Treatment Outcome of Posttraumatic Stress Disorder: A White Matter Tract Analysis

Abstract

Despite the development of empirically supported treatments for posttraumatic stress disorder (PTSD), many individuals remain symptomatic following therapy or dropout prematurely. Neuroimaging studies examining PTSD treatment outcome may offer valuable insights into possible mechanisms that may impact treatment efficacy. To date, few studies of PTSD have used neuroimaging to examine symptom change following completed treatment, and most have focused on gray matter. Studies of white matter are equally important, as changes in white matter integrity (WMI) are connected to a host of detrimental outcomes. The current study examined symptom change of 21 women with PTSD as a result of interpersonal violence who received baseline diffusion tensor imaging (DTI) scans and completed 12 weeks of Cognitive Processing Therapy (CPT). After controlling for baseline PTSD severity, fractional anisotropy (FA) in the left internal capsule, posterior limb of the internal capsule, left cingulate gyrus, superior longitudinal fasciculus, and splenium of the corpus callosum was predicted by PTSD symptom change. Results contribute to understanding neural changes within therapy and may assist in predicting individual treatment response. Namely, by identifying areas potentially impacted by PTSD treatment, future studies may be able to connect the function of these white matter areas to better predict patient PTSD treatment outcome.

Post-Traumatic Stress Disorder (PTSD) is a debilitating disorder that occurs following a traumatic event, such as domestic assault, combat, or a natural disaster. Experiencing the event can occur in a number of forms such as directly experiencing the event or witnessing an event in person (American Psychiatric Association, 2013).

PTSD Treatment Efficacy

Despite the wealth of literature documenting PTSD treatment efficacy, and the status of several treatments as either “conditionally” or “strongly recommended” for PTSD by the American Psychological Association (2017), treatment outcomes for individuals remain highly variable. Studies have noted the large rate of individuals who do not respond to treatment after completion (72% in female assault survivors: Zang, Su, McLean, & Foa, 2019; 33–46% in a mixed sample meta-analysis: Bradley, Greene, Russ, Dutra, Westen, 2005; 60–72% in military/veteran populations: Steenkamp et al., 2015). As such, recent research has increasingly focused on identifying psychosocial and biological predictors of treatment response in an effort to enhance treatment outcomes (Colvonen et al., 2017).

Neuroscience and PTSD

Neuroimaging offers a useful methodology to examine abnormalities in brain systems, and thus could inform treatment efficacy. Increasing evidence from neuroimaging studies supports the idea that pretreatment biological brain markers may be useful in predicting treatment success and that this type of research can improve the understanding of the neurobiology of individuals who experience PTSD symptom remission with treatment compared with a lack of treatment response (Kennis et al., 2015). With regard to PTSD, much of the neuroimaging research has focused on gray matter regions that play a role in the maintenance of PTSD symptoms. Studies of functional MRI (fMRI) have found differing activation in a number of brain regions in individuals with PTSD, including the amygdala, the medial prefrontal cortex (mPFC), the anterior cingulate cortex (ACC), thalamus, visual cortex, parietal cortex, inferior frontal gyrus, posterior cingulate cortex, parahippocampal gyrus (Bremner, 2007), and hippocampus (van Rooji et al., 2017).

White Matter Findings in PTSD

Though the majority of neuroimaging work in PTSD has focused on functional findings in gray matter regions, an examination of white matter tracts, which connect gray matter regions, is equally important as it provides insight to the structural integrity, functionality and efficiency of the brain. For example, abnormalities in white matter are often associated with deficits in various cognitive domains (processing speed: Cremers et al., 2016; Hong et al., 2015; visuospatial ability: Muetzel et al., 2015; executive functioning: Cremers et al., 2016; global cognition: Cremers et al., 2016; Marques, Soares, Magalhaes, Santos, Sousa, 2015) and have been connected to a range of mental disorders (schizophrenia: Lee et al., 2013; bipolar disorder: Torgerson et al., 2013; depression: Zhu et al., 2011; PTSD: Sun et al., 2015).

