Correlation Between Rostral Dorsomedial Prefrontal Cortex Activation by Trauma-Related Words and Subsequent Response to CBT for PTSD

Abstract

Trauma-focused cognitive behavioral therapy (CBT) is an important component of evidence-based treatment for post-traumatic stress disorder (PTSD), but the efficacy of treatment varies from individual to individual. It is hypothesized that some of this variability derives from interindividual differences in the brain’s intrinsic response to trauma-related stimuli and in activity of executive functional regions. We sought to characterize these differences using functional magnetic resonance imaging (fMRI) in patients about to undergo CBT for PTSD. Blood oxygenation level-dependent (BOLD) signal was measured in 12 individuals with PTSD related to sexual trauma while they read words with positive, neutral, and negative content. Some negative words had PTSD-relevant themes while others did not. We hypothesized that PTSD-relevant words, would evoke emotional processes likely to be engaged by the CBT process, and would be most likely to activate brain circuitry important for CBT success. A group-level analysis showed that the rostral dorsomedial prefrontal cortex (rdmPFC) activated to a greater degree in response to PTSD-relevant words than other word types. This activation was strongest among patients with the best CBT response, particularly in the latter part of the task, when differences between individuals were most pronounced. We propose that the rdmPFC activation observed in this study reflects the engagement of neural processes involved in introspection and self-reflection. CBT may be more effective for individuals with a greater ability to engage these processes.

Introduction:

Post-traumatic stress disorder (PTSD) is characterized by an often disabling constellation of persistent symptoms following a traumatic life experience. Symptoms include intrusive re-experiencing of the trauma via flashbacks or nightmares, emotional dysregulation, hyperarousal, and avoidant behaviors. Trauma-focused cognitive behavioral therapy (CBT), a mainstay of PTSD treatment, often combines exposure therapy with cognitive therapy (1). The cognitive component aims to challenge maladaptive trauma-related appraisals and to teach patients to think more realistically about the trauma and their symptoms. Exposure therapy is based on the theory that avoidance of negative affective states interferes with extinction learning, a process mediated by interactions between the amygdala, dorsal anterior cingulate cortex (dACC) and ventromedial prefrontal cortex (vmPFC) whereby learned threat associations are updated and extinguished when the previously learned response is not reinforced (2). The cognitive component of trauma-focused CBT aims to challenge maladaptive trauma-related appraisals and to teach patients to think more realistically about the trauma and their symptoms. This presumably depends on prefrontal regions such as dorsolateral prefrontal cortex (dlPFC) and dorsomedial prefrontal cortex (dmPFC), which are involved in higher order cognitive operations (3).

Despite its apparent efficacy, response to trauma-focused CBT is incomplete, with as many as 60% of patients in some trials still meeting criteria for PTSD diagnosis following treatment (4). Skills training in affect and interpersonal regulation (STAIR) is a particular form of trauma-focused CBT with demonstrated benefit in PTSD. When a STAIR protocol was combined with standardized exposure therapy (STAIR/Exposure), full remission (CAPS < 20) was achieved in 27% of patients with 61% no longer meeting criteria for PTSD diagnosis compared to 33% of those treated with exposure therapy alone (5). Though STAIR/Exposure performed better than exposure therapy alone, over a third of patients still met criteria for PTSD following combined treatment. While most patients benefit from treatment, the degree of improvement is variable. A better understanding of who is likely to benefit from particular forms of treatment may help guide treatment selection for individual patients and can lead to improved understanding of the differential mechanisms by which these treatments work.

