Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Aug 16.
Published in final edited form as: Biol Psychiatry Cogn Neurosci Neuroimaging. 2020 Feb 11;5(5):499–509. doi: 10.1016/j.bpsc.2020.01.011

Impact of Childhood Trauma on Executive Function in Adolescence—Mediating Functional Brain Networks and Prediction of High-Risk Drinking

Sarita Silveira 1, Rutvik Shah 2, Kate B Nooner 3, Bonnie J Nagel 4, Susan F Tapert 5, Michael D de Bellis 6, Jyoti Mishra 7
PMCID: PMC8366521  NIHMSID: NIHMS1723274  PMID: 32299789

Abstract

BACKGROUND:

Childhood trauma is known to impart risk for several adverse life outcomes. Yet, its impact during adolescent development is not well understood. We aimed to investigate the relationships among childhood trauma, functional brain connectivity, executive dysfunction (ED), and the development of high-risk drinking in adolescence.

METHODS:

Data from the National Consortium on Alcohol and Neurodevelopment in Adolescence (sample size = 392, 55% female) cohort were used. This included resting-state functional magnetic resonance imaging at baseline, childhood trauma and ED self-reports, and detailed interviews on alcohol and substance use collected at baseline and at 4 annual follow-ups. We used longitudinal regression analyses to confirm the relationship between childhood trauma and ED, identified the mediating functional brain network hubs, and used these linkages to predict future high-risk drinking in adolescence.

RESULTS:

Childhood trauma severity was significantly related to ED in all years. At baseline, distributed functional connectivity from hub regions in the bilateral dorsal anterior cingulate cortex, right anterior insula, right intraparietal sulcus, and bilateral pre- and postcentral gyri mediated the relationship between childhood trauma and ED. Furthermore, high-risk drinking in follow-up years 1–4 could be predicted with high accuracy from the trauma-affected functional brain networks that mediated ED at baseline, together with age, childhood trauma severity, and extent of ED.

DISCUSSION:

Functional brain networks, particularly from hub regions important for cognitive and sensorimotor control, explain the relationship between childhood trauma and ED and are important for predicting future high-risk drinking. These findings are relevant for the prognosis of alcohol use disorders.

Keywords: Adolescence, Binge drinking, Brain networks, Childhood trauma, Development, Executive function

Exposures to neglect and/or abuse during childhood are core dimensions of childhood trauma (CT). Much research associates CT with adverse physical and mental health outcomes (1,2), as well as increased rates of risky alcohol use, younger drinking onset, and alcohol use disorder (AUD) diagnosis (3–7). However, most findings are from adult population-based data with retrospective reports of CT (8,9). There are very few prospective studies that longitudinally follow adolescents, and notably, their findings have been inconsistent in terms of identifying CT and risky alcohol use relationships, partly because of a focus on different alcohol use outcome variables based on behavioral/diagnostic criteria. Thus, it still remains unclear how heavy drinking risk in youths with CT develops during adolescence. Furthermore, the neurocognitive mechanisms underlying susceptibility to adverse alcohol use patterns in this vulnerable population are not known.

Identifying mechanisms and pathways that link CT and adverse life outcomes is important for informing prevention and intervention efforts during adolescence and young adult-hood. Previous literature suggests that CT triggers a cascade of neurobiological events that lead to enduring alterations in susceptible individuals. These include neurohumoral and proinflammatory responses to severe and chronic stress, as well as changes in brain development, brain network architecture, and brain circuits (9,10). Previous research in clinical and nonclinical populations has established a link between the neurodevelopmental changes caused by CT and deficits in executive function, that is, the human capacity to control and coordinate thoughts and behavior (11,12). Executive dysfunction (ED) is a transdiagnostic factor in various psychopathologies (13,14) and a neurofunctional domain that is postulated to be relevant to early onset of risky alcohol use and the development and maintenance of AUDs (15). Furthermore, behavioral studies have corroborated that dysregulated self-control can mediate the relationship between CT and alcohol use (16).

A few recent neuroimaging studies demonstrate an impact of CT on resting-state functional brain connectivity (17–20). Although those studies provide evidence for trauma-related reconfiguration of functional brain networks (20), they have been limited by small samples and/or have primarily focused on sampling adults with a specific mental disorder, such as patients with major depressive disorder (MDD) versus healthy control subjects. Resting-state functional brain networks, especially those that subserve executive function, are rapidly developing during typical adolescence (21–23), yet how CT relates to this network development is not understood. Typically, cortical hubs that are central to connectivity within brain networks drive network maturation (24); in this context, the dorsal anterior cingulate cortex (dACC) is posited as one of the main hubs in neurodevelopment that drives segregation and maturation of executive control networks (25,26). In related research, we recently found that CT is associated with weaker functional connectivity strength from the dACC to the anterior insular cortex, one of the core regions of the cingulo-opercular network (CON), during adolescence (27). In addition, in adults, maltreatment-related changes of functional connectivity of the dACC to cortical areas including the inferior frontal cortex and precuneus have been linked to dysfunctional self-regulation and inhibitory control (18,28). In the present study, we aimed to bridge several lines of evidence, and directly investigated the relationships among CT, functional brain connectivity, executive functioning, and the development of at-risk alcohol use behaviors in adolescence. Given that different forms of CT, abuse versus neglect, can have differential biological and neurofunctional impacts (29–32), we also investigated those categories separately. We used a large sample of adolescents who completed CT reports from the longitudinal National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA) study (33) and hypothesized that CT will relate to weaker intrinsic functional brain network connectivity of hub regions, especially dACC connectivity to other maturing cognitive control regions, and that these networks will mediate ED in adolescence. Finally, we explored whether trauma-related ED and functional network connections mediating these impairments would together predict susceptibility to high-risk drinking in future years.

METHODS AND MATERIALS

Participants

In this study, we used data of 392 adolescents (mean age ± SD = 17.36 ± 2.53 years, range = 12–22 years) from the multisite NCANDA cohort. For a detailed description and comparison of the current sample with the full NCANDA cohort (N = 831), see Participants: Comparison of Selected Study Sample to Full NCANDA Cohort in the Supplement. Self-identified ethnicity was assessed with predefined categories. Highest grade of parental education from 1 (first grade) to 20 (≥4 years of graduate/professional school) on the Pediatric Imaging, Neurocognition, and Genetics Study Demographics and Child Health History Questionnaire served as an indicator for socioeconomic status (34), as used in previous NCANDA analyses (35). AUD family history density index in first- and second-degree relatives was calculated ([no. of positive parents] + [no. of positive grandparents × 0.5]; range of 0–4) (36).

