Perceived Racism, Brain Development, and Internalizing and Externalizing Symptoms: Findings From the ABCD Study

Abstract

Objective:

Racial discrimination drives health disparities among racial/ethnic minority youth, creating chronic stress that affects brain development and contributes to mental and behavioral health issues. This study analyzed data from the Adolescent Brain Cognitive Development (ABCD) Study to examine the neurobiological mechanisms linking discrimination to mental and behavioral health outcomes.

Method:

A sample of 3,321 racial/ethnic minority youth was split into training (80%, n = 2,674) and testing (20%, n = 647) groups. Propensity score–weighted machine learning was used to assess the effects of perceived discrimination on 2-year changes in resting-state functional connectivity between 3 subcortical regions (nucleus accumbens, amygdala, and hippocampus) and large-scale brain networks. Mediation analyses evaluated whether brain changes mediated sex-specific effects on internalizing or externalizing symptoms.

Results:

Perceived discrimination was significantly associated with 2-year changes in connectivity of the nucleus accumbens, amygdala, and hippocampus in both cross-validation and independent testing. Key findings included decreases in nucleus accumbens connectivity with retrosplenial–temporal and sensorimotor (hand) networks, decreases in amygdala connectivity with the sensorimotor (mouth) network, and increases in hippocampal connectivity with the auditory network. These changes suggest accelerated maturation in these connections among youth reporting higher discrimination levels. Moderated mediation analyses revealed sex differences, with discrimination-related changes in nucleus accumbens connectivity linked to poorer internalizing outcomes in female participants.

Conclusion:

The results indicate that perceived racial discrimination experienced in adolescence have an impact on subcortical–cortical brain development, which affects mental and behavioral health outcomes in a sex-specific manner.

Diversity & Inclusion Statement:

We worked to ensure race, ethnic, and/or other types of diversity in the recruitment of human participants. One or more of the authors of this paper self-identifies as a member of one or more historically underrepresented racial and/or ethnic groups in science.

Racism is a fundamental cause of health disparities between racial/ethnic minorities and non-Hispanic White people in the United States.^1,2 Racism operates through a complex system of structural inequities that restrict these minority groups’ full participation in society across various domains, ultimately undermining their health.³ Experiences of interpersonal racism, defined as unfair treatment of individuals based on race/ethnicity, are common in US society,⁴ and have consistently been linked to compromised mental and behavioral health.^5,6 The detrimental effects of interpersonal racism are particularly pronounced during early adolescence, a critical period when youth begin to recognize discriminatory actions but often lack effective coping strategies.⁷

According to the biopsychosocial model of racism and health,⁸ racism triggers neurobiological processes that have negative impacts on health outcomes. Indeed, the brain plays a central role in adapting to stress, and during adolescence—a critical period of rapid brain development—it exhibits remarkable structural and functional plasticity in response to environmental stressors.⁹ Thus, understanding how racism affects the developing adolescent brain is crucial for addressing racial/ethnic health disparities.

The limited literature on discrimination and brain functioning has mostly focused on individual brain regions.^10–12 For example, studies have found that Black and Latino youth living in high structural stigma environments exhibit smaller hippocampal volumes compared to those in lower stigma contexts.¹¹ In addition, greater exposure to discrimination has been associated with increased amygdala activity among Black and Hispanic older adults.¹² However, individual brain regions rarely function in isolation, and data on the effects of discrimination on functional connections, which can offer a more holistic understanding of brain function, are lacking.

