Sweat pore reactivity as a surrogate measure of sympathetic nervous system activity in trauma-exposed individuals with and without posttraumatic stress disorder

Abstract

Stress analysis by FLIR (forward-looking infrared) evaluation (SAFE) has been demonstrated to monitor sweat pore activation (SPA) as a novel surrogate measure of sympathetic nervous system (SNS) activity in a normal population. SNS responses to a series of 15 1-s, 82 dB, white noise bursts were measured by skin conductance (SC) and SAFE monitoring of SPA on the fingers (FiP) and face (FaP) in 10 participants with posttraumatic stress disorder (PTSD) and 16 trauma-exposed participants without PTSD (M_age = 48.92 ± 12.00 years; 26.9% female). Within participants, SC and FiP responses across trials were strongly correlated (r = .92, p < .001). Correlations between SC and FaP (r = .76, p = .001) and between FiP and FaP (r = .47, p = .005) were smaller. The habituation of SNS responses across the 15 trials was substantial (SC: d = −2.97; FiP: d = −2.34; FaP: d = −1.02). There was a strong correlation between habituation effects for SC and FiP (r = .76, p < .001), but not for SC and FaP (r = .15, p = .45) or FiP and FaP (r = .29, p = .16). Participants with PTSD showed larger SNS responses to the first loud noise than those without PTSD. PTSD reexperiencing symptoms assessed by the PTSD Checklist on the day of testing were associated with the SNS responses to the first loud noise measured by SC (d = 1.19) and FiP (d = .99), but not FaP (d = .10). This study confirms convergence of SAFE and SC as valid measures of SNS activity. SAFE FiP and SC responses were highly predictive of self-rated PTSD reexperiencing symptoms. SAFE may offer an attractive alternative for applications in PTSD and similar populations.

Measurement of sympathetic nervous system (SNS) activity is important for a number of conditions, including posttraumatic stress disorder (PTSD). Evolutionary adaptation of the SNS has helped humans survive threats, real and perceived, great and small, ranging from the saber-toothed tiger to hallway bullies in elementary school (Fanselow, 1984; Lazarus, 1966; Lazarus & Folkman, 1984). However, this survival mechanism is overexpressed and poorly moderated in persons with PTSD (e.g., Pitman et al., 2012). Therefore, assessing SNS activity through surrogate measures may provide clues to conditioned threat cues or other factors that trigger and sustain PTSD, whether or not the individual is aware of the trigger, intent on avoiding its acknowledgment, or reluctant to acknowledge his or her struggle with PTSD.

Many of the methods available for measuring change in SNS activity require contact with the subject such as when collecting blood or saliva samples, or the attachment of sensors to monitor blood pressure, skin conductance, cardiac activity, respiration, perspiration, eye-blink, or orbicularis oculi electromyography. For individuals with PTSD, in whom heightened arousal (hypervigilance, irritability, and emotional or physiological responses to conditioned cues) may be triggered by novel situations or threat cues, and be already exaggerated at baseline, such methods may be suboptimal. Measurement of skin conductance (SC) activity, which is related to the number of active eccrine sweat glands, has long been established and widely accepted as a useful psychophysiological measure of SNS (Freedman et al., 1994; Harris, Polk, & Willis, 1972; Juniper, Blanton, & Dykman, 1967), but again with the aforementioned shortcomings.

A number of noncontact measures of SNS activity have been developed. They include measures based on ocular variables such as saccade, slow eye movement, pupil dilation, eyeblink, or eyelid closure (Abe et al., 2011; Bernard, Deuter, Gemmar, & Schachinger, 2013), and on thermal imaging (Di Giacinto, Brunetti, Sepede, Ferretti, & Merla, 2014; Shastri, Merla, Tsiamyrtzis, & Pavlidis, 2009; Engert et al., 2014; Fei & Pavlidis, 2010; Pavlidis et al., 2012; Shastri, Papadakis, Tsiamyrtzis, Bass, & Pavlidis, 2012). Poh, McDuff, and Picard (2011) reported the use of a web camera to assess multiple physiological parameters, an important development in noncontact SNS measurement. Camera images of human faces illuminated by ambient sunlight were used to extract measures of heart rate and heart rate variability. Their hardware, the ubiquitous laptop webcam, was comparatively simple. However, the protocol required participants to remain still for a 1-min measurement. Measurement of ocular variables in eye movement studies also require participants to remain still (Abe et al., 2011). Such procedures are not practical for psychophysiological measurement of discrete SNS responses, as may be desired in investigations of SNS responses to trauma reminders or other variably timed, amygdala-activating stimuli in patients with PTSD.