Diffusion tensor imaging (DTI) identifies WM tracts that connect certain areas of the brain by measuring the three-dimensional diffusion of water within the WM tracts in vivo. Fractional Anisotropy (FA) quantifies the fraction of diffusion that is anisotropic by computing the relative difference between the largest and smallest eigenvalues (Alexander, Lee, Lazar, & Field, 2007; Feldman et al., 2014; Soares, Marques, Alves, & Sousa, 2013; Mori, Zhang, 2006). Typically, lower FA values relate to impaired white matter integrity (WMI), as increased diffusion and lack of consistency in movement direction most likely indicates a dysfunction in fiber integrity.

An overview of studies investigating WM microstructure in populations with PTSD showed that WM volume reductions were more common than WM volume expansions. A majority of these findings involved the corpus callosum, which is the largest WM fiber bundle in the brain (Daniels, Lamke, Gaebler, Walter, Scheel, 2013). Indeed, a study examining the microstructural WMI in women exposed to interpersonal violence found higher FA in the genu of the corpus callosum in participants with PTSD compared with trauma-exposed controls (Graziano et al., 2019). Additionally, a meta-analysis of both ROI- and whole-brain DTI studies in individuals with PTSD found robustly decreased FA in the genu of the corpus callosum (GCC) and the left corticospinal tract (CST), higher FA in inferior fronto-occipital fasciculus (IFOF) and left inferior temporal gyrus, and lower FA in subregions of the corpus callosum. Each of these WMI abnormalities were found to be modulated by age and trauma type (Ju et al., 2020).

Few studies have examined DTI data in conjunction with treatment outcomes in PTSD. One study of 72 veterans found greater pretreatment activation in the dorsal ACC, an area connected to the cingulum, to be predictive of poorer treatment outcome in individuals with PTSD (van Rooij, Kennis, Vink, & Geuze, 2015). In terms of WMI, Kennis et al. (2015) found the cingulum to predict PTSD treatment success using eye movement desensitization and reprocessing (EMDR). Specifically, higher FA at baseline was found in the right cingulum in patients who showed poor treatment outcome. Though lower levels of FA are typically thought to represent dysfunction, abnormal levels in either direction can indicate dysfunction depending on the brain region (Soares et al., 2013). As the cingulum is part of the limbic system, higher FA may be an indication of dysfunction in fear circuitry, possibly hindering treatment progress and predicting less positive treatment outcome.

Current study

There is a paucity of studies examining how WMI may predict treatment outcome in PTSD, and findings from treatment outcome studies may identify possible predictors of treatment efficacy. Neuroimaging studies examining treatment efficacy may aid clinicians in the selection of the most appropriate treatment and improve personalized medicine, such that treatments could be selected based on a patient’s individual baseline neuroimaging signature. This study addresses this gap in the literature explicitly by examining WMI in a sample of females with PTSD following interpersonal violence and connecting pretreatment WMI to treatment outcome. This is the first study to our knowledge to examine WMI and PTSD treatment outcomes in victims of interpersonal violence. Interpersonal violence is particularly common (Breslau, 2002; Flett et al., 2004) and is more predictive of PTSD than noninterpersonal traumas (Breslau, 2002; Luthra et al., 2009). Additionally, research has shown significant differences in PTSD symptoms with varying trauma types (Haldane & Nickerson, 2016; Wanklyn et al., 2016). As such, neural presentation may also vary by trauma type, making this sample an important one to study.

Methods

Twenty-one women meeting criteria for a DSM-IV-TR diagnosis of PTSD were recruited at a Midwest trauma center or through local advertisements. All participants reported PTSD following an interpersonal trauma and completed a full course of Cognitive Processing Therapy (CPT). All participants were right-handed; handedness has been linked to brain lateralization and differences in lateralization could confound results. Exclusion criteria consisted of (a) a diagnosis of a neurological disorder such as dementia, stroke, brain tumors, seizure disorder, multiple sclerosis, or Parkinson’s disease; (b) current comorbid alcohol or substance use disorder, schizophrenia or other psychotic disorder, obsessive compulsive disorder (OCD), or bipolar disorder; (c) currently being administered with psychotropic drugs or drugs that affect the central nervous system (CNS) such as beta-blockers, mood stabilizers, antipsychotics, or other antidepressants; (d) active suicidal risk as assessed by the investigator; (e) significant cognitive limitations that may interfere with testing procedures; (f) history of head injuries; and (g) implanted devices (such as a pacemaker) or other metallic objects that are contraindicated for magnetic resonance imaging (MRI).