Functional neuroimaging studies have revealed differences in pre-treatment brain function that are associated with subsequent CBT response, though the findings may depend in part on the nature of the particular task used. Studies comparing pre-treatment neural responses to emotionally negative images vs. matched neutral or positive images have shown that greater activations of the amygdala (6, 7) and dACC (8, 9) correlate with a poorer response to treatment. However, greater dACC was associated with better treatment response when it was seen in response to anticipation of negative vs. positive images, in contrast to the opposite effect during image viewing (8). Fonzo and colleagues (7), found that while looking at faces depicting expressions associated with threat and fear as compared to neutral facial expressions, greater dACC activation (as well as greater activation of other frontal regions and anterior insula) was associated with a better treatment response—an effect apparently opposite that shown by Aupperle and colleagues. An opposite effect on amygdala activation to facial expressions associated with threat and fear compared with neutral expressions was also demonstrated in a population of adolescent girls with PTSD (10), where CBT non-responders demonstrated less differential amygdala activation, but it was noted that this reflected greater activation to neutral stimuli among the non-responders suggesting an overactive threat response to neutral stimuli. Studies utilizing behavioral inhibition tasks have shown that greater activation of dorsal striatum, pre-frontal networks (11) and inferior parietal lobule (12) were associated with better response to treatment. In general, these studies have been interpreted as demonstrating that greater activation of emotion-responsive regions prior to treatment indicates a poorer prognosis, while greater activation of brain areas involved in cognitive control and emotion regulation indicates a better prognosis.

All of these studies have utilized standard emotion or cognitive tasks to probe the relevant circuitry. To our knowledge, trauma-specific stimuli have not been used in the context of an activation task for the purposes of identifying neural circuitry associated with PTSD treatment response. By way of their semantic content, visually-presented trauma-specific words can convey themes that are highly personally and emotionally relevant to individuals with PTSD while controlling for low-level visual features. Trauma-related words can elicit early amygdala hyperactivation in PTSD patients to a degree not seen with nontrauma-related emotionally negative words (13). We hypothesized initially that amygdala activation would correlate negatively with treatment response while activation in higher order frontal lobe regions involved in emotion regulation such as vmPFC, dACC, and dlPFC would correlate positively with treatment response. Thus, we sought to investigate the relationship between trauma-related brain activation and subsequent response to trauma-focused CBT. Significant correlations in these specific brain regions were not found in our analyses. However, we conducted a whole brain analysis to address a more exploratory hypothesis—that brain activations specifically associated with reading trauma-related words, even when outside of regions typically associated with PTSD, can have relevance for how the brain processes trauma-related content during CBT and can therefore have bearing on CBT success. Certain neural responses to trauma-related content may facilitate better response to CBT while others may interfere. We therefore sought to identify trauma word-related brain activations that correlated with subsequent response to treatment.

Methods:

Participants:

Among 36 participants who enrolled in the study, 4 dropped out before starting treatment and 12 dropped out during treatment. Complete data sets including imaging pre- and post-treatment CAPS scores were available for 12 participants. These were all right-handed females with a mean age of 35.4 (range 23–48) who met DSMIV criteria for PTSD related to sexual/physical assault as their primary diagnosis. All participants were native English speakers, and were free of other psychiatric diagnoses, substance abuse, and significant neurological or medical disorders. The Clinician Administered PTSD Scale (CAPS) (14) was used to establish a diagnosis of PTSD. To meet criteria for diagnosis of PTSD, the participant must have had the following symptoms: for symptom B cluster (re-experiencing symptoms) at least one symptom with a frequency rating of 1 and intensity rating of 2; for symptom C cluster (avoidance and numbing symptoms) at least three symptoms; and for symptom D cluster (startle response/hyperarousal) at least two symptoms. All participants gave informed consent prior to participation in the study, which was part of a protocol approved by the Institutional Review Board at New York-Presbyterian Hospital/Weill Medical College of Cornell University. Data analyses and manuscript preparation for this protocol were approved by the Partners Human Research Committee. Participants were scored on the CAPS before and after CBT.

