Greater affective network maturity is associated with better clinical outcomes in women with early sexual trauma

Abstract

Early-life adversity accelerates the maturation of affect-related circuitry, which might be a short-term adaptation with long-term tradeoffs. Sexual trauma is associated with a particularly strong impact on pubertal development and mental health outcomes. Our objective was to test the relations between trauma type, affective network maturity, and mental health outcomes in young women with trauma history. Trauma-exposed women aged 18–29 completed a clinical interview (n = 35) and an fMRI scan (n = 28). We used a public data set to train a machine learning algorithm to predict age from resting-state affective network connectivity and calculated network maturity as the difference between predicted and true age. We also performed principal component analysis on mental health outcomes and retained two components: clinical and state psychological outcomes. Compared to nonsexual trauma (n = 17), sexual trauma (n = 11) was associated with greater affective network maturity. In addition, for sexual trauma only, greater affective network maturity was associated with better clinical but not state psychological outcomes. These results suggest that sexual trauma during development might uniquely alter the maturational trajectory of affect-related circuitry, with distinct mental health consequences in emerging adulthood. Whereas delayed affective network maturation is associated with adverse clinical outcomes, accelerated affective network maturation might confer resilience in survivors.

Introduction

Traumatic stress during sensitive periods of development is a risk factor for mental health problems in adulthood (e.g., Green et al., 2010; Hoppen & Chalder, 2018). Compared to other types of adversity, early-life sexual trauma is associated with higher risk and symptom severity across a range of diagnostic domains, including posttraumatic stress disorder (PTSD), depression, anxiety, and substance use (e.g., Dworkin et al., 2017; Molnar et al., 2001; O’Brien et al., 2016). Disrupted brain development is a potential mechanism that might account for the relation between early-life trauma and adverse mental health outcomes (e.g., Andersen & Teicher, 2008; Bick & Nelson, 2016). Further, trauma type, and in particular sexual trauma, might moderate this relation by exerting circuit-specific effects on neurodevelopment (e.g., Cohodes et al., 2020).

Growing evidence suggests that early-life adversity accelerates development and sexual maturation, particularly in girls, as indexed by earlier pubertal onset (Gur et al., 2019), younger age at menarche (Colich et al., 2019), and accelerated cellular aging (e.g., Sumner et al., 2019). The developmental effects of early-life adversity depend on the type of adversity, such that threat-related adversity (e.g., sexual abuse, violence) is associated with accelerated pubertal development and cellular aging, whereas deprivation-related adversity (e.g., neglect) is not (Colich et al., 2020). Compared to other trauma types, early-life sexual trauma appears to have an especially strong association with abnormally advanced pubertal development (e.g., Mendle et al., 2014, 2016; Noll et al., 2017). Accelerated maturation in turn predicts a range of mental health problems in adolescence, such as internalizing symptoms (Mendle et al., 2014), and mood, anxiety, and substance use disorders (e.g., Platt et al., 2017). Moreover, accelerated maturation, as indexed by earlier menarche and advanced cellular aging, mediates the relation between early-life adversity and later psychiatric symptoms (Colich et al., 2019; Sumner et al., 2019).

In addition to advancing pubertal development, early-life trauma might alter the developmental trajectory of brain maturation. For example, adolescent girls who have been exposed to childhood trauma (Gur et al., 2019) or physical neglect (Keding et al., 2021) show greater structural maturity of the brain as a whole compared to non-trauma-exposed peers. Furthermore, the relation between early-life adversity and greater structural brain maturity is particularly strong among adolescents exposed to threat-related adversity, such as sexual abuse, home violence, or emotional abuse (Drobinin et al., 2022). However, emerging longitudinal evidence points to complex and nonlinear effects of early-life adversity on neurodevelopment, such that initially accelerated brain maturation may be followed by delayed brain maturation later (Rakesh et al., 2021).