The NCANDA study was approved by the institutional review board of each NCANDA site. All subjects provided informed assent or consent annually. When under the age of 18 years, the participant’s legal guardians provided written informed consent/permission to participate and the youths provided written informed assent. When aged 18 years or older, the participant provided written informed consent.

Measures

CT was assessed with the 28-item Childhood Trauma Questionnaire (CTQ) (37), a self-administered retrospective lifetime measure that inquires about child neglect and abuse experiences in 5 categories: emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect. Each category entails 5 items that are rated on a 5-point Likert-type scale ranging from 1 = never true to 5 = very often true. Scores for the 5 items in each category were summed. For the purpose of sample description, participants were classified as positive for a history of “low to moderate” trauma severity for a specific trauma category based on validated cutoff scores (38) (Supplemental Table S1). In the following analyses, we used individual CTQ scores averaged across all subscales for all participants (overall score, mean ± SD = 1.35 ± 0.38, range = 1–3.48, Cronbach’s α = 0.68). We also tested whether an association between CT and executive function is consistent across averaged abuse and neglect subscores. Some of the 392 adolescents completed CTQ reports at multiple time points. We used the first CTQ score that was acquired for each individual, with 245 responses at the baseline visit (BL), 102 at year 1 (Y1), and 45 at Y2; CTQ score reports were highly consistent across years with Cronbach’s α = 0.87. Mean age of CTQ report acquisition was 17.86 ± 2.85 years. Each NCANDA study site had an institutional review board–approved protocol and specific training for staff as mandated reporters whenever it was learned that a child (or elder) had been abused. There were situations in which reports were filed with the appropriate authorities following the guidelines specified in the site’s approved human subjects protocol.

Executive function was assessed using the 80-item self-report version of the Behavior Rating Inventory of Executive Function (BRIEF) (39,40) at BL and yearly follow-up visits. Items are rated on a 3-point Likert-type scale (0 = never, 1 = sometimes, and 2 = often), subsumed in 2 indices as well as a global executive composite. The Behavioral Regulation Index comprises scales on inhibitory control and impulsivity, cognitive flexibility, emotional control, and self-monitoring, and the Metacognition Index comprises scales on working memory (i.e., the maintenance of information), task planning, initiation, monitoring, and completion, and includes anticipation of action outcomes. The BRIEF is the most commonly used rating scale of executive function, developed to provide an ecologically valid indicator of behavior in real, goal-directed environments (41). Owing to inflated frequencies of low scores in our sample, raw BRIEF scores were transformed to normality using an inverse distribution function applied to the scores’ fractional ranks (42).

Substance use disorders and other psychiatric conditions were assessed in accordance with DSM-IV and DSM-5 criteria (43) using the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) (Supplemental Table S2). SSAGA was assessed at BL and follow-up years Y1, Y2, and Y3 but not Y4. Because of low prevalence of AUD (1.1%–2.9% at Y1, Y2, and Y3, total n = 13), we chose an alcohol use behavior outcome to represent high-risk use in follow-up years (see below). The inclusion of comorbid mental health disorders in further analyses was limited to MDD and a general factor for anxiety disorders based on binarized prevalence of generalized anxiety disorder, panic disorder, manic episode, posttraumatic stress disorder, and obsessive-compulsive disorder.

Alcohol use behavior was measured using the Customary Drinking and Drug Use Record (44), completed by participants at BL and each annual follow-up visit (Y1, Y2, Y3, and Y4). Empirically, binge drinking thresholds are defined as ≥4 drinks for women and ≥5 drinks for men per occasion. Particularly high frequency of binge drinking episodes is defined as at least once per week in the past 12 months (National Epidemiological Survey on Alcohol and Related Conditions) (45), which confers risk for alcohol-related morbidity and mortality. Thus, to classify high-risk drinkers, we used the item “During the past year, have you regularly consumed 4+ (females)/5+ (males) drinks within an occasion (at least once a week)?” and binarized individuals who reported “yes” in any year. In the current sample, this high-risk drinking criterion showed significant overlap with AUD individuals in each year (Y1: χ2 = 13.75, p = .019; Y2: χ2 = 11.96, p = .013; Y3 χ2 = 10.59, p = .010). We further included binarized variables for frequent use (i.e., at least once per week) of tobacco, marijuana, and other illicit drugs (see Table 1) because of potential substance use–induced changes in cognitive systems (46,47).

Table 1.

Prevalence of Alcohol and Substance Use in Number of Individuals at Baseline and Follow-up Assessments After 1, 2, 3, and 4 Years

High-Risk Behavior Prevalence
BL (n = 241) Y1 (n = 350) Y2 (n = 330) Y3 (n = 318) Y4 (n = 289)
High-Risk Drinking 20 22 26 34 46
Frequent Tobacco Use 5 12 19 16 17
Frequent Marijuana Use 13 30 34 45 51
Frequent Use of Other Illicit Drugs 1 2 2 2 3
Consistency Across Years
χ2 BL-Y1 χ2 Y1-Y2 χ2 Y2-Y3 χ2 Y3-Y4
High-Risk Drinking 39.30a 61.40a 99.48a 94.66a
Frequent Tobacco Use 63.32a 112.01a 152.52a 124.02a
Frequent Marijuana Use 56.84a 193.47a 179.95a 171.52a
Frequent Use of Other Illicit Drugs 64.50b 0.01 (ns) 0.01 (ns) 546.00a
Open in a new tab

n denotes the total number of completed Customary Drinking and Drug Use Record reports in each year among N = 392 participants with baseline imaging and lifetime Childhood Trauma Questionnaire data. High-risk drinking and frequent use of tobacco, marijuana, or other illicit drugs is defined as consumption at least once per week.

BL, baseline; ns, nonsignificant; Y1–Y4, 1–4 years after baseline.

a

p < .001.

b

p < .05.

Imaging Acquisition

We used functional brain connectivity measures that were assessed at BL for all 392 participants using resting-state functional magnetic resonance imaging (rs-fMRI). Neuroimaging data were acquired in the NCANDA study on 2 different 3.0T MRI scanner models: 38% of participants were scanned with a Magnetom Trio Trim (Siemens Healthcare, Erlangen, Germany) and 62% on an MR750 (GE Healthcare, Waukesha, WI). Scan protocols were consistent between sites and scanners. High-resolution anatomical T1 scans (slice thickness = 1.2 mm, repetition time = 1900 ms, echo time = 2.92 ms, flip angle = 11°, field of view = 240 mm) and gradient echo-planar images were collected (slice thickness = 5 mm, 32 slices, repetition time = 2200 ms, echo time = 30 ms, flip angle = 79°, field of view = 240, 240 × 240 matrix).