Cognitive Behavioral Therapy Protocol:

Following the fMRI scanning session, patients underwent a 16-session course of CBT with the STAIR/Exposure protocol, a CBT program specifically targeting PTSD related to childhood abuse. The treatment consisted of 8 one-hour weekly sessions on skills training in affect and interpersonal regulation (STAIR) followed by 8 one-hour bi-weekly exposure therapy sessions (5). The first four STAIR sessions focus on identifying and labeling feelings, emotion management, distress tolerance, and acceptance of feelings, while the next four focus on exploration and revision of maladaptive schemas, effective assertiveness, awareness of social context, and flexibility in interpersonal expectations and behavior. Exposure sessions involve reviewing trauma narratives. The protocol was flexibly applied, allowing up to 21 actual sessions to take into account circumstances such as crisis sessions and need to repeat sessions. The degree of treatment response was estimated by subtracting the post-CBT Total CAPS score from the pre-CBT total CAPS score such that a higher positive response score reflects a greater degree of symptom improvement. Normalized CAPS improvement indexes were calculated as (CAPS Total_Pre - CAPS Total_Post)/CAPS Total_Pre.

Functional MRI Scans:

Task-based fMRI Experiment: Emotional Word Paradigm

Before beginning CBT, fMRI scans were obtained while participants completed an emotional word task.

Stimuli consisted of 48 negative/anxiety words (24 PTSD, 24 panic), 48 neutral (NU), and 48 positive (PO) words, balanced across categories for frequency, length, and part of speech. The words used in this study were cultivated by our laboratory and based on a similar list of words that was piloted on 34 healthy volunteers who rated the three word categories (positive, negative, and neutral) and significantly different in valence. Posttraumatic stress disorder (PT) words were designed to be relevant to individuals with a history of physical/sexual trauma; panic (PA) words were designed to relate to panic attack symptoms and somatic/illness-related anxiety (a negative control condition). Words were selected for suitability for this task by a panel of three experienced clinicians.

The three valences of words were presented within a block design (six words per block, eight blocks per valence). Each word appeared for 2 seconds, followed by an inter-stimulus interval jittered around an average of 2.8 seconds, for a total of 28.8 seconds per block. During presentation of stimuli, subjects were instructed to read each word silently and to then immediately press a button under their right index finger. After the scan, participants were tested for their ability to recognize the words from amongst a list of similar distractors and to rate the valence of each word on a scale of −3 to +3. Recognition performance was quantified as a sensitivity index (d’). Ratings, recognition performance, and reaction times were compared across conditions using two-tailed Wilcoxon signed-rank tests. Further details of the task are presented in a previous publication (13).

fMRI Image Acquisition and Analysis

Details of the fMRI methods are provided in supplementary materials. Briefly, images were acquired with a research-dedicated GE Signa 3 Tesla MRI scanner (max gradient strength 40mT/m; max slew rate 150T/m/s). Blood Oxygenation Level-Dependent (BOLD) contrast imaging was employed using a gradient echo EPI sequence (TR=1200ms, TE=30ms, 15 or 21 slices of 5mm in thickness, 1mm gap, FoV=240mm, matrix=64×64). Functional image analysis was performed using customized fmristat software (15). A two-level voxel-wise linear fixed-effects model was utilized. Effects at every brain voxel were estimated using the EM (expectation maximization) algorithm, and effects of word type were then compared using linear contrasts. At the group level, the within-group pre-treatment effects of the hypothesis-driven contrasts (PTvNU, PTvPA, PTvPO) were examined for their association with the normalized CAPS improvement index via a multiple regression model, with the CAPS improvement index as the main regressor, and age and scanning protocol as a covariates of no-interest. To examine the effects of experiment time, we also tested the effects of each half of the experiment via linear contrasts within the voxel-wise general linear model. Maps of the t-statistic were thresholded initially at a voxel-wise two-tailed p-value < 0.01 and a cluster spatial extent > 250 mm³. The p-values at the peak voxels were then corrected for multiple comparisons using random field theory, based on family-wise error rate over the whole brain at p < 0.05. The group-level correlation effect of interest at a peak coordinate was considered significant if the corrected p-value (p_corr) was less than 0.05 based on family-wise error rate correction of the voxel-wise p-values over the entire brain.