Although early-life adversity has been associated with accelerated global maturation of the brain as a whole, there is also growing evidence of circuit- and network-specific effects (e.g., Colich et al., 2020; Keding et al., 2021). Frontolimbic circuitry centered on the amygdala, hippocampus, and medial prefrontal cortex (mPFC) might be particularly susceptible to the neurodevelopmental effects of early-life adversity, given its role in affective processing, threat detection, and stress response regulation (VanTieghem & Tottenham, 2017). For example, early maternal separation and low maternal care accelerate the maturation of the hippocampus (e.g., Bath et al., 2016), amygdala (Ono et al., 2008), and frontolimbic circuits (Callaghan et al., 2014; Manzano Nieves et al., 2020) in non-human animals. Similarly, maternally deprived children have been shown to display a more mature pattern of amygdala-mPFC coupling during emotion processing (Gee et al., 2013) and at rest (Herzberg et al., 2021) compared to peer controls, although the reliability of this association and the use of task-based amygdala-mPFC coupling as an index of brain maturity has been called into question (Bloom et al., 2022; Colich et al., 2020) Childhood adversity also has been associated with accelerated structural maturation of the frontolimbic circuitry in adolescence, as indexed by lesser gray matter volumes, thinner cortex, greater white matter integrity, and greater longitudinal reductions in gray matter volume spanning prefrontal, insular, medial temporal, and subcortical limbic regions (e.g., Busso et al., 2017; Gur et al., 2019; Kelly et al., 2013; Tyborowska et al., 2018). In addition, early-life adversity has been associated with accelerated functional maturation of frontolimbic circuitry in young adolescents (Brieant et al., 2021; Thijssen et al., 2022). Furthermore, accelerated frontolimbic maturation has been specifically linked to threat-related adversity (Colich et al., 2020). Despite much evidence in support of accelerated maturation, some studies have shown delayed maturation of affective circuitry in trauma-exposed youths (e.g., Keding et al., 2021; Whittle et al., 2013). This apparent discrepancy in the literature might point to nonlinear effects of early-life adversity on the neurodevelopmental trajectory. For example, a recent longitudinal study showed that neighborhood disadvantage is associated with accelerated brain maturation in early adolescence followed by delayed maturation in late adolescence in at-risk youths (e.g., Rakesh et al., 2021).

Differences in the neurodevelopmental effects of early-life trauma (i.e., acceleration or delay) could also underlie different functional outcomes and thus reflect vulnerability or resilience. Earlier frontolimbic maturation has been proposed to have adaptative value in harsh early-life environments by promoting earlier emergence of affect-related functions and stress-regulatory behaviors (Callaghan & Tottenham, 2016). For example, accelerated limbic maturation is associated with earlier development of stress reactivity and fear learning in non-human animals (e.g., Bath et al., 2016; Callaghan et al., 2014). Similarly, a more mature pattern of amygdala-mPFC coupling during emotion processing has been associated with less separation anxiety in maternally deprived children (Gee et al., 2013), although recent research has not replicated this association in more normative samples (e.g., Bloom et al., 2022). Additionally, greater functional maturity of resting corticolimbic connectivity in adolescents has been shown to mediate the relation between early-life adversity and less internalizing symptoms two years later (Brieant et al., 2021). Consistent with the hypothesized adaptive value of accelerated maturation, delayed frontolimbic maturation has been associated with psychiatric symptoms (e.g., Albaugh et al., 2017; Ducharme et al., 2014; Vanes et al., 2020), particularly in trauma-exposed youths (e.g., Keding et al., 2021; Whittle et al., 2013). Furthermore, slower development of frontal and limbic structures has been shown to mediate the relation between early-life adversity and poor outcomes in adolescents (e.g., Barch et al., 2022; Hanson et al., 2015; Whittle et al., 2016).

Accelerated pubertal development and frontolimbic maturation might constitute a short-term adaptation with long-term mental health consequences (e.g., Belsky and Shalev, 2016; Callaghan & Tottenham, 2016). For example, earlier pubertal onset has been associated with psychiatric symptoms in adolescence (e.g., Colich et al., 2019; Mendle et al., 2014; Platt et al., 2017). However, the precise relationship between altered development and short-term vs. long-term outcomes remains unclear given prior reports of both positive and negative mental health outcomes of accelerated maturation. One possibility is that negative outcomes of early-life trauma might be mediated by accelerated sexual maturation and biological aging (e.g., Colich et al., 2019; Sumner et al., 2019), whereas positive outcomes might be mediated by selectively accelerated maturation of neural circuitry involved in affective processing and stress regulation (e.g., Brieant et al., 2021). An alternative explanation supported by emerging evidence suggests that the mental health tradeoffs of accelerated brain maturation in early adolescence might result from a subsequent deceleration of brain maturation in late adolescence (e.g., Rakesh et al., 2021). For example, a prospective study of youths with a family history of mood disorders showed that the risk of developing a mood disorder was associated with delayed brain maturation during late adolescence and emerging adulthood (de Nooij et al., 2020).

Affect-related frontolimbic circuits follow protracted maturational trajectories well into adulthood, as indexed by age-related changes in synaptic density (Petanjek et al., 2011), subcortical gray matter volume (Narvacan et al., 2017), white matter structure (Lebel et al., 2012), intracortical myelination (Grydeland et al., 2013), and functional connectivity (Gabard-Durnam et al., 2014). Due to this prolonged immaturity and a high concentration of receptors for various stress mediators (e.g., Joëls & Baram, 2009), frontolimbic circuitry might be particularly sensitive to traumatic stress not only in childhood and adolescence but also in early adulthood. However, as research on posttrauma maturation of affect-related circuitry beyond childhood and adolescence is still lacking, the relation between altered frontolimbic development and later mental health outcomes remains unclear.

