, and degree (see Supplement for definitions). Age, sex, and head movement were entered as covariates. Next, we explored the group difference of graph theory measures of all other frontolimbic nodes. To correct for the number of nodes, we adopted a statistical threshold of two-sided pFDR < .05. To compare group differences in the integrated and segregated nature of the frontolimbic network as a whole, we conducted ANCOVA on the network-level global efficiency, local efficiency, and clustering coefficient (see Supplement for the calculation and choice of measures). Any measures that showed a significant group difference were submitted to a pairwise comparison t test (Tukey corrected) to specify the pattern of group difference. We conducted multiple regression analyses with the independent variable of log-transformed MFQ-C and covariates of age, sex, and head movement with the same statistical threshold to examine associations with continuous depression symptoms. To facilitate interpretation of the results, we identified the anatomical label of the nodes with significant results based on the Automated Anatomical Labeling atlas 3 (
Individuals’ rsFC maps with the seed regions of the left and right amygdala from the Automated Anatomical Labeling atlas 3 were estimated. We first conducted an ANCOVA with the search volume of an mPFC/ACC mask. The mPFC/ACC mask was created by combining medial orbitofrontal regions, rectus, superior medial prefrontal regions, and ACC of the Automated Anatomical Labeling atlas 3. We conducted a small-volume correction for the search region with familywise error rate correction provided in SPM12. The α value was divided by 2 to correct for 2 tests for left and right amygdala seeds. Then, whole-brain results were examined with a whole-brain pFDR < .025 threshold (two-sided). To explore the association between continuous self-reported depressive symptoms and amygdala connectivity with mPFC/ACC and whole brain, we ran multiple regression analyses with the independent variable of log-transformed MFQ-C and covariates of age, sex, and head movement. The same statistical threshold was used as ANCOVA.
After identifying any significant result, we identified outliers, defined as 3 standard deviations from the mean of the risk group that each participant belonged to (for ANCOVA) or the mean of all participants (for the regression analysis). We planned to report only the results that remained significant after excluding the outliers (see Table S2 for the number of excluded participants).
Demographic and Clinical Data
Demographic and clinical data are presented in Table 1.
Table 1Demographic, Clinical, and Head Motion Data
IDEA-RS was developed with data from the Pelotas 1993 Cohort Study. Using 11 sociodemographic variables measured at age 15, the model predicted risk of a current unipolar depressive episode at age 18 years [see (18,19) for more details].
The effect was driven by the difference in LR < HR (t123 = 3.40, p = .003), LR < MDD (t123 = 18.83, p < .001), and HR < MDD (t123 = 16.05, p < .001).
Censored Number of Volumes
Mean Framewise Displacement
HR, high-risk; IDEA-RS, Identifying Depression Early in Adolescence Risk Score; LR, low-risk; MDD, major depressive disorder; MFQ-C, Mood and Feelings Questionnaire-Child; WASI, Wechsler Abbreviated Scale of Intelligence.
a Categories were based on the Brazilian national census classification of race.
b IDEA-RS was developed with data from the Pelotas 1993 Cohort Study. Using 11 sociodemographic variables measured at age 15, the model predicted risk of a current unipolar depressive episode at age 18 years [see (
ROI-to-ROI Connectivity Within Frontolimbic Circuitry
We found that dACC connectivity with the posterior orbitofrontal cortex (OFC) (center of mass: 27, 20, −21) showed a significant group difference (F2,119 = 8.98, pFDR = .01), with the effect driven by greater connectivity of the MDD group compared with the LR group (t119 = 3.86, p < .001) and HR group (t119 = 3.39, p = .003) (Figure 2). Effects of the exploratory analysis with other nodes were not significant.
Graph Theory Measures Within Frontolimbic Circuitry
In the analysis with one-sided cost thresholding, the ANCOVA targeting dACC revealed a significant group difference in global efficiency driven by greater global efficiency of the MDD group compared with the LR group (Figure 3). In the analysis with two-sided cost thresholding, the analysis exploring all other frontolimbic nodes found significant group differences in local efficiency and clustering coefficient of left subgenual ACC (sgACC). For sgACC local efficiency, the group difference was driven by reduced local efficiency in the MDD group compared with the 2 risk groups. For the sgACC clustering coefficient, the group difference was driven by a decreasing pattern in the order of LR, HR, and MDD groups (Figure 4). The statistical values and center of mass of each node are presented in Table 2. Note that the results of the local efficiency and clustering coefficient of the sgACC could not include 18 participants because they did not have a neighboring subgraph, and 3 participants were excluded as outliers from the LE analysis (Table S2). There was no difference between included and excluded participants in clinical status and risk score (Tables S3 and S4).