Data Analysis

Statistical analyses were conducted using MATLAB (The Math-works, Inc., Natick, MA) and SPSS 19.0 (IBM Corp., Armonk, NY). In a generalized linear mixed model (GLMM), individual CTQ mean scores were used as fixed regressor to explain variance in BRIEF executive function indices, with time of assessment (BL, Y1, Y2, Y3, Y4) as repeated-measures factor. Sex, ethnicity, years of parental education, and AUD family history density at BL, as well as 6 repeated measures of time, age, high-risk drinking, and frequent use of tobacco, marijuana, or other illicit drugs, were modeled as fixed factors, as all these variables can have an impact on executive function. Subject identifiers and time of assessment were modeled as random factors to account for individual differences in intercepts of ED at BL and in slopes of growth over time. A total of 743 out of 1960 cases were excluded from this GLMM analysis because of missing values at follow-up visits. Each participant contributed an average of 4.67 ± 2.21 data points on the 6 repeated measures in respective years to the analysis. We also computed this repeated-measures GLMM for BL through Y4 data for the BRIEF metacognition and behavioral regulation sub-indices, to test consistency of results with the overall BRIEF scores. Furthermore, we tested whether resulting patterns are consistent across forms of CT, that is, abuse and neglect, using averaged scores from the respective CTQ subscales. Prevalence of psychiatric diagnoses, specifically of MDD and anxiety disorders, was not included in the above models owing to unavailable SSAGA data in Y4; therefore, a further repeated-measures GLMM was computed to test consistency of results across BL through Y3 including these SSAGA variables. In this GLMM, 555 out of 1568 cases were disregarded because of missing values at follow-up visits. Each participant contributed an average of 6.57 ± 2.61 data points on 8 repeated measures in respective years to the analysis.

Functional Connectivity Analyses.

Resting-state fMRI data were analyzed using the CONN functional connectivity toolbox (https://www.nitrc.org/projects/conn) (48) (see MRI Data Preprocessing in the Supplement). To investigate functional connectivity mediators of CTQ effects on ED (BRIEF), we used the Mediation Effect Parametric Mapping toolbox (https://github.com/canlab/MediationToolbox) (49); this analysis was conducted on BL data, because all 392 adolescents had complete datasets for CTQ, BRIEF, and neuroimaging measures with no missing data at this time point (see Mediation Effect Parametric Mapping in the Supplement for detailed description). The Mediation Effect Parametric Mapping analysis is based on a standard 3-variable path model (50) with a bootstrap test for the statistical significance of the product a×b (51,52), where a represents the relationship between the predictor (CTQ) and mediator, b represents the relationship between the mediator and outcome (BRIEF), and a×b is the difference between the direct predictor-outcome relationship (c) and the predictor-outcome relationship taking the mediator into account (cʹ).

Prediction of Future High-Risk Drinking.

Our goal was to test whether we can predict high-risk drinking in follow-up years (i.e., at least once-per-week binge drinking in the past year) based on BL data. While multiple statistical and machine learning approaches are available for binary classification problems, the real-world diagnostic potential of individual techniques remains unclear in many cases. Therefore, in these exploratory analyses, we compared performance of different approaches, including logistic regression, support vector machine, and ensemble learning methods of classification to predict the future high-risk binge drinking outcome (see Prediction of Future High-Risk Drinking in the Supplement for detailed description).

RESULTS

Effects of Childhood Trauma on Executive Dysfunction

Within the GLMM repeated-measures model for BL through Y4 data, we found that CTQ severity predicted BRIEF global executive composite scores (β = .14, 95% confidence inerval [CI] = .07–.21, p < .001). No significant effect of time nor interaction between CTQ and time (p > .05) was found. Age, sex, ethnicity, years of parental education, familial history of AUD, high-risk drinking, or frequent use of tobacco, marijuana, or other illicit drugs did not predict BRIEF global executive score (see Supplemental Table S1 for sample demographics). The effect of CTQ severity was consistent for both BRIEF indices of ED, that is, the Behavioral Regulation Index (β = .13, 95% CI = .07–.20, p < .001) and the Metacognition Index (β = .13, 95% CI = .06–.20, p < .001) (Figure 1). We also found the effect to be consistent across CTQ subscales for CT forms of abuse (β = .13, 95% CI = .08–.18, p < .001) and neglect (β = .04, 95% CI = 0.01–.07, p = .006).

Figure 1.

Figure 1.

Open in a new tab

Association between indices of the Childhood Trauma Questionnaire and the Behavior Rating Inventory of Executive Function (BRIEF) at baseline (BL) and at 1- (Y1), 2- (Y2), 3- (Y3), and 4-year (Y4) follow-ups.

A second GLMM was carried out including repeated measures from BL up to Y3 that also included the SSAGA variables of MDD and anxiety disorders (these variables were not acquired in Y4). CTQ was found to significantly predict BRIEF ED (β = .11, 95% CI = .04–.17, p = .002). Furthermore, MDD (β = .10, 95% CI = .05–.15, p < .001) and anxiety disorders (β = .11, 95% CI = .02–.20, p = .013) were significantly linked to ED, with higher BRIEF scores in those with mental health diagnoses. When including interaction terms between CTQ and clinical disorders in the model, we further found that the CTQ effect is moderated (p > .05) by both anxiety disorders (β = .23, 95% CI = .04–.42, p = .018) and MDD (β = .14, 95% CI = .01–.26, p = .030), with significant relationships between CTQ and BRIEF scores in participants without anxiety disorders (r = .22, p < .001) and MDD (r = .26, p < .001), but not in those with mental health diagnoses (p > .05).

Mediation Between Childhood Trauma and Executive Dysfunction by Functional Connectivity

We computed CTQ-BRIEF mediation analyses using whole-brain functional connectivity among 236 node ROIs. After correcting for multiple comparisons, 11 mediation hubs emerged, that is, node ROIs that showed a significant sum total of functional connections from that region serving as CTQ-BRIEF mediators; in total, there were 46 unique functional connections serving as significant mediators from these 11 hub nodes. Greater CT severity was associated with weaker functional connectivity of significant mediators, and in turn, weaker connectivity was associated with higher scores on BRIEF indices of ED. As per our hypothesis, mediation hub nodes were located in the bilateral dACC, as well as in the right anterior insula, the right intraparietal sulcus, and bilateral pre- and postcentral gyri. Thus, mediating brain connectivity was primarily distributed within and across hub regions in the salience network, CON, and motor networks (Figure 2; Table 2).