The importance of clarifying the relation between trauma-related alterations in frontolimbic maturity and mental health outcomes in emerging adults is threefold. First, studies that focus on trauma-exposed children and adolescents miss the delayed effects of early-life trauma on mental health, including adult-onset psychopathology (e.g., Green et al., 2010). Conversely, assessment of mental health outcomes in young adulthood permits stronger conclusions about vulnerability vs. resilience to adverse outcomes of early-life trauma. Second, because frontolimbic circuits remain malleable into early adulthood (e.g., Petanjek et al., 2011), temporary deviations from normative maturity in childhood or adolescence that eventually normalize are unlikely to account for adverse mental health outcomes in adulthood. Finally, emerging research suggests that the neurodevelopmental alterations of early-life trauma and their functional significance are age-dependent (Rakesh et al., 2021), such that potentially beneficial early acceleration might be followed by potentially detrimental late deceleration of maturation. Thus, it is important to characterize the trauma-related effects on brain maturation in emerging adulthood separately from earlier stages of development.

Trauma exposure during sensitive periods of frontolimbic development might interact with the maturational trajectories of affect-related circuits to produce distinct mental health outcomes later in life (e.g., Gee & Casey, 2015). Of the various types of childhood adversity, sexual trauma has been robustly associated with stronger impact on pubertal maturation (e.g., Noll et al., 2017) and poorer mental health outcomes (e.g., Dworkin et al., 2017), however the effect of sexual trauma on the maturation of affect-related circuitry remains unclear. Early-life sexual trauma is thought to be distinct from other trauma types in that it represents a particularly damaging combination of traumatic sexualization, betrayal, powerlessness, and stigmatization (Finkelhor & Browne, 1985), which together exert profound effects on survivors’ affective function and emotion regulation abilities. Emotion dysregulation in turn mediates the relation between sexual trauma and psychopathology (e.g., Raudales et al., 2019). Given the particular vulnerability associated with early-life sexual trauma, the impact of trauma on the maturation of affect-related circuitry might be especially pronounced in survivors of sexual trauma compared to other types of adversity and might account for vulnerability or resilience to adverse mental health outcomes.

To investigate this question, we used a machine-learning-based approach to measuring functional maturity of affect-related brain circuits. Our first objective was to test differences in the functional maturity of affective networks in young women who had been exposed to either sexual or nonsexual trauma prior to study enrollment (i.e., during childhood, adolescence, or emerging adulthood), which we refer to as early-life trauma. Given the absence of a universally accepted definition of “early life” and the growing evidence of protracted brain maturation well into young adulthood (e.g., Grydeland et al., 2013; Lebel et al., 2012), here we define early-life trauma as traumatic stress during the first three decades of life when the brain continues to undergo structural and functional maturation. Our second objective was to quantify the relation between affective network maturity and trauma-related mental health outcomes in emerging adulthood. We used a public neuroimaging data set of healthy no-trauma controls ages 6–45 to train a machine learning algorithm to predict age based on resting-state connectivity of affect-related networks (orbito-affective and cingulo-opercular). We then used the trained algorithm to predict age from functional affective network connectivity in a separate sample of trauma-exposed women ages 18–29 with a history of either sexual or nonsexual trauma.

We hypothesized that, compared to nonsexual trauma, sexual trauma would be associated with greater functional affective network maturity, as indexed by greater difference between predicted age and true age. Such a result would be consistent with accelerated frontolimbic maturation following early-life sexual trauma. Given the well-documented differences in the maturational trajectories of different brain circuits (e.g., Grydeland et al., 2013; Lebel et al., 2012), the calculated discrepancy between predicted and true age should not be interpreted as an index of global brain maturity, but rather as an index of network-specific maturity (i.e., the functional maturity of affect-related circuitry). We also hypothesized that affective network maturity would be associated with mental health outcomes, which we derived by performing dimensionality reduction on eight measures of trauma-related outcomes. A negative association between affective network maturity and adverse outcomes would suggest that greater affective network maturity has adaptive value even in emerging adulthood. Conversely, a positive association would be consistent with mental health tradeoffs of accelerated affective network maturation. Finally, we hypothesized that trauma type would moderate the relation between affective network maturity and mental health outcomes, such that the association between affective network maturity and outcomes would be more pronounced in participants with sexual vs. nonsexual trauma.

Materials and methods

Participants

We recruited 36 trauma-exposed women ages 18–29 (M = 21.2, SD = 2.7) from an urban university in the northeastern United States based on their responses to the Life Events Checklist, a self-report screen for potential trauma exposure (Gray et al., 2004). We verified trauma exposure during a structured clinical interview. Participants were excluded if they had a history of head trauma, had a history of a developmental or neurological disorder, or had MRI contraindications.

Procedure

Participants completed two study sessions: a lab visit, which included a clinical interview and questionnaires, and an MRI scan. The study protocol was approved by the Institutional Review Board and conducted in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki).