Researchers at the U.S. Army’s Night Vision and Electronic Sensors Directorate have developed a noncontact, passive method of quantifying SNS reactivity via high resolution, forward-looking infrared (FLIR) thermal imaging of sweat pore openings. Using this method, Krzywicki, Berntson, and O’Kane (2014) demonstrated that FLIR imaging of sweat pores on the finger (FiP) of healthy subjects (i.e., with no mental health diagnoses) was highly correlated with SC responses generated by using a breathing manipulation.

Stress analysis by FLIR evaluation (SAFE), an adaptation of the system described by Krzywicki et al. (2014), passively records thermal images of sweat pore openings on the face and fingers, and quantifies these openings within a defined area of skin during exposure to stressful stimuli. This stress-sensing procedure is unique in that it employs counts of the transiently opened sweat pores, rather than temperature change, as a surrogate for SNS reactivity.

The primary aim of the current study was to examine the convergent validity of SAFE and SC measures of SNS reactivity in trauma-exposed individuals with and without PTSD, during exposure to a series of brief, loud noise stimuli. Given previous findings of an association between PTSD and a variety of SNS responses to a similar loud tone task (Metzger et al., 1999; Orr et al., 2003; Shalev et al., 1992), a secondary aim of the study was to evaluate the ability of SAFE to discriminate individuals with current PTSD from trauma-exposed individuals without PTSD. Finally, we explored potential effects of dependent smoking on SAFE and SC responses to the loud noise stimuli. Given the neuroendocrine impact of acute and dependent smoking (Rasmusson et al., 2006), we hypothesized that recent smoking would suppress SNS reactivity and, consequently, should be considered as a potential confound in the psychophysiological assessment of SNS reactivity in trauma-exposed individuals with and without PTSD.

Method

The study procedures were approved by the Institutional Review Boards at the Veterans Affairs Boston Healthcare System (VABHS) and the Office of Research Protections, Human Research Protection Office of the U.S. Army Medical Research and Materiel Command. The study was conducted at the National Center for PTSD, Women’s Health Science Division, VABHS.

Screening Procedures

Participants were recruited through advertisements, including flyers placed on VA bulletin boards and notices placed online (e.g., Craigslist). Potential participants completed an initial telephone-screening interview to establish preliminary eligibility. At the subsequent in-person evaluation, participants engaged in the informed consent process before completing a general screening for physical health (e.g., medical history, current medications, menstrual cycle functioning), a diagnostic interview (administered by trained doctoral level psychologists), and self-report questionnaires.

The psychiatric evaluation included assessment of Axis I psychopathology using the Structured Clinical Interview for DSM-IV (SCID-IV; First, Spitzer, Gibbon, & Williams, 1995). Trauma exposure was assessed with the Trauma Life Events Questionnaire (Kubany et al., 2000). The Clinician-Administered PTSD Scale (CAPS), Version IV (Blake et al., 1995) was administered to confirm DSM-IV criteria A1 and A2 trauma exposure and to establish current PTSD status. Current smoking status, including dependence, was assessed using the Fagerstrom test for nicotine dependence (Heatherton, Kozlowski, Frecker, & Fagerstrom, 1991). To qualify as a smoker, participants had to meet or exceed a minimum urinary cotinine concentration of 4 ng/mL. Cotinine is a metabolite of nicotine that reliably reflects chronic smoking intensity (Pomerleau, Fertig, & Shanahan, 1983).

Exclusion criteria included meeting diagnostic criteria for a psychotic disorder (except psychosis not otherwise specified [NOS] due to the presence of trauma-related auditory, olfactory, tactile, or visual hallucinations), Bipolar I disorder, substance or alcohol dependence within the past 3 months (except for nicotine dependence), and homicidal or suicidal ideation requiring clinical intervention. Psychosis NOS includes psychotic symptoms such as sensory hallucinations experienced during intrusions of trauma memories that do not otherwise fit into a spectrum of symptoms characteristic of a more pervasive psychotic disorder, such as schizophrenia or schizoaffective disorder. Additional exclusion criteria included pregnancy, having an unstable medical condition or medication regimen, and testing positive for alcohol or illicit drugs.