Figure 2.

Figure 2.

Open in a new tab

(A) Functional region-to-region connections that mediate the Childhood Trauma Questionnaire–Behavior Rating Inventory of Executive Function relationship. Hub nodes were defined based on the sum total of significant connections from that region that fulfill criteria for mediation. Hubs with >3, >4, and >5 sum total mediating connections surpassing p < .05, p < .005, p < .0005 bootstrap thresholds for multiple comparisons are shown in cyan, blue, and yellow, respectively. (B) Sum total of independently significant mediating connections from each of the 236 regions of interest. Mediating hubs (i.e., regions of interest for which the total sum of mediating connections surpasses the bootstrap threshold for multiple comparisons) are labeled and correspond to the 11 hubs pictured in (A). Suffixes (1) and (2) refer to distinct mediating hubs in the same cortical region (see Table 2). aI, anterior insula; dACC, dorsal anterior cingulate cortex; IPS, intraparietal sulcus; l, left; post-CG, postcentral gyrus; pre-CG, precentral gyrus; r, right.

Table 2.

Region-to-Region Functional Connectivity Mediators Between Childhood Trauma and Executive Dysfunction

Hub Node; Functional Network Connecting Region of Interest Coordinates z Score z Score a/b
x y z
R Intraparietal Sulcus; SN 55 −45 50 −3.03/−2.72
aa = −3.45
an = −2.22
R middle occipital cortex 40 −72 14 1.97
L middle occipital cortex −28 −79 19 2.22
R superior occipital cortex 15 −87 37 2.18
R middle occipital cortex 29 −77 25 2.25
R Anterior Insula (1); SN 36 22 3 −2.73/−2.08
aa = −2.78
an = −1.84
R inferior frontal gyrus 43 49 22 2.29
L postcentral gyrus (dorsal)a1 −23 −30 72 2.38
L postcentral gyrus (ventral) −49 −11 35 2.10
R postcentral gyrus (ventral)a2 66 28 25 2.48
R Anterior Insula (2); SN 34 16 −8 −3.23/−2.80
aa = −3.90
an = −2.16
L postcentral gyrus (dorsal)a3 −23 −30 72 2.68
L postcentral gyrus (ventral) −49 −11 35 2.30
R postcentral gyrus (ventral) 51 −6 32 2.30
L postcentral gyrus (ventral)a4 -53 −10 24 2.74
R postcentral gyrus (ventral)a5 66 −8 25 2.74
R middle frontal gyrus 31 33 26 2.52
L Dorsal Anterior Cingulate Cortex (1); SN −1 15 44 −3.39/−3.53
aa = −3.57
an = −2.51
R calcarine gyrus 8 −72 11 2.69
L calcarine gyrus −8 −81 7 2.29
R lingual gyrus 20 −66 2 3.08
L lingual gyrus −15 −72 -8 2.84
L calcarine gyrus −18 −68 5 2.52
L cuneus −3 −81 21 2.27
L postcentral gyrus (ventral)a6 −53 −10 24 2.26
R postcentral gyrus (ventral)a7 66 −8 25 2.30
R Dorsal Anterior Cingulate Cortex; SN 5 23 37 −2.65/−3.59
aa = −2.98
an = −1.67
L postcentral gyrus (caudal) −54 −23 43 2.37
R precentral gyrus (dorsal) 29 −17 71 2.15
L postcentral gyrus (dorsal)a8 −23 −30 72 2.65
R postcentral gyrus (ventral)a9 66 −8 25 2.49
R Dorsal Anterior Cingulate Cortex; SN 10 22 27 −1.52/−2.49
aa = −1.84
an = −0.67
R superior temporal gyrus 52 −33 8 2.15
L precentral gyrus (dorsal)a10 −38 −15 69 2.17
R postcentral gyrus (ventral)a11 66 −8 25 3.76
R precentral gyrus (rostral) 42 0 47 2.15
L Dorsal Anterior Cingulate Cortex (2); CON −5 18 34 −0.91/−1.39
aa = −0.78
an = −0.65
R calcarine gyrus 8 −72 11 2.43
R middle occipital cortex 37 −84 13 2.14
L postcentral gyrus (ventral)a12 −53 −10 24 2.10
R postcentral gyrus (ventral)a13 66 -8 25 2.37
R precentral gyrus (rostral) 42 0 47 2.08
L Postcentral Gyrus (1) (Dorsal); MOT −23 −30 72 −2.77/−2.82
aa =−2.84
an = −2.22
R anterior insulaa1ʹ 36 22 3 2.38
R anterior insulaa3ʹ 34 16 −8 2.68
R midcingulate cortexa8ʹ 5 23 37 2.65
L superior temporal pole −51 8 −2 2.12
R anterior insula 36 10 1 2.29
L pallidum −15 4 8 2.28
R angular gyrus 52 −59 36 2.12
L Precentral Gyrus (Dorsal); MOT −38 −15 69 −2.72/−3.21
aa = −1.96
an = −2.40
L anterior cingulate cortex 0 30 27 2.42
R anterior cingulate cortexa10ʹ 10 22 27 2.17
R anterior insula 36 10 1 2.07
L supplementary motor area −10 11 67 2.13
L Postcentral Gyrus (2) (Ventral); MOT −53 −10 24 −3.14/−2.87
aa = −3.31
an = −2.12
R anterior insulaa4ʹ 34 16 −8 2.74
L midcingulate cortexa6ʹ −1 15 44 2.26
R anterior insula 49 8 −1 2.32
L anterior cingulate cortexa12ʹ −5 18 34 2.10
R Postcentral Gyrus (Ventral); MOT 66 −8 25 −4.10/−3.11
aa = −4.78
an = −2.79
L anterior insula −35 20 0 2.35
R anterior insulaa2ʹ 36 22 3 2.48
R anterior insulaa5ʹ 34 16 −8 2.74
L midcingulate cortexa7ʹ −1 15 44 2.30
R midcingulate cortexa9ʹ 5 23 37 2.49
R anterior cingulate cortexa11ʹ 10 22 27 3.76
L anterior insula −34 3 4 2.27
L anterior cingulate cortexa13ʹ −5 18 34 2.37
R anterior insula 36 10 1 2.43
Open in a new tab

Coordinates are in the Montreal Neurological Institute stereotaxic space. z score a/b = z score of paths a/b for median functional connectivity of hub region to its relevant connecting regions of interest; a = direction of the path childhood trauma → median functional connectivity of hub node; aa = path of abuse; an = path of neglect; b = direction of the path median functional connectivity of hub node → executive dysfunction. (1) and (2) indicate distinct mediating hubs in the same cortical region. a# and a#ʹ indicate bidirectional functional connectivity mediators (total of 13). z score of each mediation informs the relative quality of the observed solution vs. 1000 bootstrap permuted data solutions (61).