Clinical interview

Following informed consent, participants completed a clinical interview that included the Clinician-Administered PTSD Scale (CAPS-5; Weathers et al., 2018) and the Structured Clinical Interview for DSM-IV Disorders (SCID; First et al., 2002a). We used the CAPS to assess trauma exposure and PTSD symptoms and the SCID to assess lifetime history of major psychiatric disorders. Trauma exposure was defined according to Criterion A of the CAPS.

Questionnaires

After the interview, participants completed “state” psychological measures; although the timeframe for each measure varied slightly, we conceptualized the responses to each to reflect the participant’s general state during participation, and we distinguished between these phenomena and clinical outcomes. The measures were the Perceived Stress Scale (PSS, Cohen et al., 1983), the Beck Depression Inventory II (BDI-II, Beck et al., 1996), the State-Trait Anxiety Inventory-State Version (STAI-S, Spielberger et al., 1983), and the Positive and Negative Affect Schedule (PANAS, Watson et al., 1988). The PSS is a 10-item measure of the degree to which life events over the past month are perceived as stressful, uncontrollable, and unpredictable. The BDI-II is a 21-item measure of depressed mood over the past two weeks. The STAI-S is a 20-item measure of state anxiety. The PANAS is a 20-item measure of state negative and positive affect.

MRI scan

The MRI session included 13 minutes of structural scans followed by a 7-min resting scan during which the participants were instructed to fixate on a white cross on a black background and let their minds wander. Imaging data were acquired on a Siemens MAGNETOM Prisma 3T MRI scanner using a 64-channel gradient head coil. We acquired T1-weighted (T1w) images using the MPRAGE sequence (TR/TE/flip angle = 2.4 s/2.28 ms/8°; FOV = 256 × 256 mm; voxel size = 0.8 mm³) and resting-state functional data using multiband image acquisition and a gradient-echo echo-planar imaging T2*-weighted sequence (TR/TE/flip angle = 995 ms/34 ms/52°; FOV = 208 × 208 mm; voxel size = 2 mm³; MB = 6; 425 volumes).

Training data set for machine learning

To build a normative model of affective network maturation, we used the enhanced Nathan Kline Institute-Rockland Sample (NKI-RS; Nooner et al., 2012) as the training data set for machine learning. Our goal was to build a sex-specific normative model of functional maturation of the affective networks from middle childhood through middle adulthood in the absence of prior trauma exposure or clinical history. To this end, we selected the normative sample using the following inclusion criteria: female, age 6–45, right-handed, no neurological, developmental, or psychiatric disorders, and no prior history of trauma. These stringent selection criteria were used to minimize known sources of variability in functional connectivity that are unrelated to age, such as sex, handedness, psychiatric history, and trauma exposure. Our choice of a wider age range for the training data set compared to the test data set was motivated by both conceptual and practical considerations. First, because the degree of the hypothesized maturation acceleration or delay was unclear, we chose relatively wide “safety margins” (±12–16 years) to avoid underestimating any deviations from the normative trajectory. In addition, our wide age range was meant to increase the model’s prediction accuracy for the age group of interest (age 18–29) by exposing the machine learning algorithm to relative extremes of the brain age spectrum relative to the age group of interest (i.e., middle childhood and middle adulthood). Finally, by including a wider age range, we were able to not only increase the sample size but also increase age-related variability while decreasing age-unrelated variability through stringent exclusion criteria.

Clinical history. In the NKI-RS sample, trauma history was assessed using the University of California at Los Angeles Posttraumatic Stress Disorder Reaction Index (Pynoos et al., 1998). Lifetime neurological and psychiatric history was assessed using the Structured Clinical Interview for DSM-IV-TR (First et al., 2002b) and the Adult ADHD Clinical Diagnostic Scale (Kessler et al., 2010) for participants above age 17 and the Kiddie Schedule for Affective Disorders and Schizophrenia for participants under age 18 (Kaufman et al., 1997).

Neuroimaging data. NKI-RS imaging data were acquired on a Siemens MAGNETOM Trio Tim 3T MRI scanner using a 32-channel gradient head coil. We used the selected participants’ T1w MPRAGE images (TR/TE/flip angle = 1.9 s/2.52 ms/9°; FOV = 250 × 250 mm; voxel size = 1 mm³) and the 9.5-min resting-state scans which were acquired using multiband image acquisition and a gradient-echo echo-planar imaging T2*-weighted sequence (TR/TE/flip angle = 1.4 s/30 ms/65°; FOV = 224 × 224 mm; voxel size = 2 mm³; MB = 4; 404 volumes). During the resting scan, NKI-RS participants were instructed to fixate on a white cross on a black background.