Measures

PTSD symptoms were examined with the Clinician-Administered PTSD Scale–IV (CAPS-IV; Blake et al., 1998; Blake et al., 1995). The CAPS-IV is a 25-item semi-structured interview that measures both the intensity and frequency of PTSD symptoms (0–4 scale). The CAPS-IV has high test–retest reliability (.77-.96 for symptom clusters, .90-.98 for total score), internal consistency (.85-.87 for symptom clusters, .94 for total score; Blake et al., 1995), and high interrater reliability (.92–1.00 for frequency, .93-.98 for intensity; Hovens et al., 1994). Participants must have had a CAPS score above 45 (Orr et al., 1997) and met the original scoring criteria by Blake et al. (1995); PTSD symptoms are deemed present if the frequency is rated as 1 or higher and the intensity is rated as 2 or higher.

Image Acquisition

A Siemens 3 T TrioTim MRI scanner (Erlangen, Germany) was utilized to collect DTI scans. First, participants underwent a T1 3D magnetization-prepared rapid gradient echo image (MP-RAGE). With a 1 Å~ 1 Å~ 1 resolution, the structural images were garnered using a sagittal 3D T1-weighted sequence with a repetition time (TR) of 2.4 s, a time-to-echo (TE) of 3.13 ms, a flip angle of 8°, and an inversion time (TI) of 1,000 ms. For DTI, the b-value was set at 1,400 s/mm2 with an acquisition of a reference image (b = 0). One acquisition entailed 25 transverse slices, 25 directions, 2 mm thickness (no gap), and an in-plane resolution of 2 Å~ 2 mm2. A second DTI scan was gathered for each participant in case of artifacts.

Image Analysis

T1 images were examined for any obvious abnormalities (e.g., enlarged ventricles, cysts, tumors, hyperintensities) by an MRI technician. This study utilized a whole brain analysis, with imaging data preprocessed with FMRIB Software Library (FSL) to account for asynchronous slice acquisition and odd/even slice intensity differences from interleaving. Images were then corrected for spatial distortion with an eddy current correction (Friston, Williams, Howard, Frackowiak, & Turner, 1996; Snyder, 1996). Next, a brain mask for the data was created to ensure that areas beyond the scans were not included. Finally, tensors were fitted with the generated brain mask. Processing was done using the ENIGMA protocol, which is detailed elsewhere (Jahanshad et al., 2013) and is available online (http://enigma.ini.usc.edu/protocols/dti-protocols/). Briefly, the data were processed with FSL’s tract-based spatial statistics (TBSS; http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/TBSS) and were changed to display individual FA values on the hand-segmented ENIGMA-DTI skeleton mask. After extracting the skeletonized WM and projections of individual FA values, ENIGMA tract-wise regions of interest were transferred to extract the mean FA across the full skeleton and average FA values for all regions. Regions included the entire white matter tract, as well as the left and right sides of each tract when applicable. Last, images were examined once more to check for artifacts or abnormalities.

Cognitive Processing Therapy

All participants completed Cognitive Processing Therapy (CPT; Resick & Schnicke, 1992) after their first imaging scan. CPT is a 12-session cognitive therapy developed for PTSD, and focuses on altering distorted thoughts. CPT begins with psychoeducation of PTSD and treatment rationale. At the conclusion of Session 3, the client is asked to write a detailed account of their trauma to be read aloud to the therapist in Session 4. This is repeated in Session 5. Throughout this process, the therapist helps with the client’s emotional processing and challenges erroneous beliefs about the client’s role in the trauma. The remaining modular sessions focus on adjusting global views on safety, trust, power, and esteem that may be affected by cognitions from the index trauma.

CPT has been well established as an effective intervention to significantly reduce PTSD symptoms (Alvarez et al., 2011; Chard, 2005; Iverson, King, Cunningham, & Resick, 2015; Resick et al., 2008; Resick et al., 2015). Notably, CPT is effective in the reduction of both PTSD and depression symptoms (Iverson et al., 2015; Resick et al., 2008) across different trauma types (rape, Iverson et al., 2015; combat, Alvarez et al., 2011; interpersonal violence, Resick et al., 2008; childhood sexual abuse, Chard, 2005).