CON, cingulo-opercular network; MOT, motor network; R/L, right/left hemisphere; SN, salience network.

Predicting High-Risk Drinking in Future Years

A comparison of cumulative recall rates of high-risk and low-risk drinkers, that is, sensitivity and specificity of each predictive model are presented in Figure 3B. The random forest algorithm based on demographic, behavioral, and functional connectivity data showed the best classification results, with combined highest sensitivity and specificity of all models. This random forest model had an overall accuracy of 0.80; high-risk drinkers were predicted with a precision of 0.79, recall (or sensitivity) of 0.83, and F1 score of 0.81, and low-risk drinkers were predicted with a cumulative precision of 0.83, recall (or specificity) of 0.78, and F1 score of 0.80. The predictive odds ratio of this model was 17.31, suggesting that it has diagnostic value.

Figure 3.

Figure 3.

Open in a new tab

(A) Predictive model of future high-risk drinking in years 1–4 using baseline data. (B) Cumulative sensitivity and specificity of predicting high-risk drinkers in follow-up years 1–4 based on demographic, behavioral, and functional connectivity data at study baseline, with synthetic minority oversampling technique applied. Results of logistic regression (LR), support-vector machine (SVM), gradient boost (GB), AdaBoost (AB), and random forest (RF) algorithms are compared. For LR, results without any synthetic minority class over-sampling are also shown (LR SM–) to exemplify the poor sensitivity obtained without applying this technique. Model predictions are displayed against the 95% confidence interval = 44%–56% (blue region) for 50% chance level prediction (black line). (C) Relative feature importance for prediction of high-risk versus low-risk drinkers in follow-up years 1–4 using the random forest ensemble learning algorithm, which was most predictive (i.e., with combined highest sensitivity and specificity) among the tested models (B). Means and SDs of feature importance are shown across five folds of cross-validation. Suffixes (1) and (2) refer to distinct hubs in the same cortical region that mediate the childhood trauma → executive dysfunction relationship (see Table 2). aI, anterior insula; AUD, alcohol use disorder; Bl, baseline; dACC, dorsal anterior cingulate cortex; IPS, intraparietal sulcus; l, left; MDD, major depressive disorder; post-CG, postcentral gyrus; pre-CG, precentral gyrus; r, right.

Relative feature importance scores extracted for the random forest ensemble learning model (Figure 3C) indicated that the factor of age was of highest classification relevance: older adolescents were more likely to develop a pattern of high-risk drinking. Also, functional connectivity from the left postcentral gyrus and the left dACC (part of the CON) hub regions ranked in the top 3 most important features (specifically, these were the median functional connectivity of the hubs to other connecting regions that mediate the relationship between CT and ED) (see Table 2): lower absolute functional connectivity strengths were observed in high-risk versus low-risk drinkers (postcentral gyrus: 0.044 ± 0.015 vs. 0.049 ± 0.012; dACC: 0.000 ± 0.011 vs. 0.016 ± 0.008) at follow-up years Y1 to Y4. All other mediating hub region functional connections were within the top two-thirds highest feature ranks (up to rank 15), including mediating connection strengths from the right anterior insula, bilateral postcentral gyri, left precentral gyrus, and bilateral dACC region. CTQ scores and the BRIEF executive function indices ranked as second and third most relevant behavioral predictors (at ranks 10 and 12, respectively): higher CT severity and higher levels of ED predicted high-risk drinkers. Feature importance of BL high-risk drinking (see “frequent binge BL” factor in Figure 3C) was equivalent to that of ED. Finally, we also found that all models based on behavioral data alone, which excluded functional connectivity data, had much poorer sensitivity than when functional connectivity was included (Supplemental Figure S1).

DISCUSSION

Using data from the large NCANDA longitudinal neuroimaging study, we showed significant associations between CT exposure and ED during adolescence. As a novel contribution, we demonstrated the brain region hubs whose functional connections mediate the relationship between CT and poor executive function. We further used these neurobiological data to predict future alcohol use behavior, particularly high-risk drinking in these adolescents.

We found that exposure to CT was associated with lower executive function across all 5 assessment years. There was neither an effect of time nor an interaction of CT with time. Hence, this trauma-related ED seems to be consistently present, predating longitudinal data collection in adolescence. Notably, while the relationship between frequent alcohol/substance use and executive function could be bidirectional, here we did not find that high-risk drinking status at BL or frequent use of tobacco, marijuana, or other illicit drugs had any significant effect on ED. However, when additionally modeling prevalence of psychiatric disorders at BL and the first 3 years of follow-up assessments, we found that MDD and anxiety disorders significantly affect ED, which is in line with the notion of a common transdiagnostic factor of impaired executive function in mental illness (14), and that associations between CT and executive function are only relevant in adolescents without mental health diagnoses. These findings highlight a distinct and moderating effect of mental health status on trauma-related ED.

To test whether functional brain connectivity mediates the relationship between CT and ED, we examined brain-wide functional connections. We identified 11 hub seed regions and their connectivity strengths to distributed cortical regions that qualified as neurobiological mediators at BL, with trauma severity corresponding to weaker functional connectivity, and lower connectivity values being associated with higher impairments in executive function. These contributing region-to-region functional connections were found within and between known functionally organized brain networks. Detected hub nodes were located within the CON (i.e., the left dACC), the salience network (53) (i.e., the right anterior insula, right intraparietal sulcus, and bilateral dACC), and the motor network (i.e., bilateral pre- and post-central gyri). Emergence of the dACC hubs as mediators confirmed our hypotheses of this region being sensitive to trauma and critical to development of executive function. Prior evidence has also linked cognitive control development in typical adolescence to functional connectivity of the dACC as a key hub that drives segregation and maturation of the CON (25,26). Furthermore, we recently found that the dACC’s maturing connectivity within the CON is hampered by CT in an international sample, which led to the hypothesis that functional connections from dACC will mediate the relationship between CT and executive function (27). Notably, we found that neurobiological mediators between CT severity and ED were not limited to cognitive control networks. Alterations in connectivity strength within and between “low-level” network regions in the pre- and post-central gyri and in the visual cortex (Table 2) were found to affect high-level executive function. This finding supports the important role of sensorimotor networks in decision making and action regulation, complementing the role of the dACC in performance/feedback monitoring. Together these distributed brain networks integrate bottom-up attention with top-down control and generate appropriate behavioral responses to salient stimuli, thereby contributing to behavioral regulation and sound executive function (51–57).