Data preparation

Neuroimaging data preprocessing

Usable MRI data were available for 85 healthy no-trauma controls in the training data set (NKI-RS sample) and 29 trauma-exposed participants in the test data set. The training and test data sets had comparable neuroimaging parameters and therefore were preprocessed similarly (see Supplementary Information, SI). Briefly, anatomical and functional scans were preprocessed using fMRIPrep (Esteban et al., 2019) and converted to the Connectivity Informatics Technology Initiative format using the fmriprep_ciftify BIDS-app (Dickie et al., 2019). All subsequent denoising operations and analyses were performed on the preprocessed Connectivity Informatics Technology Initiative data.

Denoising of resting-state functional data

To minimize the influence of head motion on estimates of functional connectivity and brain maturity, we excluded high-motion participants and performed nuisance regression and high-pass filtering (see SI). After exclusion of 6 high-motion participants from the training data set (see SI for exclusion criteria), the final sample included 79 participants (age 6–45, M = 23.2, SD = 11.8). No participants were excluded from the test data set.

Functional connectivity estimation

All functional connectivity estimation operations were performed on the cleaned and filtered functional data using Connectome Workbench and custom Matlab code (see SI). Briefly, we used the Cole-Anticevic Brain-wide Network Partition (Ji et al., 2019) to parcellate the brain into 718 regions of interest (ROIs) and excluded ROIs with group median temporal signal-to-noise ratio of <100. We calculated correlation coefficients between mean time series from each pair of the remaining 567 ROIs and performed Fisher’s r-to-Z transformation. To minimize the effects of residual motion on functional connectivity, we excluded edges that were correlated with head motion at uncorrected p < .05 (39%) from all subsequent analyses. The features that were used to train the final model are presented in SI.

Data analysis

Affective network maturity estimation

All machine learning operations were performed in Matlab R2019b using the Statistics and Machine Learning Toolbox and the Regression Learner App. We used support vector regression (SVR; Smola and Schölkopf, 2004; Vapnik, 1995) to train the normative model of affective network maturation using the selected NKI-RS participants as the training data set (Dosenbach et al., 2010; Nielsen et al., 2019) and the trauma-exposed participants as the test data set (see SI for details). Briefly, the affective network maturation model was trained on 500 features sampled from the cingulo-opercular and orbito-affective networks (Ji et al., 2019) based on the strength of their correlations with age. We performed leave-one-out cross-validation on the training sample to determine the accuracy of the trained SVR model and used permutation testing to determine whether the model’s performance was statistically different from chance. We then used the trained SVR model to predict the ages of trauma-exposed participants and calculated affective network maturity as the difference between predicted age and true chronological age.

The trained SVR model showed high accuracy (R ² = .86, r = .93, RMSE = 4.6, MAE = 3.8) and was statistically significant at p < .001 in the training data set (Figure 1). The MAE for the test group as a whole was 2.5. The performance of our model is consistent with reported model performances in previous life span studies that used a range of machine learning algorithms to predict age based on various neuroimaging modalities and feature sets (Franke et al., 2020). Although most prior studies used training data sets ranging from n = 122 (R ² = .57; age range: 7–31; Nielsen et al., 2019) to n = 2354 (R ² = .87; age range: 19–82; Liem et al., 2017), successful “brain age” prediction has been achieved with as few as 74 participants (R ² = .46; age range 8–18: Keding et al., 2021), 63 participants (R ² = .61; age range 18–45: Tian et al., 2016), or even 44 participants (R ² = .48; age range: 23–74; Neeb et al., 2006) in the training data set. Despite our relatively small training data set, our trained model showed better performance compared to previous studies that used resting-state fMRI to predict age in more heterogenous mixed-gender samples and achieved R ² in the range of .55–.61 (Dosenbach et al., 2010; Nielsen et al., 2019; Tian et al., 2016).

Figure 1. Performance of the machine learning algorithm trained to predict age based on affective network connectivity (MAE = 3.8, RMSE = 4.6, R ² = .86, r = .93, p < .001). Linear support vector regression was trained on 500 features sampled from the cingulo-opercular and orbito-affective networks (“affective networks”) based on the strength of their correlations with age. Affective network maturity was calculated as the difference between predicted age and true age (i.e., prediction errors).

We did not perform bias correction associated with regression to the mean, because prediction errors were not correlated with chronological age, r = −.33, p = .087, and because bias correction is inaccurate when the age ranges of the training and test samples are different, such as in our case (Liang et al., 2019). To test the degree to which regression to the mean may have influenced our results, we repeated our analyses of affective network maturity with chronological age as a covariate. Chronological age did not improve model fit and was not a statistically significant predictor.

