Psychophysiological Response to Virtual Reality and Sub-Threshold PTSD Symptoms in Recently

Psychophysiological Response to Virtual Reality and Sub-threshold PTSD Symptoms in Recently Deployed Military

ABSTRACT

Subthreshold posttraumatic stress disorder (PTSD) has garnered recent attention due to the significant distress and functional impairment associated with the symptoms as well as the increased likelihood of developing full PTSD. Unfortunately, it is not uncommon for individuals to under- or over-report symptoms, so an independent method for identifying those with significant subthreshold symptoms would be useful. The purpose of this report is to document the utility of a virtual combat environment in eliciting distinctive psychophysiologic responses associated with PTSD symptoms in a population of recently deployed military service members who do not meet full criteria for PTSD. Among a range of psychophysiologic measures that were studied, regression analysis revealed heart rate and galvanic skin response as the best predictors for each of the three symptom clusters--hyperarousal, re-experiencing and avoidance—as well as for global PTSD symptoms. We believe this represents the first report to document the utility of a virtual reality environment in the identification of subthreshold PTSD.

Introduction

Posttraumatic stress disorder (PTSD) is frequently a disabling and chronic condition that has been reported in 5-17% of service members after deployment to Afghanistan or Iraq [1] . PTSD is characterized by three clusters of symptoms including reexperiencing (cluster B), avoidance and numbing (cluster C) and hyperarousal (cluster D) [2]. In addition to reacting to a traumatic event with intense fear, horror, or helplessness (criterion A), the diagnosis of PTSD requires the presence of at least one of five cluster B symptoms, three of seven cluster C symptoms, and two of five cluster D symptoms that persist for at least one month and are associated with functional impairment [2]. There is evidence that those with subthreshold PTSD, while lacking a sufficient number or distribution of symptoms to meet full criteria do nevertheless have clinically significant functional impairment [3, 4].

Subthreshold PTSD, sometimes referred to as partial or subsyndromal PTSD [5], is not yet well-defined (see Table 1). Nevertheless, several studies have documented that the degree of functional impairment associated with subthreshold PTSD is comparable to that of full-blown PTSD [5-13], and one in four with subthreshold PTSD symptoms go on to develop full PTSD during longitudinal follow up [5, 14, 15]. Furthermore, recent evidence that low-intensity treatment may be sufficiently effective for subthreshold PTSD, but not full PTSD [16], underscores the potential value to screening for subthreshold symptoms. Service members who have recently returned from the wars in Iraq and Afghanistan represent an important population to study, because of the high prevalence, evidence of PTSD-related impaired social functioning [17, 18] and difficulties in postdeployment adjustment [4, 19]. The specific aim of this report is to document a novel method for identifying subthreshold PTSD symptoms by measuring psychophysiologic responses to virtual representations of Iraq and Afghanistan among active and reserve component U.S. military service members recently returned from deployment.

In addition to looking at subthreshold PTSD symptoms globally, we also examine the relationship between psychophysiologic reactions and the three distinct clusters of PTSD symptoms. We hypothesize that hyperarousal (Criterion D) would have the strongest correlation with psychophysiologic responses, which is relevant because recent evidence suggests that it the severity of the Cluster D symptoms that best predicts the severity of the avoidance and re-experiencing symptoms [20-22]. There is some evidence that hyperarousal may even lead directly to the emotional numbing of PTSD [21, 23]. Hyperarousal is readily detectable physiologically; there is good evidence that facial electromyography (EMG), heart rate (HR), skin conductance (SC), and blood pressure all distinguish PTSD-related symptoms [24], providing a strong rationale for conducting physiological assessment during a stressful challenge to identify subthreshold symptoms. Moreover, it stands to reason that a stressful challenge, such as combat-related stimuli, may be especially potent in identifying subthreshold symptoms as well as the vulnerability to developing combat-related PTSD in military service members.

VR has shown success in the treatment of PTSD [25-28], but scant attention has been given to the potential of VR as a means for early detection of those at risk for PTSD. We used three unique combated-related VR scenarios taken from Virtual Iraq as a novel means of assessing subthreshold symptoms in combat veterans recently returned from Iraq or Afghanistan, with the expectation that this virtual combat environment would be particularly useful in detecting physiological changes in this population.

Methods

Participants

The sample consisted of 78 SMs (11 women) with a mean age of 29.72 (SD 7.91), range = 19 – 51 within two months after return from a deployment to either Iraq or Afghanistan. All participants completed an initial brief in-person or phone screen after which 85 service members were invited and presented for the baseline assessment. Four of those were excluded because they already met criteria for PTSD (2) or Major Depressive Disorder (2) and three were excluded in the analysis due to recording problems with the physiological data.