Given the limited literature on the neural consequences of discrimination on brain development, we develop hypotheses using the growing literature on brain functional development following exposure to other forms of stressors. Recent well-powered studies have leveraged resting-state functional magnetic resonance imaging (fMRI) functional connectivity data from the large prospective Adolescent Brain Cognitive Development (ABCD) Study.¹³ For example, considering large-scale brain networks, low socioeconomic status has been associated with alterations in functional connections within and between sensorimotor, auditory, and frontal networks.¹⁴ Other studies have focused on functional connections between large-scale networks and subcortical regions (ie, amygdala and hippocampus) and have identified greater decreases in connectivity associated with lifetime adverse events, which in turn was associated with lower internalizing symptoms.¹⁵ Our recent work¹⁶ built on this prior literature and suggested the following: (1) exposure to adverse events during early adolescence particularly affects connections between subcortical and large-scale cortical networks (more than within or between large-scale networks); (2) these brain developmental effects represent accelerated brain development relative to adolescents with limited exposure to stress; and (3) somewhat counterintuitively, greater adversity-related changes in these brain connections relate to attenuated development of internalizing symptoms, suggesting that they reflect adaptive changes to cope with a stressful environment. Given these prior findings, we sought to test, using ABCD Study resting-state fMRI data, whether racial discrimination, an uncontrollable social evaluative stress, would have similar effects on brain development in racial/ethnic minority youth.

Based on prior studies of adversity, we examined whether racism is associated with brain functional connectivity development between large-scale cortical networks and subcortical brain regions, focusing on 3 bilateral subcortical regions: amygdala, hippocampus, and nucleus accumbens. These regions were selected for their well-known associations with exposure to stressors and the development of stress-related psychopathology.¹⁷ The amygdala and hippocampus are involved in associative fear learning and response to threat, and have been implicated in the pathophysiology of stress-related disorders such as anxiety, depression, and substance use.¹⁸ The nucleus accumbens plays a key role in reward processing and reinforcement learning,¹⁹ as well as responses to threat.¹¹ This region is robustly associated with depression, substance use, anxiety, and attention-deficit/hyperactivity disorder.^20,21 In particular, the psychiatric outcomes associated with these subcortical brain regions are known to be mediated by their connections with prefrontal cortical regions.²² Thus, we predicted that the development of the functional connections between these subcortical regions and large-scale cortical networks would serve as biological mechanisms linking racism with mental health outcomes among early adolescents. We intentionally focus on early adolescence, a critical period for brain development that is particularly sensitive to social stressors such as racism.²³ Early adolescence is also a time when interventions can be especially impactful. An improved understanding of how racism influences brain development during this stage can potentially be leveraged to develop interventions that effectively mitigate its negative effects.

Even though a few studies have examined the link between racial discrimination and brain functioning,^24,25 a key challenge is that exposure to racial discrimination is often entangled with numerous confounding variables, making it difficult to isolate the causal effect of racial discrimination on brain functioning. One way to address this is to use propensity score–weighted analysis (PSW). This approach weights participants based on a set of covariates associated with treatment assignment (in this case, discrimination) and the outcome of interest, aiming to balance covariate distribution across “treatment” and “control” groups, mimicking randomization. By adjusting for these covariates, PSW allows the closest possible approximation to a randomized experiment.²⁶ Using PSW can strengthen the causal estimates for the links between racial discrimination and brain functioning.

Sex differences in the brain targets of racial discrimination may account for sex differences in the expression of mental and behavioral health. Previous research has shown that boys and girls differ not only in how they experience racial discrimination²⁷ but also in how it affects them.²⁸ For example, boys tend to show stronger links between discrimination and conduct disorders, whereas girls are more likely to exhibit stronger connections between discrimination and internalizing symptoms.²⁸ In addition, there are salient sex differences in adolescent brain development,²⁹ which may further shape how brain pathways mediate the link between discrimination and adolescent outcomes.

In the current study, we conducted a PSW analysis of ABCD data, one of the first national, multisite neuroimaging studies in childhood and adolescence,¹³ to examine the relationship between experiences of racial discrimination during a 2-year period and concurrent brain development in a sample of racial/ethnic minority youth. Specifically, we hypothesized that perceived racial discrimination would alter development of subcortical-to-cortical functional connectivity of amygdala, hippocampus, and nucleus accumbens as measured with resting-state fMRI. Similar to prior studies of other forms of adversity,^14–16 we hypothesized that discrimination would be associated with more rapid brain development. In addition, we hypothesized that greater changes in the brain associated with discrimination would be associated with less expression of internalizing and externalizing symptoms. Finally, we explored sex differences in these relationships.

