|Identifying Sex-Dependent Connectivity patterns in Trauma and PTSD|
A. Research Question, Goal, or Specific Aims
Aim 1: We plan to replicate our findings (Helpman et al., 2021) regarding sex-dependent functional connectivity pathways in PTSD within a larger, global database by analyzing ENIGMA data, using the most recent data release.
Aim 2: We plan to extend these findings by adding a trauma-nonexposed group (lacking in the original sample) as well as by further elucidating underlying mechanisms by utilizing machine learning approaches and exploring other imaging modalities (structural MRI, DTI). Aim 3: We plan to further extend our analyses to examine potential differences in sex-dependent patterns between those with adult vs. childhood trauma and interpersonal vs. non-interpersonal trauma.
B. Analytic Plan
Posttraumatic stress disorder (PTSD) is a prevalent and debilitating disorder, associated with significant burden1. This disorder disproportionately affects women, who have twofold the prevalence relative to men2. Recently, neural substrates of PTSD have been elucidated, and these include specific regions and functional circuits (e.g., limbic system, prefrontal cortex, striatal regions, as well as the hippocampus) as well as larger systems, in which they are organized (salience network, the default mode network, and the central executive network3. Neural activation in many of these appears, at least in part, to be regulated by steroid hormones such as estrogen4, progesterone5, and allopregnanolone6. Therefore, the function of these neural circuits and systems may be key in understanding sex-dependent bias in PTSD. Neurobiological research in PTSD has, to date, commonly treated sex as a potential confounder, rather neglecting to examine sex-dependent patterns of rs-FC in PTSD. Recently, we have analyzed and published data that has demonstrated, for the first time, sex-specific patterns of neural functioning among trauma-exposed individuals with and without PTSD7. The findings suggest that the patterns of connectivity distinguishing trauma-exposed individuals who develop PTSD from those who do not are reversed for males and for females. For males, weaker DMN and DMN-SN (AMG-PCC, HIP-precuneus) connectivity was found in those with PTSD compared to those without PTSD, and stronger SN-DMN (ACC-precuneus) connectivity in those with PTSD compared to those without PTSD. For females, the opposite pattern emerged: stronger connectivity was found among individuals with PTSD than in trauma-exposed healthy individuals, for AMG-PCC and HIP-precuneus, and weaker connectivity for ACC
precuneus among individuals with PTSD than in trauma-exposed healthy controls, for the third. Main effects for sex and PTSD diagnosis were consistent with those documented so far in the literature, suggesting reliable findings 8–12.
Previous studies seeking to understand the sex discrepancy in PTSD prevalence have identified increased rates of exposure to interpersonal trauma, particularly gender-based violence (GBV(, as possible contributors to increased risk 13. Current data suggests female survivors of childhood sexual assault (CSA) and other forms of childhood maltreatment have distinct neurobiological profiles when compared to healthy controls, with underpinnings in delayed or premature aging of various parts of the brain associated with emotion regulation, executive functioning, and learning. Specifically, the literature highlights the differences in the connectivity and cortical thickness of fronto-parietal network (FPN) between healthy controls and survivors of sexual abuse. Concurrently, it underscores varying outcomes in GMV in the visual cortex, frontal lobe, corpus callosum, and the ventral attention (VAN) in survivors of other types of childhood maltreatment. The impact of the abuse on brain development is directly correlated with the age at which the abuse occurred, its severity, and duration 14,15. The literature also highlights the distinct severity of the neurobiological impact of abuse prior to the age of 12 16,17, thereby providing support for critical periods of neurological development in females. With adults, it appears that exposure to interpersonal, gender-based violence (intimate partner violence and sexual abuse) among females is associated with changes in GMV in the amygdala and insula, and reduced rsFC between the insula, amygdala, and prefrontal cortex 18–22. Therefore, it would appear that not only sex itself but its interaction with age and type of trauma may be key in informing the patterns of structural and functional neurobiological trauma signature.
Further clarifying the sex, development, and trauma type dependence of PTSD neurobiological substrates is instrumental in tailoring individually appropriate, neurobiology informed treatments. As endocrine data for the sample is unavailable, underlying neuroendocrine mechanisms can only be inferred from results, but may serve as basis for further investigation of this substrate.
The ENIGMA Consortium PTSD workgroup, has aggregated and preprocessed resting-state fMRI data from 29 international sites totaling 2,625 individuals, who are divided roughly evenly into PTSD cases and trauma-exposed controls. Harmonization of DSM-5 and DSM-IV PTSD diagnosis and severity has been achieved by translating DSM-5 to DSM-IV with established procedures (Delin Sun, 2019). Childhood Trauma Questionnaire (CTQ) will be used to stratify subjects based on the number of categories of childhood trauma (0, 1, 2+) (Edward A. Walker, 1999). All cohorts in ENIGMA-PTSD excluded bipolar disorder, psychotic disorders, substance dependence, alcohol dependence, and neurological disorders including TBI and most cohorts have excluded Axis I disorders except MDD and PTSD.
Replication of previous resting-state functional connectivity analyses will be carried out using a seed-based and ICA approaches. ROI-to-ROI correlation matrix and ICA component maps will be used using ROIs identified as key nodes in CEN (lateral prefrontal cortex [lPFC]), SN (anterior cingulate cortex [ACC], insula, amygdala [AMG]) and DMN (medial prefrontal cortex [mPFC], posterior parietal cortex [PCC], and hippocampus [HIP]).
We utilize random forest, SVM, ML methods in order to identify sex-specific biomarkers. Logistic regression will be performed using scikit-learn library in Python. Both filter-based and wrapper-based approaches will be used for feature elimination. The leave-one-out cross validation (LOOCV) strategy will be used to validate the performance of our proposed approach. Accuracy, sensitivity and specificity are defined based on the prediction results of LOOCV to quantify the performance of all approaches.
ROI-to-ROI connectivity matrix and ICA component maps will be used using ROIs identified as key nodes in CEN, DMN, SN, and VAN (such as ventral frontal cortex [VFC]; temporoparietal junction [TPJ], dorsolateral prefrontal cortex [dlPFC]; inferior parietal lobe [IPL]; intraparietal sulcus [IPS]; lateral parietal cortex [LPC]; medial cingulate cortex [MCC]; striatum; thalamus). We will focus on pathways that have shown the best discriminatory power (in aim 2) and test sex by trauma type interactions there using a 2 (PTSD/non) X 2 (male/female) X 2 (child/adult trauma) ANOVA with interactions. Despite being a saturated model, the current sample should provide ample power given current calculations. Exploratory aim: Repeat aim 3 with interpersonal/non rather than child/adult trauma. With current information, only approximately 121 subjects have reported interpersonal trauma, therefore a sample of 242 may be reached, with matched participants. This falls below the minimal N=271 required for moderately powered results with a moderate effect size, thus making this an exploratory aim.