We found that severity of CT, associated functional connectivity, and executive function can be used to predict high-risk drinking in adolescents. Particularly, functional connectivity of the left dACC seed region within the CON was found to be within the top 3 features important for classifying high-risk drinkers in follow-up years Y1 to Y4. This finding demonstrates that high-risk drinking in adolescents can be inferred from trauma-related impairments in underlying neurocognitive functional networks. Notably, these functional connectivity features inform prediction beyond effects of drinking status at an earlier age, i.e., high-risk drinking at BL. Our study thus extends prior work that has been limited to probing subjective self-regulation behaviors as underlying mechanisms (16). By including brain function, our findings further complement prior longitudinal studies in adolescent samples that have not been able to prospectively resolve the relationship between CT and excessive drinking (58,59).

As a limitation, our research is not based on an enriched CT sample, with reports confined to low to moderate levels of trauma. Furthermore, the CTQ is a retrospective measure regarding childhood experiences before the age of 18 years. While this allows assessment of long-term effects and developmental trajectories after the time of assessment, the cause and effect relationship with executive function deficits at study BL remains unclear; it is possible that children and adolescents with ED are more prone to experience trauma or to report trauma experience. That we found consistent effects of abuse and neglect categories of the CTQ may suggest that multiple measures that are differentially sensitive to these categories need to be used to better tease apart their effects (29–32). A further limitation is the low prevalence of AUDs in the current sample (only up to 2.9%), demanding caution in a generalization of the results to the development of AUDs, although notably, we did find a significant overlap between high-risk drinkers and adolescents that fulfill clinical criteria of alcohol abuse. Regarding assessments of executive function, this study is limited by subjective self-report instead of objective performance measures. Mapping onto different mental constructs, rating measures assess rational control toward goal achievement in everyday environments but not objective performance, while objective performance measures assess efficiency of cognitive control processes but not degree of rational goal pursuit. In the context of substance use, the BRIEF thus provides ecologically valid dimensions of behavioral executive function in complex situations, which have no parallel performance-based measures (41). In addition, self-reports are complemented by relevant functional connectivity measures from hub regions that are known to correspond to executive functions (22,26).

Novel to this study, we found that prediction of high-risk drinking is critically informed by the trauma-affected brain networks that mediate poor executive function; this result underscores the value of functional brain connectivity in precision psychiatry. Of interest, even though our sample reports low to moderate CT, our neural findings are consistent with recent findings of anatomical differences between severely maltreated youths with posttraumatic stress disorder (60). Taken together, our findings highlight the important role of distributed networks relevant for cognitive and behavioral control, as well as sensorimotor integration in the prediction of high-risk alcohol use behavior.

In conclusion, our results provide evidence for the relationship between CT, neurobiological functional network alterations, and executive function, which may be used to predict high-risk drinking in adolescence (Figure 3). The present study demonstrates that functional brain connectivity from a distributed set of hub regions may serve as functional biomarkers mediating the relationship between early-life adversity and executive behaviors. Future research is needed to address remaining gaps in understanding, such as what neurobiological responses to early life stress endure longitudinally, and hence either contribute to a susceptibility to adverse life outcomes in terms of psychopathology and alcohol/substance use behaviors, or in contrast, facilitate resilience. An identification of these functional markers can further serve as a target to promote personalized network-targeted interventions in this at-risk population.

Supplementary Material

Supplementary Material

ACKNOWLEDGMENTS

Project funding was provided by the University of California San Diego School of Medicine (to JM) and the German National Academy of Sciences Leopoldina Fellowship Grant No. LPDS 2017–09 (to SS). Reported data were obtained from the multisite National Consortium on Alcohol and Neuro-development in Adolescence (NCANDA) study, which is supported by the U.S. National Institute on Alcohol Abuse and Alcoholism with cofunding from the National Institute on Drug Abuse, the National Institute of Mental Health, and the National Institute of Child Health and Human Development (Grant Nos. AA021697 [to Adolf Pfefferbaum and Kilian M. Pohl], AA021695 [to Susan F. Tapert and Sandra A. Brown], AA021692 [to Susan F. Tapert], AA021696 [to Fiona C. Baker and Ian M. Colrain], AA021681 [to Michael D. de Bellis], AA021690 [to Duncan B. Clark], and AA021691 [to Bonnie J. Nagel]).

Footnotes

DISCLOSURES

The authors report no biomedical financial interests or potential conflicts of interest.

ARTICLE INFORMATION

Supplementary material cited in this article is available online at https://doi.org/10.1016/j.bpsc.2020.01.011.

Contributor Information

Sarita Silveira, Department of Psychiatry, Neural Engineering and Translation Labs, University of California San Diego, La Jolla, California.

Rutvik Shah, Department of Psychiatry, Neural Engineering and Translation Labs, University of California San Diego, La Jolla, California.

Kate B. Nooner, Department of Psychology, University of North Carolina, Wilmington, North Carolina

Bonnie J. Nagel, Departments of Psychiatry and Behavioral Neuroscience, Oregon Health and Science University, Portland, Oregon

Susan F. Tapert, Department of Psychiatry, University of California San Diego, La Jolla, California

Michael D. de Bellis, Department of Psychiatry and Behavioral Sciences, Duke University Medical Center, Durham, North Carolina

Jyoti Mishra, Department of Psychiatry, Neural Engineering and Translation Labs, University of California San Diego, La Jolla, California.