METHOD

ABCD Data Set and Participant Characteristics

Data were downloaded from the ABCD 5.1 data release.

Youth from racial/ethnic minority groups were included in the analysis if they had resting-state fMRI data at both baseline (8.9–11.1 years of age) and 2-year follow-up (10.6–13.8 years of age) and racial discrimination data for at least 1 of the first 2 follow-ups. A flow diagram of subjects included or excluded from analyses is provided in Figure S1, available online. Racial categories to determine inclusion in the analysis were defined from the parent-reported demographics survey at baseline with the race_ethnicity variable (ie, Asian, Black, Hispanic, Other, White), and all categories other than White were included. The total sample of 3,321 youth was divided into a training sample (80%, n = 2,674) and a testing sample (20%, n = 647). The sample was divided using the partition() function in R to balance the samples with regard to ABCD site and discrimination scores; sibling sets were constrained to be in the same sample. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

Racial Discrimination Measure

Racial discrimination was calculated as a mean score of 7 child-reported items assessing perceptions of racial discrimination (dim_y_ss_mean). Three of the questions queried, “How often do the following people treat you unfairly or negatively because of your ethnic background?” which corresponded with “teachers,” “other adults outside school,” and “other students.” The other questions were: “I feel that others behave in an unfair or negative way toward my ethnic group,” “I feel that I am not wanted in American society,” “I don’t feel accepted by other Americans,” and “I feel that other Americans have something against me.” Possible responses to each question included, 1 = almost never, 2 = rarely, 3 = sometimes, 4 = often, and 5 = very often.

The mean of the dim_y_ss_mean variable at the first and second follow-up visits were calculated to represent reported discrimination in this analysis. Values from a single time point were used in cases in which data were missing for either time point.

Adolescent Behavioral Outcomes

Internalizing and externalizing behaviors were measured by caregiver report on the Child Behavior Checklist³⁰ at both baseline and the 2-year follow-up. All analyses were conducted using raw scores where higher values correspond with more severe internalizing and externalizing symptoms.

Resting-State Functional Connectivity Measures and Processing

Brain development scores were estimated by calculating the residualized change between baseline and the 2-year follow-up for selected resting-state functional connections. Specifically, we tested functional connectivity between the 13 large-scale networks defined by the Gordon parcellation³¹ and bilateral subcortical regions-of-interest (ROIs): amygdala, hippocampus, and nucleus accumbens. Before calculating the functional connectivity change values, we cleaned the functional connectivity estimates at each time point by removing effects estimated by regression of age in months, scanner manufacturer, and mean and maximum motion estimates and their squares.

Propensity Score Calculation

Observational studies of the effects of discrimination benefit from the use of methods that address selection biases. Here we used inverse probability weighting, which is a robust propensity score method used to adjust for confounding and address missingness.³² Given the dimensional nature of the discrimination measure in this study, we used a generalized form of propensity score calculation that employs linear regression for continuous treatments.³³ Specifically, we calculated inverse probability weights to report discrimination using the propensity score method in WeightIt() in R software and including the following baseline variables: age, sex, pubertal status, scanner manufacturer, site, marital status of parents, household income, parental education, race and ethnicity, externalizing and internalizing scores (Child Behavior Checklist), adverse family experiences (sum of 7 items from the Parent Demographics Survey), functional connectivity values (within and between all Gordon networks, and between Gordon networks and all subcortical ROIs), and mean and maximum motion estimates for the baseline scan. When creating propensity score weights, race was defined from the parent demographic survey question, “What race do you consider the child to be? Please check all that apply.” Responses were binary indicator variables for each of the included races (eg, Black, Filipino, Korean, Native Hawaiian, White, etc), enabling inclusion of multiracial identities. Many included participants had White indicated as part of their racial identity (Table 1), reflecting multiracial identities and/or Hispanic ethnicity. Missing data in the predictors were handled with the “ind” option in Weightit(), which includes indicator variables (ie, 1 or 0) to denote missing observations for variables with missing data

TABLE 1.