Statistical Analysis

All statistical analyses were conducted using SPSS 25.0 (SPSS, Inc., Chicago, IL). First, DTI and PTSD data were screened and reviewed for potential outliers with box plots, linearity with scatter plots, normality with P-P plots, multicollinearity with regression statistics, and homoscedasticity with scatterplot of residuals.

Treatment improvement (i.e., change scores) was calculated by subtracting posttreatment CAPS scores from pretreatment CAPS scores. As a preliminary investigation to determine regions of interest (ROI), a correlation was conducted between each region and change scores. WM tracts that were significantly related to change scores were deemed ROIs. Next, separate hierarchical regressions were conducted with each ROI as the dependent variable. To control for initial symptom severity, baseline PTSD scores were included in block 1 of the regression. Change scores were included in block 2. Finally, the Benjamini-Hochberg (B-H) procedure with a maximum false discovery rate (Q) of 0.1 was used to account for multiple comparisons within hierarchical regression analyses (Glickman, Rao, Schultz, 2014).

Results

DTI and PTSD data passed screening tests for outliers, linearity, multivariate normality, multicollinearity, and homoscedasticity. Table 1 reports the demographic and clinical data for the sample. Of note, the there was a substantial reduction in CAPS scores for most participants ($\overline{\text{x}}=43.857$), with a mean pretreatment CAPS score of 66.524 and a mean posttreatment CAPS score of 22.667. The mean age was 31.905 and 71.4% of the sample identified themselves as Caucasian. One participant was missing both race and education data.

Table 1-.

Participant Demographics

	Participants (n=21)
Mean Age (SD)	31.905 (11.040)
Race (%)
Caucasian	15 (71.4)
African American	4 (19.0)
Hispanic	1 (4.8)
Years of Education (%)
12	2 (10)
13–16	12 (60)
>17	6 (30)
Mean Pretreatment CAPS (SD)	66.524 (16.467)
Mean Posttreatment CAPS (SD)	22.667 (23.133)
Mean CAPS Change (SD)	43.857 (16.951)

ROI analyses showed FA in the following eight regions to be significantly related to change scores: internal capsule (r=.442; IC; p=.045), left cingulate gyrus (r=.461; CGC-L; p=.035), the posterior limb of the internal capsule (r=.524; PLIC; p=.015), left posterior limb of the internal capsule (r=.456; PLIC-L; p=.038), right posterior limb of the internal capsule (r=.497; PLIC-R; p=.022), superior longitudinal fasciculus (r=.484; SLF-L; p=.026), splenium of the corpus callosum (r=.472; SCC; p=.031), and left internal capsule (r=.464; IC-L; p=.034). After controlling for initial PTSD symptom severity, FA in the CGC-L (β=.469; p=.034), IC-L (β=.473; p=.032), PLIC (β=.528; p=.017), SLF-L (β=.494; p=.022), and SCC (β=.475; p=.034) significantly predicted change in CAPS scores with treatment. All results passed the B-H multiple comparisons procedure.

Discussion

This is the first study to our knowledge to examine WMI and PTSD treatment outcomes in victims of interpersonal violence. Overall, this study showed that the left internal capsule, posterior limb of the internal capsule, left cingulate gyrus, superior longitudinal fasciculus, and splenium of the corpus callosum predicted change in PTSD symptoms with treatment. Of note, as these results are correlational, these regions represent potential predictors of change in PTSD symptoms; more studies are needed to confirm these results. Further, across multiple WM tracts, FA was positively correlated with treatment change. This positive association between FA and treatment change could indicate that: 1) lower FA at baseline was predictive of higher PTSD change scores (i.e. greater symptom reduction) and/or 2) higher FA at baseline was predictive of lower PTSD change scores (i.e. less symptom reduction). Importantly, abnormal FA levels in either direction can suggest dysfunction (Soares et al., 2013). The only other DTI study (Kennis et al., 2015) examining treatment outcome in PTSD identified higher FA in the cingulum bundle, an important WM tract that connects the frontal lobe to the limbic system, to predict worse treatment outcome. This is consistent with the current study, as the cingulum bundle is located in the cingulate gyrus. However, more studies are needed to elucidate the cause and direction of this relationship, particularly because DTI studies of PTSD have shown both lower (Fani et al., 2012; Kim et al., 2005; Wang et al., 2010) and higher (Abe et al., 2006; Graziano et al., 2019; Kennis et al., 2015) FA to relate to dysfunction.