REFERENCES

  • 1.Felitti VJ, Anda RF, Nordenberg D, Williamson DF, Spitz AM, Edwards V, et al. (1998): Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The adverse childhoos experiences (ACE) study. Am J Prev Med 14:245–258. [DOI] [PubMed] [Google Scholar]
  • 2.Kessler RC, McLaughlin KA, Green JG, Gruber MJ, Sampson NA, Zaslavsky AM, et al. (2010): Childhood adversities and adult psychopathology in the WHO World Mental Health Surveys. Br J Psychiatry 197:378–385. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Dube SR, Anda RF, Felitti VJ, Edwards VJ, Croft JB (2002): Adverse childhood experiences and personal alcohol abuse as an adult. Addict Behav 27:713–725. [DOI] [PubMed] [Google Scholar]
  • 4.Fenton MC, Geier T, Keyes K, Skodol AE, Grant BF, Hasin DS (2013): Combined role of childhood maltreatment, family history, and gender in the risk for alcohol dependence. Psychol Med 43:1045–1057. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Fuller-Thomson E, Roane JL, Brennenstuhl S (2016): Three types of adverse childhood experiences, and alcohol and drug dependence among adults: An investigation using population-based data. Subst Use Misuse 51:1451–1461. [DOI] [PubMed] [Google Scholar]
  • 6.Oberleitner LMS, Smith PH, Weinberger AH, Mazure CM, McKee SA (2015): Impact of exposure to childhood maltreatment on transitions to alcohol dependence in women and men. Child Maltreat 20:301–308. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Rothman EF, Edwards EM, Heeren T, Hingson RW (2008): Adverse childhood experiences predict earlier age of drinking onset: Results from a representative US sample of current or former drinkers. Pediatrics 122:e298–e304. [DOI] [PubMed] [Google Scholar]
  • 8.Just-Østergaard E, Mortensen EL, Flensborg-Madsen T (2017): Major life events and risk of slcohol use disorders: a prospective cohort study. Addiction 113:25–30. [DOI] [PubMed] [Google Scholar]
  • 9.McEwen BS, Gray JD, Nasca C (2015): Redefining neuroendocrinology: Stress, sex and cognitive and emotional regulation. J Endocrinol 226:T67–T83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Teicher MH, Samson JA, Anderson CM, Ohashi K (2016): The effects of childhood maltreatment on brain structure, function and connectivity. Nat Rev Neurosci 17:652–666. [DOI] [PubMed] [Google Scholar]
  • 11.Gould F, Clarke J, Heim C, Harvey PD, Majer M, Nemeroff C (2012): The effects of child abuse and neglect on cognitive functioning in adulthood. J Psychiatr Res 46:500–506. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Aas M, Dazzan P, Mondelli V, Melle I, Murray RM, Pariante CM (2014): A systematic review of cognitive function in first-episode psychosis, including a discussion on childhood trauma, stress, and inflammation. Front Psychiatry 4:1–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Millan MJ, Agid Y, Brüne M, Bullmore ET, Carter CS, Clayton NS, et al. (2012): Cognitive dysfunction in psychiatric disorders: characteristics, causes and the quest for improved therapy. Nat Rev Drug Discov 11:141–168. [DOI] [PubMed] [Google Scholar]
  • 14.McTeague LM, Goodkind MS, Etkin A (2016): Transdiagnostic impairment of cognitive control in mental illness. J Psychiatr Res 83:37–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Kwako LE, Momenan R, Litten RZ, Koob GF, Goldman D (2017): Addictions neuroclinical assessment: A framework for addictive disorders. Biol Psychiatry 80:179–189. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Shin SH, Jiskrova GK, Wills TA (2019): Childhood maltreatment and alcohol use in young adulthood: the role of self-regulation processes. Addict Behav 90:241–249. [DOI] [PubMed] [Google Scholar]
  • 17.Cisler JM, James GA, Tripathi S, Mletzko T, Heim C, Hu XP, et al. (2013): Differential functional connectivity within an emotion regulation neural network among individuals resilient and susceptible to the depressogenic effects of early life stress. Psychol Med 43:507–518. [DOI] [PubMed] [Google Scholar]
  • 18.Elton A, Tripathi SP, Mletzko T, Young J, Cisler JM, James A, et al. (2015): Childhood maltreatment is associated with a sex-dependent functional reorganization of a brain inhibitory control network. Hum Brain Mapp 35:1654–1667. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Wang L, Dai Z, Peng H, Tan L, Ding Y, He Z, et al. (2014): Overlapping and segregated resting-state functional connectivity in patients with major depressive disorder with and without childhood neglect. Hum Brain Mapp 35:1154–1166. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Nooner KB, Mennes M, Brown S, Castellanos FX, Leventhal B, Milham MP, et al. (2013): Relationship of trauma symptoms to amygdala-based functional brain changes in adolescents. J Trauma Stress 26:784–787. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Uddin LQ, Supekar K, Menon V (2010): Typical and atypical development of functional human brain networks: insights from resting-state fMRI. Front Syst Neurosci 4:21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Ordaz SJ, Foran W, Velanova K, Luna B (2013): Longitudinal growth curves of brain function underlying inhibitory control through adolescence. J Neurosci 33:18109–18124. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Luna B, Marek S, Larsen B, Tervo-Clemmens B, Chahal R (2015): An integrative model of the maturation of cognitive control. Annu Rev Neurosci 38:151–170. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.van den Heuvel MP, Sporns O (2013): Network hubs in the human brain. Trends Cogn Sci 17:683–696. [DOI] [PubMed] [Google Scholar]
  • 25.Dosenbach NUF, Visscher KM, Palmer ED, Miezin FM, Wenger KK, Kang HC, et al. (2006): A core system for the implementation of task sets. Neuron 50:799–812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Fair DA, Dosenbach NUF, Church JA, Cohen AL, Brahmbhatt S, Miezin FM, et al. (2007): Development of distinct control networks through segregation and integration. Proc Natl Acad Sci U S A 104:13507–13512. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Mishra J, Sagar R, Modi K, Alam S, Gazzaley A (2017): T13. Plasticity of cognitive control in adolescents with adverse childhood experiences. American College of Neuropsychopharmacology 56th Annual Meeting: Poster Session II, December 5, 2017. Neuro-psychopharmacology 42:S294–S475. [Google Scholar]