Study Sample Descriptive Statistics of Racial/Ethnic Minority Sample (N = 3,321)

	n/ Mean	%/SD	Skewness	Kurtosis
Sociodemographic characteristics (baseline)
Age	9.47	0.51	0.08	−1.26
Sex at birth, n (%)
Female	1,609	48
Male	1,712	52
Race, n (%)
Asian	772	23
Black	985	30
Hispanic	1,425	43
Other	139	4
Household financial adversity	0.71	1.28	2.09	4.15
Parental education, n (%)
Less than high school	412	12
High school	515	16
Some college	1,182	36
Bachelor's degree	665	20
Post–bachelor's degree	547	17
Parental marital status, n (%)
Married	2,042	38
Not married	1,235	62
Key study variables, mean (SD)
Racial discrimination (mean of 1- and 2-y follow-up)	1.23	0.37	2.62	9.16
Internalizing problems (2-y follow-up)	4.78	5.70	2.04	5.28
Externalizing problems (2-y follow-up)	4.18	5.89	2.47	7.94

Elastic Net Prediction Analysis

Next, we tested whether functional connectivity development of the amygdala, hippocampus, and/or nucleus accumbens was related to perceived racial discrimination during the 2-year study period. For each of the 3 subcortical regions, we conducted regularized regression using elastic net regularization, because of the large number of functional connectivity predictors and high correlations between predictors. Because elastic net regression is a data-driven analysis, this approach allowed us to examine multiple brain connections in the absence of clear a priori hypotheses regarding the specific brain connections that may be affected. Specifically, there were 13 large-scale network connections for both the left and right ROI, resulting in 26 predictors in each of the 3 regression equations.

To optimize each model, analyses were conducted within a machine learning framework, following the methods of similar work examining effects of adversity on functional connectivity development in this dataset.¹⁶ The framework includes nested loops to prevent overfitting of the model and thereby optimizing prediction of out-of-sample observations, which is tested in cross-validation and independent test groups (Supplement 1, available online). We explicitly balanced racial identities (collapsed into Hispanic, Black, Asian, Hawaiian and Pacific Islanders, and other categories) for each sex across cross-validation partitions so that the estimated effects would be more likely to generalize across groups. Prediction accuracy was assessed as the Spearman correlation between the actual and predicted discrimination values in the training and testing samples to determine model significance. The elastic-net analysis was repeated once without weights, and a second time with observations weighted according to the calculated propensity scores of reported discrimination (ie, PSW) to minimize effects of confounding variables.

Because of the “backward” prediction of these models, namely, brain changes predicting discrimination, a Haufe transformation³⁴ was conducted to identify the predictor loading of each of the included connections (in this case, the Pearson correlation of each connection with the predicted values of discrimination) rather than examining the regression coefficients. For interpretative purposes, we highlight predictor loadings of |r| ≥ 0.5 for significant models.

Post hoc correlational analyses between discrimination scores and the individual functional connections identified in the machine learning analysis quantified the brain–behavioral relationships for each racial/ethnic group separately. Correlations were weighted using PSW.

Finally, we explored sex differences in the effects of racial discrimination on the brain by developing elastic-net models for male and female participants separately. Because these analyses reduced the sample size for machine learning analyses approximately by half, these analyses are considered secondary to the full-sample analyses described above.

Mediation Analyses

Finally, we conducted mediation analyses to examine whether brain developmental changes associated with discrimination mediate the relationship between perceived discrimination and internalizing and externalizing symptoms at the 2-year follow-up (Figure S2, available online, provides a conceptual model). Data were analyzed using Mplus 8.3 with maximum likelihood estimation with robust standard errors (MLR), which provides estimations of standard errors that are robust to nonnormal data.³⁵ Internalizing and externalizing symptoms were analyzed in the same model, whereas different brain connections were analyzed in separate models. The indirect effects of discrimination, changes in brain connectivity, and internalizing and externalizing symptoms were estimated. All analyses controlled for youth age, assigned sex at birth (ie, female or male), race, ethnicity, household financial adversity, parental educational attainment, and parental marital status. Race was collapsed into 4 categories (ie, Asian, Black, Hispanic, and other; Hispanic was the reference group) to ensure model convergence. To address the missing data, we used full-information maximum likelihood (FIML), which estimates model parameters using all available data without imputing missing values.³⁶

To examine sex differences in the mediating pathways, we conduced moderated mediation analyses by adding the discrimination–sex interaction in A path and brain connectivity–sex interaction in B path (Figure S2, available online). If the interaction was significant, we conducted simple slope analyses.³⁷ We also estimated the indirect effects separately for boys and girls.