Though few prior DTI treatment outcome research studies have been conducted, our results align with several previous findings. As previously mentioned, Kennis et al. (2015) found higher cingulum bundle FA at baseline to predict worse treatment outcome. Other DTI studies have also connected changes (both increases and decreases in FA) in the cingulum to PTSD (Abe et al., 2006; Kim et al., 2005; Sanjuan et al., 2013). Indeed, one study (Hu et al., 2016) suggested that damage to the cingulum may explain previous fMRI findings of increased activity in the amygdala and decreased activity in the prefrontal cortex. Regarding the corpus callosum, PTSD has been associated with WM volume changes in the corpus callosum (Daniels et al., 2013), and a previous study also demonstrated higher FA in the corpus callosum in participants with PTSD compared to trauma-exposed controls (Graziano et al., 2019). Additionally, a study using whole brain DTI (tract based spatial statistics) found increased FA values in the posterior limb of the left internal capsule in participants with acute PTSD following single-prolonged stress compared to healthy controls (Xi et al., 2013). Finally, Hu et al. (2016) connected altered FA in the SLF in the acute phase following motor vehicle accidents in subjects who would later develop PTSD. In line with this, a meta-analysis of DTI studies (Daniels et al., 2013) identified changes in FA in the SLF as frequently connected to traumatic exposure. Taken together, previous research supports the findings of this study; namely, that the WM tracts identified in this study as potentially important for treatment efficacy (CGC-L, IC-L, PLIC, SLF-L, and SCC) have also been shown to be affected by trauma exposure.

By identifying WM tracts that potentially predict treatment efficacy, the current study contributes to understanding mechanisms of change within therapy. If future studies can concretely identify the function of the WM tracts of interest, these findings may assist in elucidating why some patients respond or fail to respond to PTSD treatment. As this DTI study examined women with PTSD as a result of interpersonal violence, it makes particular contributions in understanding the possible neural mechanisms of change within a common, often severe trauma type (Breslau, 2002; Luthra et al., 2009). Additionally, connecting these WM tracts to specific, observable symptoms may assist in providing an indication of a patient’s WMI without the need for neuroimaging, progressing treatment towards the goal of personalized medicine in which treatments are chosen for each individual based on a set of observable, studied criteria.

Though this study contributes to the literature in a meaningful way, it is not without limitations. First, this study is cross-sectional in nature, which prevents firm conclusions about changes in FA over the course of therapy. Future treatment outcome studies using DTI should examine the regions of interest identified in this study. Additionally, as data collection began prior to DSM-5, this study used DSM-IV-TR criteria for PTSD. However, Kilpatrick and colleagues (2013) found that the vast majority of individuals meeting DSM-IV criteria for PTSD would also meet criteria for PTSD using DSM-5 criteria (Kilpatrick et al., 2013). Further, as this study was limited to studying women, results may not be generalizable across gender. Finally, due to the clinical, treatment-seeking nature of the sample, this study used a relatively low sample size. Thus, this study should be viewed as a preliminary study, with larger, more diverse longitudinal studies needed in the future.

Despite these limitations, results from this study add to a small but growing body of literature examining PTSD treatment outcome utilizing DTI. Our results suggest that the CGC-L, IC-L, PLIC, SLF-L, and SCC may be predictive of PTSD treatment efficacy in women with PTSD as a result of interpersonal violence. By identifying these regions, this study contributes to the understanding and possible neural mechanisms of change after PTSD treatment. In combination with future longitudinal studies, the identification of these WM tracks may contribute to the refinement of current PTSD therapies or the development of new interventions that will improve treatment response.

Highlights:

• Few neuroscience studies of white matter have examined PTSD treatment outcome.

• White matter tracts predict posttraumatic stress disorder treatment outcome.

• Changes in white matter underlie PTSD treatment mechanisms of change.