Principal component analysis on mental health outcomes

To generate summary indices of mental health outcomes in trauma-exposed women, we performed dimensionality reduction on eight measures of psychological functioning and past and current clinical history. To estimate the amount of shared variance among different measures of mental health outcomes, we calculated Pearson’s correlations between perceived stress (PSS), depressed mood (BDI), state anxiety (STAI), negative affect (PANAS), PTSD symptom number and severity (CAPS), and number of current and past psychiatric diagnoses (SCID). The predominance of strong positive correlations among measures of mental health outcomes (Figure 2A) reflected a substantial amount of shared variance and warranted dimensionality reduction. We performed principal component analysis on the eight measures of mental health outcomes and retained the first two principal components with eigenvalues above 1, which together explained 72% of total variance (Figure 2B). After varimax rotation, self-reported “state” measures of perceived stress, depressed mood, anxiety, and negative affect loaded most strongly on the first rotated component (PC1), whereas number of current and past psychiatric diagnoses loaded most strongly on the second rotated component (PC2). Number and severity of PTSD symptoms loaded moderately on both components (Figure 2C). For each participant, we calculated individual PC1 scores (“state psychological outcomes”) and PC2 scores (“clinical outcomes”), which we used as two composite measures of mental health outcomes.

Figure 2. Principal component analysis on indices of mental health outcomes. Panel A: The high number of strong positive correlations among indices of mental health outcomes warranted dimensionality reduction. Panel B: The first two principal components together captured 72% of total variance. Panel C: The pattern of loadings of the original measures of mental health outcomes on the first two components after varimax rotation suggests that PC1 captures state psychological outcomes, whereas PC2 captures clinical outcomes.

Associations between trauma type, affective network maturity, and mental health outcomes

To determine the effect of trauma type on mental health outcomes, we tested group differences in PC1 scores (“state psychological outcomes”) and PC2 scores (“clinical outcomes”) between participants with sexual vs. nonsexual trauma. To determine the effect of trauma type on affective network development, we tested group differences in affective network maturity (age prediction errors) between participants with sexual vs. nonsexual trauma. To clarify the functional significance of affective network maturity, we tested associations between affective network maturity (age prediction errors) and mental health outcomes (PC1 and PC2 scores). Finally, we tested trauma type as a moderator of the relations between mental health outcomes and affective network maturity. To minimize the influence of any outliers, we used a robust general linear model with the bisquare weight function to test all associations and group differences. All statistical analyses were performed using Matlab R2019b and the Statistics and Machine Learning Toolbox.

Results

Descriptive data

Participant information is presented in Table 1. Thirteen participants (37%) endorsed sexual violence as their index trauma, 11 of whom had usable fMRI data. Twenty-two participants endorsed nonsexual trauma as their index trauma, 17 of whom had usable fMRI data. One participant did not provide the details of her trauma and was excluded from the analysis. Age at the earliest trauma ranged from 3 to 27 years, and time since the most recent trauma ranged from less than a year to 15 years. Participants with sexual vs. nonsexual trauma did not differ in age, Wilcoxon rank-sum test: Z = 0.81, p = .418, r = .14; number of traumatic events, Z = 0.15, p = .883, r = .02; age at the earliest trauma, Z = 1.47, p = .141, r = .25; or time since the most recent trauma, Z = 0.86, p = .389, r = .15.

Table 1. Participants

Note. PSS = Perceived Stress Scale; STAI-S = State-Trait Anxiety Inventory-State Version; BDI-II = Beck Depression Inventory II; PANAS = Positive and Negative Affect Schedule; NACM = negative alterations in cognitions and mood.

^a Any major psychiatric diagnosis other than PTSD, as assessed using the SCID.

^† p < 0.1; * p < 0.05; ** p < 0.01.

Participants with sexual vs. nonsexual trauma did not differ in perceived stress, Z = 0.86, p = .392, r = .14; depressed mood, Z = 0.79, p = .432, r = .13; negative state affect, Z = 0.55, p = .583, r = .09; or positive state affect, Z = −1.06, p = .288, r = −.18. However, participants with sexual vs. nonsexual trauma showed a nonsignificant trend (medium effect size) towards greater state anxiety, Z = 1.90, p = .058, r = .32. Almost half (46%) of the participants with sexual trauma met criteria for a provisional PTSD diagnosis compared to almost one-third (27%) of the participants with nonsexual trauma. Although there was no difference in probability of a provisional PTSD diagnosis by trauma type, χ² = 1.29, p = .256, participants with sexual trauma endorsed more, Z = 2.87, p = .004, r = .48, and more severe, Z = 2.00, p = .045, r = .34, PTSD symptoms compared to participants with nonsexual trauma. There were no differences by trauma type in number of current psychiatric diagnoses, Z = 0.74, p = .462, r = .12, or past psychiatric diagnoses, Z = 0.14, p = .888, r = .02. We repeated the above analyses in participants with imaging data and obtained a similar pattern of results, except the probability of a provisional PTSD diagnosis was now higher in the sexual trauma group, χ² = 4.04, p = .044.