RESULTS

Participants

Sociodemographic characteristics and discrimination frequency data are presented in Table 1. Average discrimination scores ranged from 1.0 to 4.4, with a median of 1.1. The sociodemographic characteristics of adolescents who were excluded from the analyses are presented in Table S1, available online.

Elastic Net Prediction

Univariate correlational relationships between each tested functional connection are reported in Table S2, available online. Prediction accuracies of the multivariate elastic net machine learning analysis for each subcortical region are presented in Table S3, available online. Elastic net regression models, which included bilateral subcortical-to-cortical network functional connectivity changes as predictors, significantly predicted reported discrimination in both the training and test sets for all 3 subcortical regions tested. Results from models that did not incorporate propensity weights were similar (Table S3, available online).

Predictor loadings for each connection between cortical networks and the subcortical regions are displayed for the PSW analysis in Figure 2 and reported in Table S2, available online. For the nucleus accumbens model, key predictors, determined a priori as a predictor loading of |r| ≥ 0.5, included the right nucleus accumbens to retrosplenial–temporal network connection (r = −0.67) and right accumbens to sensorimotor (hand) network connection (r = −0.74). For the amygdala model, the connection between the left amygdala and sensorimotor (mouth) network had a loading of r = −0.70. For the hippocampus model, the predictor loadings were more distributed, and no connection exceeded the a priori determined threshold. However, the predictor with the strongest loading for that model was the connection between the right hippocampus and auditory network (r = 0.46). We highlight each of these connections in Figure 1 and subsequent analyses. Results from the unweighted models were highly similar (Tables S2 and S3, available online, and Figure S3, available online).

FIGURE 2.

Line Plots of Functional Connectivity at Baseline and Two-Year Follow-Up

Note: Data are plotted by median split of reported discrimination median (dim_y_ss_mean = 1.071). Means (propensity score–weighted) and standard errors of functional connectivity are plotted for each subgroup at each timepoint. L = left; R = right.

FIGURE 1.

Results From Propensity Score–Weighted Elastic Net Regression

Note: Brain maps of predictor loadings (based on Pearson r values) for connections between each subcortical region for each large-scale cortical network as defined by the Gordon brain parcellation. Parcels corresponding to the same cortical network are mapped with identical values.

Functional connectivity at baseline and 2-year follow-up are plotted by median split of reported discrimination in Figure 2. Consistent with the negative predictor loadings, higher levels of discrimination were associated with greater decreases in functional connectivity relative to those in individuals who experienced less discrimination for the right nucleus accumbens to retrosplenial–temporal network connection, the right accumbens to sensorimotor (hand) network connection, and the left amygdala to sensorimotor (mouth) network connection. On the other hand, the connection between the right hippocampus and auditory network exhibited greater increases in functional connectivity in individuals reporting greater discrimination. Taken together, these results suggest that the brain changes associated with discrimination represent an acceleration of typical brain maturational processes. Finally, sex-specific models were explored; results are available in Table S4, available online, and in Figures S4 and S5, available online. Briefly, for female participants, models were significant in the training group for the nucleus accumbens and amygdala, whereas a model with hippocampus connectivity was significant only for males. It is important to note that power was reduced for these exploratory analyses because of the divided sample, which likely reduced the number of significant findings, particularly in the independent test group.

Post hoc correlational analyses of the effects of discrimination on brain functional connectivity development are reported in Table S5, available online, for each racial/ethnic group separately.