Testing Facility and Equipment

The testing facility consisted of a testing room adjacent to a control room (Figure 1). Ambient temperature in the testing room was maintained between 78°F and 85°F. The recording sensors were connected by cables to the data acquisition system in the adjoining control room.

Figure 1.

Two high-resolution infrared cameras form the backbone of the SAFE system. An RFHD High Performance IR thermal imager (L3-CCE, Mason, OH) mounted on a pan and tilt unit (Burchfield Automation) for imaging the face is shown in the left background. The data acquisition system is shown in the left foreground. The SC5600 (FLIR Systems, Willsonville, OR) is mounted under a table and aimed upward at the distal phalanx of the index and middle fingers (right panel).

A detailed description of the sensors, hardware, and software constituting the SAFE system is provided in Krzywicki et al. (2014). Two separate high-resolution infrared cameras imaged the distal phalanges of the index and middle fingers of the nondominant hand from a distance of 11.5 cm, and the face from 2.3 m (Figure 1). The finger sensor was a 640 × 512 InSb FLIR Systems SC5600 with a 27-mm focal length lens. The face sensor was an RFHD High Performance IR thermal imager (L3-CCE, Mason, OH) with a 100-mm lens (FLIR Systems, Willsonville, OR). It was mounted on a model CLU 555 HD pan and tilt unit (Burchfield Engineering, Covington, LA) on a tripod.

Skin conductance level (SCL) was recorded from two EL507 disposable Ag-AgCl electrodes applied to the volar surface of the proximal phalanges of the same middle and index fingers imaged by the SAFE sensor. The electrodes were connected via LEAD110A electrode leads to a GSR 100C BIOPAC biopotential amplifier (BIOPAC Systems, Inc., Goleta, CA). Thermal video of sweat pore activation and SCL were concurrently sampled at 25 Hz, synchronized, and stored for analysis on a custom-designed SAFE processing system.

Experimental Procedures

Preparation

On arrival at the laboratory, vital signs were obtained and urine was collected for pregnancy testing and a toxicology screen, which included tests for cocaine, methamphetamine, marijuana/THC, opiates, oxycodone, and benzodiazepines. A urine test for cotinine (metabolite of nicotine with a 34-h half-life) was administered to confirm overall smoking status. A breathalyzer test for carbon monoxide (CO) was used to assess possible recent smoking. A positive CO test was exclusionary for self-characterized nonsmokers; for smokers, it provided confirmation of recent smoking to satisfaction, per study instructions. An alcohol breathalyzer test was also administered. Participants then completed the Posttraumatic Stress Disorder Checklist–Civilian Version (Weathers, Litz, Herman, Huska, & Keane, 1993), a 17-item self-report measure that assesses the severity of PTSD symptoms according to DSM-IV criteria (Conybeare, Behar, Solomon, Newman, & Borkovec, 2012). Participants then sat quietly for at least 30 min to allow arousal stimulated by preparatory activities to return to baseline before the loud noise task commenced.

Loud Noise Task

Autonomic nervous system responses were elicited using loud noise bursts, similar to procedures previously described (Metzger et al., 1999; Orr et al., 2003; Turpin, Schaefer, & Boucsein, 1999). Participants were fitted with headphones and seated in a comfortable chair within line of sight of the face thermal sensor. The index and middle fingers of the nondominant hand were inserted into a plastic plate designed to maintain them at a fixed distance above the lens of the finger sensor (Figure 1). Skin conductance recording electrodes were attached as described above. The loud noise stimulus was provided via headphones. The software-generated pseudo-white noise stimuli were nominally set at 100 dB, had a rise time of 5 ms, and were 1 s in duration. The actual output level of the loud noise stimuli measured at the headphones in 66.7% of test sessions was 82.15 ± 0.72 dB. The stimulus for each session consisted of 15 loud noise bursts over approximately 12 min, with pseudorandom interstimulus intervals varying between 30 and 60 s, and a mean interstimulus interval of 45 s.

Data Processing

A correlation-based process (Krzywicki et al., 2014) was used to compute the pore activation index (PAI). The PAI is a weighted measure of active sweat pores visualized via thermal image video recording within a specified area of finger or face skin.