Results

Behavioral Results:

Mean total CAPS score prior to treatment was 63.2 (range 46–84). Following treatment, mean total CAPS score was 36.9 (range 9–54). The mean reduction in CAPS score following treatment (total CAPS post - total CAPS pre) was 26.3 (range 1–57). There were no significant differences in reaction time between any pair of word types. Mean reaction time ranged from 448 ms to 1586 ms with a grand mean of 950 ms. Following the scan, subjects were asked to rate the words for valence on a scale of −3 to +3. Rating data were available for 11 subjects. PT words were rated significantly more negative than all other word types (PT word mean −2.27+0.45; NU word mean 0.28+0.33; PA word mean −1.84+0.52; PO word mean 2.03+0.32; p<0.001 for all three comparisons). Memory for the words viewed in the scanner was assessed by computing a d’ value comparing hits with false positives during a post-scan recognition test. Subjects exhibited significantly better recall for PT words than NU words (mean PT d’=1.98+0.49; mean NU d’=1.46+0.58; p=0.002). Recall was also significantly better for PT words than for PO words (mean PO d’=1.30+0.70; p<0.001), but the difference between PT words and PA words was not significant (mean PA d’=1.75+0.92; p=0.13).

Imaging Results:

At the group level, responses to PTSD words (PT) were compared against responses to panic disorder words (PA), emotionally positive words (PO), and emotionally neutral words (NU). PT words activated the rostral dorsomedial prefrontal cortex (rdmPFC) more than PA, NU or PO words (Figure 1). The differences between PT and NU were more pronounced than the differences between PT and the other emotion conditions (PA, PO). PT words also activated a network of thalamic, basal ganglia and midbrain regions more than PA, NU, and PA words. The left amygdala was more activated during the PT condition than during the NU condition, but not more active than during the PA or PO conditions. (See supplementary tables 1−3 for full results).

Figure 1.

PT v NU t-scores displayed with initial voxel-wise threshold of p<0.001. The rdmPFC activates more strongly in the PT condition than in NU, PA, or PO conditions.

Correlations were calculated for each voxel for the PTvNU effect against normalized treatment response (Table 1). The most significant positive correlation with treatment response was in the rdmPFC (left: p_corr = 0.002; right: p_corr = 0.015). This correlation reflects greater activation in response to PT words relative to NU words in patients with the greatest symptomatic improvement in response to treatment (Figure 2A,B). This effect is primarily driven by the second half of the experiment (Figure 2C). During the first half of the task, even the patients with poorer response to treatment demonstrated a differential activation in this region, but these patients did not maintain this differential throughout the experiment the way those with better treatment response did.

TABLE 1.

Gray matter regions in which differential activation between PT and NU conditions correlates with normalized CAPS improvement index (p<0.05 corrected for whole brain volume, cluster extent threshold > 250 mm³)