Mediation Analyses

Mediation results of the full sample are reported in Table S6, available online. With the full sample, discrimination was significantly related to brain changes, and discrimination was significantly associated with increased externalizing symptoms and marginally associated with internalizing symptoms. However, the links between brain changes and behavioral outcomes were not significant for any of the tested connections. There was also no significant mediation (indirect) pathway detected.

However, we observed significant sex differences in these relationships. Specifically, sex significantly moderated the link between nucleus accumbens(right)–retrosplenial temporal connectivity and adolescent internalizing (b = −4.13, p = .01) and externalizing symptoms (b = −3.04, p = .04), as reported in Table S7, available online. Specifically, as seen in Figure 3A, for girls, weaker connectivity between right nucleus accumbens and retrosplenial–temporal network was significantly associated with higher levels of internalizing symptoms (b = −3.26, p = .02). In contrast, this relationship was positive but not significant for boys (b = 0.88, p = .36). For externalizing symptoms, although the simple slope analyses were not statistically significant, a similar trend of sex differences was observed (Figure 3B).

FIGURE 3.

Line Plots of the Moderating Effects of Sex

Note: The link between nucleus accumbens(right)–retrosplenial temporal connectivity and internalizing symptoms (A) and externalizing symptoms (B) was moderated by sex.

We also found significant sex difference in the indirect effects of discrimination, nucleus accumbens(right)–retrosplenial temporal connectivity, and adolescent outcomes (Figure 4; Table S8, available online). For girls, we found significant indirect effects: greater discrimination was associated with weaker nucleus accumbens (right) and retrosplenial–temporal network connectivity, which in turn was associated with higher levels of internalizing symptoms. This indirect effect was not significant for boys. There was no moderating mediation effect of sex for other brain connections.

FIGURE 4.

Model of the Moderating Effects of Sex

Note: The link between nucleus accumbens(right)–retrosplenial temporal connectivity and internalizing symptoms was moderated by sex. Solid lines indicate significant relationships, whereas dashed lines indicate nonsignificant relationships.

DISCUSSION

Perceived racial discrimination was associated with altered development of functional connections between subcortical regions and cortical networks. Specifically, propensity-weighted regularized regression models of nucleus accumbens, amygdala, and hippocampus functional connectivity development significantly predicted the severity of racial discrimination experiences. Furthermore, consistent with effects of other forms of adversity in adolescents,¹⁶ analyses revealed that these brain changes represented an accelerated maturation of these brain connections. However, unlike the adaptive brain changes associated with other forms of adversity, these brain changes were not associated with reductions in internalizing psychopathology. In fact, these changes partially accounted for the increased psychopathology associated with discrimination experiences, an effect that was driven by female sex participants. These findings support empirical evidence of the harmful effects of racial discrimination on adolescent outcomes, and highlight an apparent lack of brain functional adaptations observed for other forms of stress. The results contribute to the growing body of research suggesting that early adolescence is particularly susceptible to the negative impact of discriminatory stressors, leading to adverse mental and behavioral health outcomes.

Discrimination Effects on Subcortical Functional Connectivity

The findings highlight an important role of subcortical function in the effects of discrimination experiences. Here, we identified impacts on the amygdala and hippocampus, which have been previously associated with discrimination based on task or structural measures,^11,12,38 and the nucleus accumbens, which has not previously been identified as a brain target of discrimination. Prior work has also identified greater developmental decreases in connectivity between the hippocampus and amygdala with sensorimotor and frontal networks following exposure to stressful events in the ABCD Study.^15,16 However, previously identified connections were not major predictors of discrimination stress in this sample. Despite apparent differences in the effects of discrimination vs other forms of adversity on brain connectivity, the current findings support the impact of stressors on these subcortical regions. In the current study, a key finding was a mediation effect involving connectivity between the nucleus accumbens and the retrosplenial–temporal network. This network involves the retrosplenial cortex and proximal temporal areas, which are considered part of the extended default mode network and are functionally connected with the medial temporal lobe.³⁹ This network has roles in self-referential processing,³⁹ and, given its identified association with internalizing symptoms (Figures 3 and 4), we speculate that this connection could relate to poorer self-concept in youth experiencing discrimination.