Table 2Statistical Values for Analysis of Covariance of Graph Theory Measures
Graph Theory Measures
Pairwise Comparison Statistics
Right dACC (7, 21, 32)
F2,118 = 3.34, p = .039
LR < MDD: t118 = 2.419, p = .045
HR < MDD: t118 = 1.951, p = .129
LR < HR: t118 = 0.575, p = .834
Left sgACC (−5, 29, −10)
F2,99 = 11.6, pFDR = .001
LR > MDD: t99 = 4.71, p < .001
HR > MDD: t99 = 3.07, p = .008
LR > HR: t99 = 1.93, p = .135
Left sgACC (−5, 29, −10)
F2,102 = 11.47, pFDR = .002
LR > MDD: t102 = 4.77, p < .001
HR > MDD: t102 = 2.61, p = .028
LR > HR: t102 = 2.41, p = .047
CoM, center of mass; dACC, dorsal anterior cingulate cortex; FDR, false discovery rate; HR, high-risk; LR, low-risk; MDD, major depressive disorder; sgACC, subgenual ACC.
Multiple regression analysis with MFQ-C showed that greater self-reported depressive symptoms are associated with greater dACC global efficiency (t120 = 2.23, p = .03) and degree (t121 = 2.24, p = .03) when using one-sided cost thresholding and reduced sgACC local efficiency (t100 = −4.01, pFDR = .005) and clustering coefficient (t103 = −3.89, pFDR = .008) when using two-sided cost thresholding (Figure 5). Network-level global efficiency, local efficiency, and clustering coefficient did not show significant group differences or associations with self-reported depressive symptoms.
Amygdala Seed-Based Connectivity
Amygdala connectivity with the regions within mPFC/ACC or the whole brain did not show group differences or associations with MFQ-C.
The current study investigated how frontolimbic network connectivity is associated with the risk and presence of depression in Brazilian adolescents stratified using 11 sociodemographic variables and clinical assessment. We found that the MDD group showed greater dACC-OFC connectivity compared with the 2 risk groups and greater dACC global efficiency compared with the LR group. The MDD group showed reduced sgACC local efficiency and a lower clustering coefficient than the 2 risk groups, and the HR group showed a lower sgACC clustering coefficient than the LR group. Adolescents with greater self-reported depressive symptoms showed greater dACC global efficiency, greater dACC degree, reduced sgACC local efficiency, and a lower sgACC clustering coefficient. This study indicates that the risk and presence of adolescent depression in Brazil is associated with altered ACC connectivity patterns within frontolimbic circuitry.
We found several connectivity patterns of frontolimbic circuitry associated with the presence of depression, which supports the theory suggesting the critical role of interactions of the regions within the frontolimbic circuit in adolescent affective development (
), we found that adolescents with MDD showed greater dACC-OFC connectivity compared with adolescents in the other 2 groups and that they showed greater dACC global efficiency compared with adolescents in the LR group. This result indicates that greater and more efficient dACC connectivity may be a neural correlate or outcome of the development of depression in adolescents.
) showed heightened degree and efficiency of ACC in depressed adolescents in terms of connectivity with whole-brain regions, although the ACC was not divided into its subregions in these studies. Studies with a seed-based approach have also demonstrated that depressed adolescents showed heightened dACC rsFC with frontal (
), we speculate that heightened resting-state global efficiency of dACC in depressed adolescents may be associated with extensive transfer of salience signals detected from negative and self-relevant information or selecting avoidance behavior by weighting expected cost and deweighting expected benefit of normally rewarding events (e.g., social activities). As the prominent role of OFC is value updating (
), high dACC-OFC connectivity of depressed adolescents could be related to increased propensity to update the value of an object or behavior when it is associated with an outcome with high saliency, such as an unexpected or threatening experience. It should be noted that, in contrast to the dACC-OFC connectivity effect, there was no difference in dACC global efficiency between high-risk and depressed adolescents, which suggests that the high dACC global efficiency could not be attributed solely to the diagnosis of depression, but rather the combination of high-risk profiles and the clinical diagnosis.
Adolescents with MDD also showed reduced sgACC local efficiency and clustering coefficient compared with adolescents in the LR and HR groups. The sgACC has been a major target of depression treatment through deep brain stimulation (
), respectively. Our results extend the finding of major disruption of sgACC connectivity in adolescent depression by demonstrating the reduced interconnectedness of its neighboring regions. Importantly, we also observed a reduced sgACC clustering coefficient but not local efficiency in the HR group compared with the LR group. This result indicates that while both a reduced number of direct connections and inefficient connections among the neighboring regions of the sgACC are neural correlates or outcomes of depression, only a reduced number of direct connections serves as a potential risk factor for developing depression.