Positive Correlations	Brain Region		Brodmann Area	MNI Coordinates			z-value	p-value	Corrected p-value	Cluster	Cluster Size (mm3)
				X	Y	Z
	L	Medial Frontal Gyrus	8	−6	54	33	4.887	<0.001	0.002	1	8127
		Superior Frontal Gyrus	8	0	45	45	4.707	<0.001	0.004	1
	L	Medial Frontal Gyrus	8	−3	57	33	4.702	<0.001	0.005	1
		Medial Frontal Gyrus	9	0	60	30	4.659	<0.001	0.005	1
	L	Superior Frontal Gyrus	8	−9	54	42	4.444	<0.001	0.013	1
	L	Superior Frontal Gyrus	6	−3	21	63	5.093	<0.001	0.001	2	2943
	R	Superior Frontal Gyrus	6	9	24	66	4.415	<0.001	0.015	2
	L	Inferior Semi-Lunar Lobule		−9	−72	−39	4.546	<0.001	0.009	3	1188
	L	Cerebellar Uvula		−6	−69	−36	4.507	<0.001	0.01	3
	R	Cerebellar Tonsil		48	−51	−36	4.185	<0.001	0.036	4	891
	L	Lingual Gyrus		−30	−78	6	4.635	<0.001	0.006	5	1188
	L	Cuneus	18	−6	−81	30	4.549	<0.001	0.009	6	1431
	R	Medial Frontal Gyrus	9	12	54	18	4.258	<0.001	0.027	7	756
	R	Middle Frontal Gyrus	10	36	60	−21	4.248	<0.001	0.028	8	459
	L	Cerebellar Tonsil		−30	−33	−45	4.392	<0.001	0.016	10	459
Negative Correlations
	L	Superior Parietal Lobule	7	−45	−57	60	5.516	<0.001	<0.001	1	6885
	L	Inferior Parietal Lobule	40	−45	−54	57	5.444	<0.001	<0.001	1
	L	Supramarginal Gyrus	40	−48	−42	39	4.192	<0.001	0.035	1
	R	Inferior Occipital Gyrus	18	51	−84	−6	4.494	<0.001	0.011	2	2808
	R	Superior Parietal Lobule	7	45	−54	57	4.947	<0.001	0.002	3	4077
	R	Postcentral Gyrus	5	51	−42	69	4.44	<0.001	0.013	3
	R	Inferior Frontal Gyrus	45	72	27	12	5.257	<0.001	<0.001	4
	R	Inferior Frontal Gyrus	46	63	36	6	4.426	<0.001	0.014	4
	R	Cuneus	18	9	−93	21	4.94	<0.001	0.002	5	3537
	R	Cuneus	17	9	−96	9	4.617	<0.001	0.006	5
	R	Superior Frontal Gyrus	6	30	6	72	4.453	<0.001	0.013	6	1890
	L	Postcentral Gyrus	7	−15	−48	78	4.413	<0.001	0.015	7	3240
	L	Middle Frontal Gyrus	46	−57	36	24	4.524	<0.001	0.01	8	2538
	R	Superior Temporal Gyrus	38	33	9	−42	4.275	<0.001	0.025	9	1377
	L	Inferior Temporal Gyrus	37	−69	−51	−9	4.794	<0.001	0.003	10	1485
	R	Superior Parietal Lobule	7	36	−75	57	4.217	<0.001	0.032	11	1890
	R	Inferior Frontal Gyrus	47	45	24	−30	4.291	<0.001	0.024	13	675
	L	Superior Parietal Lobule	7	−12	−69	66	4.418	<0.001	0.015	14	918
	L	Precuneus	7	−15	−72	66	4.299	<0.001	0.023	14
	L	Superior Frontal Gyrus	6	−12	−9	78	4.119	<0.001	0.045	18	702
	R	Supramarginal Gyrus	40	42	−42	36	4.47	<0.001	0.012	19	567

Figure 2.

PT v NU correlation with treatment response. (A) Sagittal slice (x=−6) showing t-statistic displayed with initial voxel-wise threshold of p<0.001. Warm colors indicate positive correlation while cool colors indicate negative correlation. Scatter plots (B and C) depict the correlations at the peak voxel for the rdmPFC cluster illustrated with the circle in A. Panel C depicts the correlations separately for the early and late portion of the experiment, demonstrating a stronger correlation during the late part of the experiment. The x-axis depicts the normalized treatment response with higher numbers indicating a greater symptomatic improvement.

A trend toward positive correlation was seen in the dACC and negative correlation in the ventromedial prefrontal cortex (vmPFC) neither of which survived whole brain correction. This effect was driven by the early part of the experiment.

Discussion:

While the dACC (among other regions) has been previously highlighted in the PTSD treatment response literature, activations in the dmPFC have received little attention as mediators of CBT response. Here, we show that stronger activation of the rostral dmPFC in response to the reading of trauma words prior to treatment with CBT correlates with better response to treatment.