Participants with sexual trauma showed a nonsignificant trend towards poorer state psychological outcomes indexed by PC1 scores, compared to participants with nonsexual trauma, F(1,33) = 3.49, p = .071, R ² = .10 (Figure 3). There was no group difference by trauma type in clinical outcomes indexed by PC2 scores, F(1,33) = 0.84, p = .365, R ² = .03.

Figure 3. Trauma type and mental health outcomes. Compared to participants with a history of nonsexual trauma, participants with a history of sexual trauma showed a nonsignificant trend towards poorer state psychological outcomes indexed by PC1 scores (left panel) but no difference in clinical outcomes indexed by PC2 scores (right panel). Bars represent group means; error bars represent standard errors.

Trauma type, affective network maturity, and mental health outcomes

Participants with sexual trauma showed greater affective network maturity indexed by age prediction errors (M = 2.5, SD = 2.9, range: −3.6 to 5.9), compared to participants with nonsexual trauma (M = 0.6, SD = 2.7, range: −3.8 to 7.1), F(1,26) = 5.33, p = .029, R ² = .17 (Figure 4A). Affective network maturity was not associated with either state psychological outcomes indexed by PC1 scores, F(1,26) = 0.41, p = .529, R ² = .02, or with clinical outcomes indexed by PC2 scores, F(1,26) = 0.63, p = .433, R ² = .03. However, trauma type moderated the relation between clinical outcomes (independent variable) and affective network maturity (dependent variable), F(1,24) = 4.44, p = .046, η_p ² = .16. Poorer clinical outcomes were associated with less affective network maturity in participants with sexual trauma but not in participants with nonsexual trauma (Figure 4B). Trauma type did not moderate the relation between psychological outcomes and affective network maturity, F(1,24) = 0.21, p = .648, η_p ² = .01.

Figure 4. Trauma type, affective network maturity, and clinical outcomes. Panel A: Compared to participants with a history of nonsexual trauma, participants with a history of sexual trauma showed greater affective network maturity indexed by age prediction errors. Panel B: Trauma type moderated the relation between affective network maturity and clinical outcomes indexed by PC2 scores. Less affective network maturity was associated with poorer clinical outcomes in participants with sexual trauma but not in participants with nonsexual trauma. Bars represent group means; error bars represent standard errors. The lines of best fit were generated using robust regression.

Discussion

In this proof-of-concept study, we used a machine learning approach to test atypical functional maturation of affect-related brain circuitry as a candidate mechanism that might account for the relation between exposure to early sexual trauma and adverse mental health outcomes in adulthood. Our first objective was to compare the functional maturity of affective networks in young women who were exposed to either sexual or nonsexual trauma during formative years of brain development (i.e., early-life sexual trauma). Consistent with our hypothesis, young women with early sexual trauma showed greater affective network maturity compared to women with nonsexual trauma. This result aligns with the previously documented relation between early-life adversity and accelerated maturation of affect-related frontolimbic circuitry (e.g., Colich et al., 2020; Gur et al., 2019; Thijssen et al., 2022; Tyborowska et al., 2018). However, our data also suggest that the effect of early traumatic stress on neurodevelopment depends on the type of trauma. In line with previous evidence that sexual trauma is uniquely capable of advancing pubertal development (e.g., Mendle et al., 2016; Noll et al., 2017), our results suggest that sexual trauma also has a greater impact on frontolimbic maturation compared to other trauma types. These data also extend prior research in children and adolescents and show that the neurodevelopmental effects of traumatic stress are evident even in emerging adulthood, which is consistent with the protracted maturational trajectory of affect-related circuitry (e.g., Lebel et al., 2012; Narvacan et al., 2017). Together with prior literature, our results suggest that the developmental trajectory of affect-related circuitry remains sensitive to traumatic stress into emerging adulthood, and the effects of sexual trauma on affective network maturation can manifest beyond childhood and adolescence.

Our second objective was to test the functional significance of affective network maturity in relation to mental health outcomes in emerging adulthood. Counter to our hypothesis, affective network maturity was not associated with either state psychological or clinical outcomes in trauma-exposed women as a group. However, trauma type moderated the relation between affective network maturity and clinical outcomes. In women with sexual trauma, lesser affective network maturity was associated with poorer clinical outcomes, whereas there was no such relation in women with nonsexual trauma. In contrast, state psychological outcomes were not associated with affective network maturity in the trauma-exposed sample as a whole or in either trauma group separately. These results suggest that different types of trauma-related outcomes (i.e., state psychological vs. clinical) are mediated by different mechanisms and that subclinical levels of psychological distress are unrelated to the maturation of affect-related circuitry in early-trauma survivors.