Procedures

Data were derived from a larger study exploring a variety of early predictors of PTSD, Major Depressive Disorder (MDD) and Post-Concussive Syndrome with the ultimate goal of developing a risk stratification model. This paper represents the initial investigation of physiologic responses as early predictors of PTSD which may ultimately be part of the model that will be developed after completing the study. Data for this paper were obtained during baseline assessments though the study is ongoing with follow-up assessments currently being conducted. The overall study was approved by institutional review boards at Walter Reed National Military Medical Center, Uniformed Services University, and the National Institutes of Health.

All participants completed a comprehensive 2-day baseline assessment at the National Intrepid Center of Excellence in Bethesda, Maryland. The principal investigator, a board-certified internist, obtained written informed consent from each participant and then performed a medical history and physical examination, as well as a detailed assessment of vestibular and olfactory function. An experienced, licensed psychologist assessed each participant with the Clinician-Administered PTSD Scale (CAPS), considered the gold standard method for diagnosing PTSD [29]. Each participant also completed the self-administered PTSD Checklist-Military Version (PCL-M) [30] and the Patient Health Questionnaire (PHQ) [31], a screen for depression and other psychiatric conditions. To be included in the study, participants had to score less than 10 on the PHQ-9 and less than 50 on the PCL-M to rule out depression and PTSD, respectively; the CAPS was used as a confirmatory measure for the diagnosis of PTSD, to ensure that even those not achieving a PCL-M score of 50 did not have a sufficient number and severity of symptoms to be diagnosed with PTSD. The physiological assessment employed three 2-minute sequences taken from a Virtual Iraq/Afghanistan environment that we previously used to treat PTSD [32]. Two of the sequences were presented from the perspective of a HUMVEE in a convoy confronted with improvised explosive devices (IEDs) and ambushes (convoy 1, gunner position and convoy 2, cabin of a humvee), while the third conveyed a foot patrol through a village marketplace replete with explosions and terrorists firing rocket-propelled grenades (city). The sequences were presented on a computer screen providing visual and auditory stimulation only, without the tactile, olfactory, or even the full immersion effect provided by a head-mounted display as we have used in conducting virtual reality exposure therapy [32]. Each sequence was separated by 30 seconds of a blue screen baseline (Figure 1).

Psychophysiological data was collected using Biopac MP150 for Windows (Biopac Systems, Inc., Aero Camino, CA). We recorded electromyographic (EMG), galvanic skin response (GSR) electrocardiogram (ECG), and respiratory rate (RR). EMG activity was recorded with electrodes placed over the orbicularis oculi muscle; GSR from the palmar surface of the non-dominant hand; ECG from electrodes on the chest, below the right clavicle and the other inside of the left arm; and RR with a chest band across the sternum. All data were sampled at 1000 Hz, digitized at 16 bit A/D resolution, amplified and processed with the Biopac system. Blood pressure (diastolic blood pressure (DBP) and systolic blood pressure (SBP)) was measure with a Coulbourn System Strain Gage Bridge Transducer Coupler.

EMG was band-pass filtered from 28-500Hz and rectified [33]. EMG peak amplitude 20-200 ms after the startle probe offset [33] was used to compute difference scores for each startle probe, by using the startle magnitude in the presence of VR minus the startle magnitude to the noise alone during the baseline period [34]. GSR was 1 Hz low-pass filtered and a 0.05 Hz high-pass filtered. ECG was band-pass filtered from 0.5-35Hz and was converted to heart rate in beats per minute. RR was converted into breaths per minute. During each VR session, mean HR, RR, GSR and EMG (difference score) were computed during the 2 minute viewing period [35] and BP was measured following the offset of the viewing period. All data were exported into SPSS for statistical analysis.

Given the range of definitions of subthreshold PTSD, we executed a series of linear regression analyses to look at the continuous relationship between symptom clusters and physiological variables. ANOVA was first used to examine if a particular VR session (Convoy 1, Convoy 2, City) would result in significantly different physiological responses. Based on the ANOVA findings, a series of multiple regressions was performed to investigate if cluster B (reexperiencing), cluster C (avoidance), cluster D (hyperarousal) and CAPS total score were predicted by the psychophysiological variables from each period of VR. The independent variables for the multiple regression analyses included the RR, EMG, HR, GSR, DBP, SBP. Step-wise regression analysis using backward elimination was also employed for those variables that were significant predictors (from the linear regression) to examine any differences in the predictive physiological variables, given that a data driven approach is informative with this novel paradigm.