The portions of the medial prefrontal cortex dorsal and rostral to the anterior cingulate cortex consisting of the medial aspects of Brodmann areas 6, 8, 9 and 10 are often referred to as dorsomedial prefrontal cortex or medial frontal gyrus. The dmPFC can be further subdivided into regions with distinct functional associations. The most posterior portions of the dmPFC are concerned with motor functions and include the supplementary motor area (SMA) and pre-SMA. Anterior to this region is the mid-dmPFC, which is involved in internal monitoring of action, is particularly sensitive to error or conflict, and is involved in the regulation of behavior by monitoring the value of potential future actions (16). Our results point toward a region of dmPFC anterior to the mid-dmPFC referred to here as rostral-dmPFC (rdmPFC), which encompasses the medial portions of Brodmann areas 9 and 10 and is most closely linked with social cognition. This area has not been closely associated with emotion down-regulation as have other areas of the PFC such as the vmPFC, which makes strong connections with amygdala and other limbic regions and is implicated in emotion regulation, particularly in the context of extinction of a learned threat response (2, 17).

Though there are few direct connections between dmPFC and amygdala, dmPFC has been invoked (along with other brain areas such as dorsolateral PFC and dACC) in reappraisal, a method of emotion-regulation whereby a deliberate effort is made to alter the emotional interpretation of an emotionally-laden stimulus. However, meta-analyses have suggested that it is primarily mid-dmPFC rather than rdmPFC that subserves this function (18–21). Why, then, might activation of the rdmPFC in response to reading trauma-related words correlate with a better response to subsequent CBT? We propose that differences in the degree of activation of this region among PTSD patients may reflect an individual’s capacity for introspection and emotional self-awareness, psychological characteristics likely to mediate treatment response.

Among all PTSD patients in our sample about to undergo CBT, the rdmPFC activated to trauma words but not to neutral words at the group level. Positive words and panic words activated this region more than neutral words but less than trauma words. Taken together, these findings suggest that the rdmPFC may respond more strongly to personally relevant emotion stimuli. Although the stimuli were standardized across participants and were not specific to an individual’s personal story, they were chosen to reflect the experience and pre-occupations of patients with PTSD related to sexual trauma, a history common to all of the participants in this study, and the panic words would be expected to be less personally relevant to this population. The particular relevance of the PTSD words may evoke stronger self-reflection as compared to the other categories of stimuli.

Prior literature has demonstrated that the rdmPFC is activated by a variety of types of emotion and non-emotion tasks that involve mentalizing, or the meta-representation of one’s own or another’s mental processes or attributes (16). For example, while viewing affectively charged photos, rdmPFC was activated when individuals reflected on their own feelings or on the feelings of the central character in the photo but not when asked to make a determination about whether the same photos were taken outdoors or indoors (22, 23). In a task in which subjects were asked to either up-regulate or down-regulate their negative emotions in response to aversive images, rdmPFC activated with up-regulation but not down-regulation (24) again arguing against a direct role for this region in down-regulation of negative emotion. The activation of the rdmPFC in response to up-regulation of negative emotion in that study may have been due to the “self-focused” method by which half of subjects were taught to increase negative emotion—they were instructed to imagine themselves in the negative situation depicted in the picture. By contrast, the method by which they were taught to decrease negative emotion involved viewing the image from a more detached perspective in which self-referential processes are expected to be less engaged. Thus, the effect of that task on the rdmPFC may have related more to the differing degree of self-examination between the two tasks rather than on the directionality of emotional self-regulation The rdmPFC demonstrates greater activity when individuals make judgments about self-attributes (mental or otherwise) as compared to impersonal semantic judgments (25) or as compared to judgments about the accuracy of impersonal factual statements (26) further supporting a role for this region in self-referential processing. The rdmPFC also tends to be relatively active during the resting state (27, 28), a finding that has been attributed to the types of self-generated mental activities that occur spontaneously under unconstrained “resting” conditions, many of which are likely introspective or self-referential (29). Whitfield-Gabrieli (30) found an overlap in the rdmPFC between the self-referential network and the resting-state default-mode network.