It is important to note that the observed abnormal efficiency of dACC and sgACC in depressed adolescents supports the recent proposal (
) suggesting that ACC connectivity is key for healthy behavioral development in multiple domains (e.g., relationships, achievement) owing to its hublike function of integrating multimodal inputs (e.g., social, cognitive, and visceral) to guide adaptive self-regulation. Interestingly, the analysis with self-reported depressive symptoms mirrored the findings from the group analysis, with an addition of greater number of edges originating from dACC (i.e., degree), suggesting that abnormal ACC topological properties are related not only to clinical diagnosis of depression but also to subjective experience of depressive symptoms across the subclinical to clinical range.
Contrary to our expectation, there was no group difference in amygdala–mPFC/ACC connectivity, which has consistently been implicated in depression in adolescents and adults in high-income countries (
), suggesting that this association may become more apparent when connectivity is elicited by a stimulus, such as a threatening face.
This study has several limitations. First, this study is cross-sectional. A longitudinal study that examines the intraindividual change in frontolimbic network topology before and after developing depression is needed to understand the timing of changes in topology in relation to developing depression. Second, although the IDEA-RS provides a more comprehensive approach to measuring depression risk, it did not include parental history of depression, as this cannot be assessed accurately through adolescent self-report. A future study is needed to systematically compare the IDEA-RS and parental history of depression to determine the advantages and disadvantages of the 2 approaches for understanding neural correlates of depression risk. Note that a study testing the predictive validity of the IDEA-RS within a different Brazilian sample found that the risk score improved prediction of depression risk above and beyond family history (
), suggesting that it may capture risk not captured by family history. Third, we did not examine whether the three topological properties altered in adolescents with MDD were associated with different symptom dimensions (e.g., decision-making ability, rumination, anhedonia). A future study that examines specific depression symptom dimensions would have implications for personalized treatment. Fourth, we did not collect a field map and did not apply distortion correction to the images, so results and/or lack of results for regions susceptible to distortion, such as the OFC, should be interpreted with caution.
In conclusion, with an underrepresented and extensively phenotyped Brazilian adolescent sample, we found that aberrant connectivity of the ACC in frontolimbic circuitry may be involved in risk for and the presence of depression in adolescence. The present study provides the first evidence to our knowledge that high-risk adolescents and adolescents with clinical depression show altered topology of the frontolimbic network implicated in adolescent affective brain development and depression. These results advance our knowledge on the atypical neural architecture of adolescents with depression and depression risk, specifically in Brazil, and will ultimately contribute to the prevention and treatment of adolescent depression across the globe.
Acknowledgments and Disclosures
We are extremely grateful to the schools and individuals who participated in this study, and to all members of the IDEA team for their dedication, hard work, and insights. The IDEA project, to which this research belongs, is supported by the MQ Brighter Futures Programme (Grant No. MQBF/1 IDEA [to VM, CK, and JRS]), Medical Research Council (Grant No. MC_PC_MR/R019460/1 [to CK and VM]), and Academy of Medical Sciences (Grant No. GCRFNG∖100281 [to CK]) under the Global Challenges Research Fund. This work was also supported by research grants from the Brazilian public funding agencies Conselho Nacional de Desenvolvimento Científico e Tecnológico (Grant Nos. 477129/2012-9 and 445828/2014-5 [to CK and LAR]), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (Grant No. 62/2014 [to CK]), and Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul (Grant No. 17/2551-0001009-4 [to CK]). CK is a Conselho Nacional de Desenvolvimento Científico e Tecnológico researcher and an Academy of Medical Sciences Newton Advanced Fellow. VM is supported by the Medical Research Foundation (Grant No. MRF-160-0005-ELPMONDE ) and National Institute for Health Research Mental Health Biomedical Research Centre at South London and Maudsley National Health Service Foundation Trust and King’s College London. The views expressed are those of the authors and not necessarily those of the National Health Service, National Institute for Health Research, or Department of Health and Social Care. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
LY had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. LY, VM, CK, and JRS were responsible for study concept and design. All authors were responsible for acquisition, analysis, and interpretation of data. LY drafted the original manuscript. All authors critically revised the manuscript for important intellectual content. LY performed statistical analysis. LAR, VM, CK, and JRS obtained funding.
The sociodemographic and self-reported measures were used and reported in another article (
), which was a preliminary study that examined the feasibility of MRI research in adolescents in Brazil. Task-based functional MRI data and sociodemographic and self-reported measures from this cohort of participants have been reported in another article (
VM has received research funding from Johnson & Johnson, a pharmaceutical company interested in the development of anti-inflammatory strategies for depression, but the research described in this article is unrelated to this funding. All other authors report no biomedical financial interests or potential conflicts of interest.