  • 28.Cassiers LLM, Sabbe BGC, Schmaal L, Veltman DJ, Penninx BWJH, van den Eede F (2018): Structural and functional brain abnormalities associated with exposure to different childhood trauma subtypes: A systematic review of neuroimaging findings. Front Psychiatry 9:329. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Sumner JA, Colich NL, Uddin M, Armstrong D, McLaughlin KA (2019): Early experiences of threat, but not deprivation, are associated with accelerated biological aging in children and adolescents. Biol Psychiatry 85:268–278. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Lambert HK, King KM, Monahan KC, McLaughlin KA (2017): Differential associations of threat and deprivation with emotion regulation and cognitive control in adolescence. Dev Psychopathol 29:929–940. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Sheridan MA, Peverill M, Finn AS, McLaughlin KA (2017): Dimensions of childhood adversity have distinct associations with neural systems underlying executive functioning. Dev Psychopathol 29:1777–1794. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Gur RE, Moore TM, Rosen AFG, Barzilay R, Roalf DR, Calkins ME, et al. (2019): Burden of environmental adversity associated with psychopathology, maturation, and brain behavior parameters in youths. JAMA Psychiatry 76:966–975. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Brown SA, Brumback TY, Tomlinson K, Cummins K, Thompson WK, Nagel BJ, et al. (2015): The National Consortium on Alcohol and Neuro-Development in Adolescence (NCANDA): A multisite study of adolescent development and substance use. J Stud Alcohol Drugs 213:895–908. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.von Rueden U, Gosch A, Rajmil L, Bisegger C, Ravens-Sieberer U (2006): Socioeconomic determinants of health related quality of life in childhood and adolescence: results from a European study. J Epidemiol Community Heal 60:130–135. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Sullivan EV, Brumback T, Tapert SF, Fama R, Prouty D, Brown SA, et al. (2016): Cognitive, emotion control, and motor performance of adolescents in the NCANDA study: Contributions from alcohol consumption, age, sex, ethnicity, and family history of addiction. Neuro-psychology 30:449–473. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Stoltenberg SF, Mudd S, Blow FC, Hill EM (1998): Evaluating measures of family history of alcoholism: density versus dichotomy. Addiction 93:1511–1520. [DOI] [PubMed] [Google Scholar]
  • 37.Bernstein DP, Stein JA, Newcomb MD, Walker E, Pogge D, Ahluvalia T, et al. (2003): Development and validation of a brief screening version of the Childhood Trauma Questionnaire. Child Abus Negl 27:169–190. [DOI] [PubMed] [Google Scholar]
  • 38.Scher CD, Stein MB, Asmundson GJG, McCreary DR, Forde DR (2001): The Childhood Trauma Questionnaire in a community sample : Psychometric properties and normative data. J Trauma Stress 14:843–857. [DOI] [PubMed] [Google Scholar]
  • 39.Guy SC, Isquith PK, Gioia GA (2004): BRIEF-SR: Behavior rating inventory of executive function—Self-report version. Psychological Assessment Resources Available at: https://www.parinc.com/products/pkey/27.
  • 40.Clark DB, Chung T, Martin CS, Hasler BP, Fitzgerald DH, Luna B, et al. (2017): Adolescent executive dysfunction in daily life: relationships to risks, brain structure and substance use. Front Behav Neurosci 11:223. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Toplak ME, West RF, Stanovich KE (2013): Practitioner review: Do performance-based measures and ratings of executive function assess the same construct? J Child Psychol Psychiatry 54:131–143. [DOI] [PubMed] [Google Scholar]
  • 42.Templeton GF (2011): A two-step approach for transforming continuous variables to normal: Implications and recommendations for IS research. Commun Assoc Inf Syst 28:41–58. [Google Scholar]
  • 43.American Psychiatric Association (2013): Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition. Arlington, VA: American Psychiatric Association. [Google Scholar]
  • 44.Brown SA, Myers MG, Lippke L, Tapert SF, Stewart DG, Vik PW (1998): Psychometric evaluation of the Customary Drinking and Drug Use Record (CDDR): A measure of adolescent alcohol and drug involvement. J Stud Alcohol 59:427–438. [DOI] [PubMed] [Google Scholar]
  • 45.Grant BF, Chou P, Saha TD, Pickering RP, Kerridge BT, June R, et al. (2017): Prevalence of 12-month alcohol use, high-risk drinking, and DSM-IV alcohol use disorder in the United States 2001–2002 to 2012–2013. JAMA Psychiatry 74:911–923. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Lydon DM, Wilson SJ, Child A, Geier CF (2014): Adolescent brain maturation and smoking: What we know and where we’re headed. Neurosci Biobehav Rev 45:323–342. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Bechara A (2005): Decision making, impulse control and loss of willpower to resist drugs: a neurocognitive perspective. Nat Neurosci 8:1458–1463. [DOI] [PubMed] [Google Scholar]
  • 48.Whitfield-Gabrieli S, Nieto-Castanon A (2012): CONN: A functional connectivity toolbox for correlated and anticorrelated brain networks. Brain Connect 2:125–141. [DOI] [PubMed] [Google Scholar]
  • 49.Wager TD, Davidson ML, Hughes BL, Lindquist MA, Ochsner KN (2008): Prefrontal-subcortical pathways mediating successful emotion regulation. Neuron 59:1037–1050. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Baron RM, Kenny DA (1986): The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Pers Soc Psychol 51:1173–1182. [DOI] [PubMed] [Google Scholar]
  • 51.Efron B, Tibshirani RJ (1994): An Introduction to the Bootstrap CRC Press. [Google Scholar]
  • 52.Shrout PE, Bolger N (2002): Mediation in experimental and nonexperimental studies: New procedures and recommendations. Psychol Methods 7:422–445. [PubMed] [Google Scholar]
  • 53.Menon V, Uddin LQ (2010): Saliency, switching, attention and control: a network model of insula function. Brain Struct Funct 214:655–667. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Cicchetti D, Tucker D (1994): Development and self-regulatory structures of the mind. Dev Psychopathol 6:533–549. [Google Scholar]
  • 55.Luu P, Tucker DM (2001): Regulating action: alternating activation of midline frontal and motor cortical networks. Clin Neurophysiol 112:1295–1306. [DOI] [PubMed] [Google Scholar]
  • 56.Depue BE, Orr JM, Smolker HR, Naaz F, Banich MT (2016): The organization of right prefrontal networks reveals common mechanisms of inhibitory regulation across cognitive, emotional, and motor processes. Cereb Cortex 26:1634–1646. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Bressler SL, Menon V (2010): Large-scale brain networks in cognition: emerging methods and principles. Trends Cogn Sci 14:277–290. [DOI] [PubMed] [Google Scholar]
  • 58.Patten SB, Wilkes TCR, Williams JVA, Lavorato DH, El-Guebaly N, Schopflocher D, et al. (2015): Retrospective and prospectively assessed childhood adversity in association with major depression, alcohol consumption and painful conditions. Epidemiol Psychiatr Sci 24:158–165. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Scott KM, McLaughlin KA, Smith DAR, Ellis PM (2012): Childhood maltreatment and DSM-IV adult mental disorders: Comparison of prospective and retrospective findings. Br J Psychiatry 200:469–475. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.de Bellis MD, Hooper SR, Chen SD, Provenzale JM, Boyd BD, Glessner CE, et al. (2015): Posterior structural brain volumes differ in maltreated youth with and without chronic posttraumatic stress disorder. Dev Psychopathol 27:1555–1576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Cheung GW, Lau RS (2008): Mediation and suppression effects of latent variables. Organ Res Methods 11:296–325. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material
NIHMS1723274-supplement-Supplementary_Material.pdf (288.4KB, pdf)

RESOURCES