The relation between greater affective network maturity and better clinical outcomes in sexual trauma survivors is consistent with the stress acceleration theory, which posits that accelerated maturation of affect-related circuitry is adaptive under conditions of early adversity (e.g., Callaghan & Tottenham, 2016). Our results suggest that greater maturity of affect-related circuitry might reflect resilience or constitute a protective factor against adverse clinical outcomes in survivors of early-life sexual trauma. This result is consistent with prior evidence of greater frontolimbic maturity being protective against internalizing symptoms in adolescents with early-life adversity (e.g., Brieant et al., 2021). In addition, given the robust literature documenting poorer clinical outcomes overall in sexual trauma survivors (e.g., Dworkin et al., 2017; O’Brien et al., 2016), the absence of a difference in clinical outcomes by trauma type might reflect a high prevalence of resilience in our sexual trauma group. This absence might be accounted for by greater affective network maturity in our sample of sexual trauma survivors.

Whereas accelerated affective network maturation might be adaptive, failure to accelerate maturation of affect-related circuitry following early sexual trauma might reflect a lack of adaptation, which might in turn confer vulnerability to adverse clinical outcomes. This interpretation is broadly consistent with previous evidence that delayed frontolimbic maturation is associated with psychiatric symptoms in adolescence (e.g., Albaugh et al., 2017; Ducharme et al., 2014; Keding et al., 2021; Vanes et al., 2020) and that slower development of frontal and limbic structures mediates the relation between early-life adversity and negative outcomes (e.g., Barch et al., 2022; Hanson et al., 2015; Whittle et al., 2016). Our results are consistent with prior literature and suggest that early-life trauma can alter frontolimbic development, however we also demonstrate that deviation from the normative maturational trajectory is not necessarily pathological and might reflect a compensatory developmental process that is necessary for successful adaptation following trauma.

Our results are consistent with the hypothesis that delayed maturation of the affect-related circuitry leads to higher risk of adverse clinical outcomes in adults with early sexual trauma. However, the limitations of this study, such as the small sample size and cross-sectional design, preclude us from making strong claims about causal relations between sexual trauma, brain maturation, and clinical outcomes. Our data suggest that exposure to early-life sexual trauma accelerates functional maturation of affect-related networks. However, due to the cross-sectional nature of our study and our focus on emerging adults, we cannot determine whether this deviation from the normative maturational trajectory occurs shortly following trauma or later in development due to sleeper effects (e.g., Rakesh et al., 2021). In addition, because we measured brain maturity only once in emerging adulthood, we cannot assume that the observed direction or functional significance of accelerated maturation is the same at earlier stages of development. For example, risk of depression has been linked to greater global brain maturity in younger adolescents (Drobinin et al., 2022) and delayed brain maturation in older adolescents and emerging adults (de Nooij et al., 2020). Similarly, although we show that greater functional maturity of affect-related circuitry is associated with better clinical outcomes in young women with sexual trauma, we cannot determine the temporal order or causal relation between brain maturity and outcomes. For example, disrupted neurodevelopment could be an outcome rather than an underlying mechanism of trauma-related symptoms (e.g., Muetzel et al., 2018). Finally, because cross-sectional and longitudinal research on neurodevelopment can lead to different conclusions about the effects of early adversity on brain maturation (e.g., Rakesh & Whittle, 2021), it is important to replicate our results in larger, longitudinal samples.

Another promising direction will be to investigate interactions among trauma characteristics, such as trauma type, chronicity, and timing, on brain development and mental health outcomes. For example, future research with larger samples would be necessary to determine the effects of trauma timing (e.g., age of onset, duration, recency) on the relationship between frontolimbic maturation and mental health outcomes. Due to our limited sample size, we were unable to probe such interactions or control for other trauma characteristics to isolate the unique contributions of trauma type. Despite this limitation, our results underscore the importance of considering trauma characteristics as potential moderators of the relations between trauma exposure and its sequelae. In addition, because our sample included high-functioning young women on the lower end of the psychopathology spectrum, our results might have been influenced by high levels of resilience in our sample. We suggest that in a clinical sample of sexual trauma survivors with more severe symptoms and higher diagnostic rates, we would observe less affective network maturity compared to nonsexual trauma survivors.

Conclusion

Our results suggest that exposure to sexual trauma during formative years of brain development is associated with greater functional maturity of affect-related brain circuitry in emerging adulthood. Accelerated functional maturation of affective circuitry might represent an adaptation that protects survivors of early sexual trauma from adverse clinical outcomes later in life. In contrast, failure to accelerate maturation of affect-related circuitry following early sexual trauma might confer vulnerability to mental health problems in adulthood. Given the prolonged developmental trajectory of affect-related circuitry, functional maturity of affect-related circuits might hold prognostic or diagnostic promise as a potential neural marker of resilience or vulnerability to negative clinical outcomes in emerging adulthood. For example, interventions that improve clinical outcomes in young survivors of early-life sexual trauma might normalize the maturational trajectory or promote compensatory developmental processes, such as accelerated maturation, in affect-related circuits. Establishing the clinical utility of affective network maturity as a potential resilience or vulnerability marker is a promising avenue for future work.