Accelerated Brain Maturation

The hypothesis that brain changes associated with racial discrimination experiences could be related to more rapid maturation in the affected brain connections was supported by the current findings. Figure 2 supports this interpretation, as there was a relatively steeper slope in brain development for participants reporting greater discrimination. Specifically, 3 connections demonstrated decreasing functional connectivity from baseline to the 2-year follow-up, even in individuals experiencing less discrimination; however, those decreases were steeper for individuals experiencing more discrimination. Similarly, the connection between the hippocampus and auditory network was increasing with development; however, the rate of increase was steeper for individuals experiencing more discrimination. Previous studies have identified accelerated maturation of physical development based on pubertal staging in response to discrimination.⁴⁰ Relatedly, accelerated brain development in response to stressors has been shown to be mediated through earlier puberty.⁴¹ Such brain findings have been proposed to be related to a maturing emotion regulation circuitry.^15,16 However, some of the connections that were primarily affected in the current study, such as those involving sensorimotor and auditory networks, could simply reflect the enhanced maturational trajectory in connections undergoing the most rapid development during this developmental time period (ie, ages 10–12 years).^14,42 Indeed, there was no indication that the brain developmental effects of perceived discrimination in this study related to adaptive changes associated with resilient outcomes based on CBCL scores.

Role of Brain Changes in Development of Internalizing and Externalizing Outcomes: Sex Differences

Mediation models suggested an indirect effect of discrimination scores on internalizing symptoms via reduced connectivity between nucleus accumbens (right) and retrosplenial–temporal regions, among female participants. Specifically, these findings suggest that decreases in functional connectivity in response to racial discrimination in girls were linked to increases in internalizing symptoms. In contrast, developmental decreases in several other cortical–subcortical functional connections have previously been shown to have adaptive benefits in response to other forms of stress.^15,16 This divergence underscores key differences between racial discrimination and other forms of adversity. Racial discrimination is chronic, pervasive, uncontrollable, and often lacks a clear resolution.⁴³ Its direct targeting of ethnic/racial minority adolescents’ developing ethnic/racial identities may further contribute to its harmful effects.

The stronger mediation effect in girls is consistent with prior studies suggesting that stress during puberty has stronger proximal effects on girls compared with boys, including an increased risk for developing mood and stress-related disorders,⁴⁴ and steeper symptom trajectories.⁴⁵ Internalizing symptoms, encompassing emotional liability, anhedonia, anxiousness, distress, and fear, are highly prevalent among children and adolescents, particularly adolescent girls.⁴⁶ Relevant to this study, sex differences in the brain and its response to stress-related experiences involve complex independent and interdependent chromosomal, hormonal, and epigenetic differences.⁴⁷ Significant sex differences in the response to environmental threats during adolescence might also stem from early life differences in rearing and socio-cultural practices that differentially shape the developing brain.⁴⁸ It is worth noting that although models controlled for chronological age and puberty stage, residual confounding and maturational processes could lead to observed sex differences, as brain developmental trajectories differ by sex.⁴⁹ Significant gaps remain in the understanding of brain mechanisms involved in sex and sex-by-race-ethnicity differences. Our findings highlight the need to continue investigating sex differences in the circuits and mechanisms involved in the occurrence and manifestation of mental health disorders among adolescents, as well as the timing, dosing, chronicity, and agents by which social and structural etiological factors generate and sustain mental health disparities among minority youth.