The rdmPFC is also activated by tasks that require participants to make inferences about the attributes (25) and mental states (31, 32) of other people and even of dogs (33), suggesting a role for this region in social cognition. Taken together, the prior literature suggests that the rdmPFC is involved with meta-representation of one’s own mind or the minds of others and supports cognitive functions such as self-examination, introspection, and awareness of one’s emotional state. In fact, patients with alexithymia, a condition characterized by difficulty identifying and describing one’s own feelings and emotions, demonstrated reduced activation of the rdmPFC in response to a theory-of-mind task (34).

The STAIR treatment module, which precedes the exposure module, aims to help patients become more aware of their emotional responses during emotional events and recall of traumatic memories (35). During the STAIR phase of therapy, patients practice skills including identifying and labeling of feeling states and identifying and altering interpersonal schemas (36) to enhance capacity to use exposure therapy. It is possible, based on the evidence reviewed above, that the rdmPFC supports an individual’s capacity to practice such skills. Some research indicates that treatments that capitalize on patients’ strengths produce better outcomes (37). Thus, patients who already have the ability to reflect on their emotions may benefit more from therapies that include self-reflective exercises, and this may be an avenue for future investigation. It remains to be seen whether individuals who have less reflective capacity can benefit more from a therapy that teaches them these skills as opposed to therapies that utilize more concrete behavioral strategies that might be more consistent with current skills or preferences.

Individuals differ in their ability to reflect upon their own emotions, and these intrinsic characteristics could inform the type of psychotherapy recommended. We might predict that these differences in abilities would be supported by differences in the mediating functional neuroanatomy under relevant probe conditions. Ultimately, it is hoped that functional neuroimaging at the individual level, along with psychological profiling, will be clinically useful for the design or selection of an individualized psychotherapy program.

Previously, we found that amygdala BOLD responses to these linguistic stimuli in PTSD patients compared with healthy controls change from the early part of the task to the late part of the task, highlighting variable degrees of habituation or sensitization in patients and healthy controls depending on the stimulus type (13). Since neural processes engaged by the stimuli change over time as the subject gains experience with the task and as the stimulus effects build, examining effects of experiment time on the BOLD responses can provide further insight into the nature of these processes. Interestingly, an examination of the effects of experiment time on rdmPFC activation reveals that the correlation with treatment response is largely driven by the second half of the task. During the first half of the task, even the poor responders activated rdmPFC, but they were less able than the stronger responders to sustain this response during the latter part of the task such that rdmPFC activation during the late portion of the task demonstrated a stronger correlation with treatment response. This finding could potentially be explained by a tendency of the poorer responders (but not the stronger responders) to fail to sustain their emotional engagement and self-reflection throughout the scanning session, either due to fatigue or a tendency to dissociate under the stress of the repeated exposure to emotionally distressing stimuli.

Limitations of this study include the small sample size and the fixed effects analysis, which may limit the generalizability of the findings beyond the particular group studied. In particular, this study included only female participants with a PTSD related to sexual/physical assault, and it is unclear whether these findings will generalize to men or to PTSD related to other types of trauma. The lack of a control group who did not undergo CBT makes it difficult to demonstrate the extent to which improvements in symptom scores were due to treatment, but the efficacy of STAIR when combined with exposure therapy has been previously shown to be effective compared to exposure without STAIR in a randomized controlled trial (5). Though we surmise that the rdmPFC effect can be explained by an intrinsic capacity among the strong responders toward introspection and self-reflection in the context of negative emotion, we did not assess this behavioral characteristic in our participants. Future studies may utilize scales such as the self-consciousness scale (38), the self-reflection and insight scale (39) or the Toronto Alexithymia Scale (TAS) (40), to examine the extent to which these behavioral measures correlate with treatment response and/or rdmPFC activation.

Conclusions:

In summary, individuals with PTSD related to sexual trauma demonstrate differential activation of the rdmPFC while reading words related to PTSD and sexual trauma, and greater activation of this region prior to treatment is associated with a better response to CBT with STAIR/exposure therapy. We propose that rdmPFC activation during this task may be driven by self-referential thought and that the relationship between activation of this region and response to treatment may reflect a greater capacity to benefit from STAIR among individuals with intrinsically greater ability to mobilize neural resources for self-examination.