Results

ANOVA

Separate one-way analysis of variance (ANOVA) for each physiological measure was used to examine significant effects across the VR sessions in order to inform the regression analysis. Violations of sphericity were examined using Mauchly’s Test and F-values corrected Greenhouse–Geisser adjustment were used if needed. Post hoc analysis was executed using Tukey’s LSD. ANOVA for GSR showed a main effect for VR Session F (1,78)= 58.74, p<0.001. Post hoc analysis revealed significant differences between all three VR sessions: Convoy 1 (CV1) 7.771, SEM= .565, Convoy 2 (CV2) 6.591, SEM=.505, City 6.367, SEM=.520. ANOVA for HR also resulted in a main effect for VR Session F(1,78) = 5.482, p = 0.006. Post hoc analysis revealed significant differences between CV1 67.33, SEM=1.192 and CV2 68.293, SEM 1.153, and CV1 and City 68.471, SEM=1.342. ANOVA did not reveal a main effect for VR Session for EMG F(1,77) = .743, p = .421, RR F(1,77)= .737, p=.473, DBP F(1,70)= .206, p=.651 nor SBP F(1,72)= .561, p=.456.

Linear Regression

Linear regression analysis revealed that GSR across all VR sessions significantly explained the variation in Cluster D F(3,74)=3.452 p=0.021, R2=.123 and Cluster C F(3,74)=2.629, p=0.05, R2=.096 and CAPS Total F(4,74)=3.024, p=0.035, R2=.33, but not Cluster B symptoms. Regression analysis also demonstrated that HR significantly explained the variance of Cluster D F(3,74)=2.737, p=0.049, R2=.316 and CAPS Total F(3.74)=4.273, p=0.008, R2=.384, but not Cluster B or C. EMG, RR, SBP and DBP all failed to significantly enhance an explanation of the variance observed in Cluster B, C, or D, or CAPS Total.

Step-wise Regression (Backward Elimination)

In order to test which physiological measures were the best independent predictors of variance for each of the three symptom clusters as well as the CAPS Total Score, a stepwise regression analysis (backwards elimination) was used, constrained to GSR and HR, since these variables were significant predictors in the linear regression. The best model for CAPS total included GSR during all three VR sessions, along with HR for the two convoy sessions (CV1 and CV2) F(5,72)=3.717, p=0.005, R2=.205. On the other hand, HR across all three sessions provided the best model for Cluster B symptoms F(3, 74)=7.391, p<0.001, R2=.231. The final model for Cluster C approached—but did not achieve—significance, utilizing HR during City and CV2 and GSR during City and CV2 F(4, 73)=2.36, p=0.061, R2=.114. Finally, GSR during all three sessions and HR during CV1 and CV2 remained significant predictors in the final model of Cluster D, F(5, 72)=3.685, p=0.005, R2=.204 (see Table 2).

Discussion

Virtual reality appears to be a valuable method for treating behavioral disorders, including as an attractive and effective adjunct to providing exposure therapy for PTSD [25-28]. This novel technology also has the potential to elicit significant physiological responses and may ultimately be able to function as a tool for assessing PTSD symptoms. Virtual reality may be more palatable than the frank discomfort of other psychophysiologic paradigms such as electric shocks or airblasts to the larynx, attracting rather than repelling potential study participants. To the best of our knowledge, this is the first report to document that virtual reality can in fact be used to distinguish subthreshold PTSD symptoms, which are in turn linked to significant functional impairment [5-13] and risk for developing full PTSD [5, 14, 15].

We identified physiologic responses to the virtual sequences that were significantly associated with the presence of subthreshold PTSD symptoms in U.S. military service members within 2 months after their return from Iraq or Afghanistan. Heart rate (HR) and galvanic skin response (GSR) were consistently the best predictors of subthreshold PTSD symptoms in their entirety as well as within symptom clusters, across a range of VR sequences. The other physiological measures we obtained, including EMG startle, RR, and BP, added nothing to our modeling of PTSD symptoms. Following linear regression to identify predictors, stepwise multiple regression confirmed that physiologic responses to VR sequences were best at distinguishing hyperarousal symptoms (HR and GSR, 20% of the variance) and re-experiencing (HR, 23%), but were less potent with regard to avoidance symptoms (GSR, 11%). These measures (GSR and HR) also explained approximately 21% of the variance in global PTSD symptoms as measured by the gold-standard CAPS. The modest nature of this predictive power suggests that physiologic responses to a virtual combat environment should not be used in isolation to ascertain the significance of PTSD symptoms, but they can serve as a valuable element of such efforts. As indicated above, the results that we report here are part of the baseline assessment of a cohort we have assembled with whom we are continuing to conduct follow up assessments. It is likely that the physiologic responses to the virtual environment will represent only part of the risk stratification model we develop to predict who go on to develop PTSD, major depression, or postconcussive syndrome.