The current study makes significant contributions to the literature on racism. First, it is among the first to provide evidence that experiences of racial discrimination can alter brain functional connectivity in early adolescence, suggesting that such experiences may become biologically embedded during this critical period of development. By leveraging the ABCD Study—a national dataset—we had unprecedented statistical power to detect the effects of perceived discrimination on brain function at a population level. Furthermore, the use of propensity score weighting (PSW) to identify causal associations, combined with rigorous validation methods, adds robustness and precision to these findings, strengthening the validity of conclusions. Finally, this study provides novel evidence of a differential sex effects by which perceived discrimination affects the connectivity of reward-related and motivational processing networks and indirectly increases the risk for internalized symptomatology through altering developmental processes within those networks among female participants. Studies focusing on both biological sex and gender differences to further investigate these findings are warranted. This study adds to the growing body of population neuroscience research by providing information that could be incorporated in neuroscience-informed psychoeducation programs designed to reduce mental health and substance use by improving neuroscientific literacy.⁵⁰ Furthermore, our sex-specific findings help to identify and prioritize subgroups (eg, girls) who would greatly benefit from preventive interventions buffering the effect of discrimination on mental health. Finally, the findings provide preliminary insights on potential neurobiomarkers to consider when developing clinical trials examining discrimination preventive interventions.

The current study has some limitations that should be noted. First, although resting-state fMRI has many advantages, it limits our ability to make functional inferences, as it is not acquired in the context of a behavioral task. Although task-based fMRI data are also acquired in the ABCD Study, higher rates of missing and low-quality data would have limited power to detect effects, especially because the analytical sample was already substantially reduced from the full sample. Second, this study did not account for other forms of racism (eg, institutional and systemic racism), nor did it examine the role of perpetrator (eg, teacher, peer) in experiences of discrimination. Third, our study provides only a snapshot of development in early adolescence. The potential long-term effects of racism on both brain development and overall health remain unexplored and warrant further investigation. Fourth, although our study gave some consideration to subgroup variations based on biological sex or racial/ethnic identity, there are likely important racial/ethnic differences in the intensity, chronicity, and context of the lived experiences of discrimination, which can be an important research direction for future research. In addition, this study relied on retrospective reporting of discrimination and lacked precision regarding the temporal proximity of discrimination experiences to brain changes. Fifth, there is still a limited understanding of alternative and modifiable individual, interpersonal or contextual level factors that might mitigate or counteract the deleterious effects of discrimination on brain development and function. Further research examining this issue is needed, with a particular focus on the timing and dosing at which modifying these mechanisms might alleviate or offset the impact of adverse experiences. Sixth, our sample excluded a disproportionate percentage of non-Hispanic Black participants, as well as those whose parents were unmarried. These subgroups were excluded primarily for having missing imaging data needed to conduct the analyses. Therefore, our results should be interpreted in light of this limitation, and highlight the need to implement strategies that facilitate continuous participation of these subgroups in longitudinal studies.

In summary, the effects of racism are wired in brain functional connectivity. Using rigorous methodology, the findings demonstrate that youth discrimination experiences are associated with altered development of hippocampal, amygdala, and nucleus accumbens function. These brain regions offer potential targets for future investigations. Finally, there was a notable role of sex differences in discrimination effects that deserves further exploration to understand the differential impact of discrimination experiences on the development of boys and girls.

Data Sharing:

Data used in the preparation of this article were obtained from the Adolescent Brain Cognitive Development^SM (ABCD) Study (https://abcdstudy.org), held in the NIMH Data Archive (NDA). This is a multisite, longitudinal study designed to recruit more than 10,000 children age 9–10 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 numbers U01DA041048, U01DA050989, U01DA051016, U01DA041022, U01DA051018, U01DA051037, U01DA050987, U01DA041174, U01DA041106, U01DA041117, U01DA041028, U01DA041134, U01DA050988, U01DA051039, U01DA041156, U01DA041025, U01DA041120, U01DA051038, U01DA041148, U01DA041093, U01DA041089, U24DA041123, U24DA041147. A full list of supporters is available at https://abcdstudy.org/federal-partners.html. A listing of participating sites and a complete listing of the study investigators can be found at https://abcdstudy.org/consortium_members/. ABCD consortium investigators designed and implemented the study and/or provided data but did not necessarily participate in the analysis or writing of this report. This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or ABCD consortium investigators. The ABCD data repository grows and changes over time. The ABCD data used in this report came from 10.15154/1523041. DOIs can be found at https://doi.org/10.15154/1523041.