Data were processed in two steps. In the first step, responses specific to each stimulus were determined. The algorithm searched for peaks with amplitudes ≥ 10 pores for thermal channels, or conductance ≥ 0.05 μS for skin conductance. Peaks with these parameters were monitored between 1 and 4 s following a stimulus onset. These amplitude thresholds are consistent with values previously reported (Freedman et al., 1994; Juniper et al., 1967), and the time window is consistent with that previously established (e.g., Metzger et al., 1999; Orr et al., 2003). The response amplitude was calculated as the difference between the peak SCL, or number of pore openings, and the prestimulus baseline. The prestimulus baseline was defined as the mean SCL, or mean number of pore openings, over the 5-s interval immediately preceding each stimulus onset.

Machine algorithms were developed to automate determination of the amplitude of the time course for finger pore (FiP), facial pore (FaP), and SC responses to each loud noise stimulus, as shown in Figure 2. For the noise stimuli, response amplitude to the ith trial is A(i), i 5 1, .. ., 15. Habituation, H(i), may be derived from response to the ith stimulus and subsequent ones as

$$\text{H}(i)=\frac{A(i)}{A(i+1)},$$ (1)

or

$$\text{A}(i)=\text{A}(1)\cdot e^{-(i-1)h},$$ (2)

where h is a habituation regression coefficient. The linear habituation relationship of Equation (1) is simpler; however, intuitively, the regression-fitted exponential expression is likely a better representation of neurophysiological habituation. We analyzed both.

Figure 2.

Parameters of skin conductance and pore count responses. The current study examined only the amplitude of the responses.

Habituation (H₂) was calculated from the second to the 15th loud noise trial responses (i = 2, .. ., 15), as has been done in previous studies (e.g., Metzger et al., 1999; Orr et al., 2003). We also calculated habituation (H₁) using responses to all trials, that is, A(i), i = 1, .. ., 15.

Statistical Analysis

Skin conductance response and pore activation data exhibited substantial positive skew. A square-root transformation was performed on both measures, which substantially reduced skewness. Therefore, all of the analyses were conducted using square-root transformed measures. A series of multilevel regressions (i.e., mixed-effects models or hierarchical linear models) using the software program Hierarchical Linear and Non-Linear Modeling software (Raudenbush, Bryk, & Congdon, 2011) with restricted maximum likelihood estimation was used to examine associations among SC, FiP, and FaP indices, and associations between these indices and PTSD diagnostic status or severity of PTSD symptoms. Multilevel regression analysis was developed to analyze nested or hierarchical data structures (Raudenbush & Bryk, 2002). For the current study, multiple responses to the loud noises (within-subject component or Level 1) were nested within individuals (between-subjects component or Level 2). Strengths of this multilevel regression include (a) capability of handling missing data and unbalanced designs (i.e., the number of valid loud noise measurements can vary across subjects), (b) highly efficient and powerful estimation procedures that include all data points available, and (c) modeling flexibility that allows for the inclusion of predictors and covariates that are continuous or categorical, as well as time invariant or time varying.

The first set of analyses examined the within-subject associations between PAI and SC measurements across loud noise trials. In separate analyses, all pairwise associations (i.e., SC–FiP, SC–FaP, FiP–FaP) were estimated. One of the measures was specified as the outcome and the other was specified as a Level 1 (time-varying) predictor. Because these estimates represent bivariate associations, which variable is specified as the outcome and which is specified as the predictor should not influence the strength of the association. However, for completeness, we analyzed each pair both ways (e.g., PAI as the outcome and SC response as the predictor, and vice versa).¹

A second set of analyses investigated the change in reactivity across loud noise presentations. For these analyses, trial number (coded as 0–14) was entered as a Level 1 predictor of each measure. This growth curve modeling approach produces two Level 1 coefficients. With the trial number of the first loud noise trial coded as zero, the Level 1 regression intercept represents initial response amplitude for the first loud noise. The regression coefficient for the trial number predictor estimates change in response amplitude across loud noises. It is typical for responding to decrease across trials, representing a habituation effect (e.g., Orr et al., 2003). We examined linear habituation by including the loud noise trial number described above as a Level 1 predictor. We examined nonlinear habituation by computing the natural log of the loud noise number and including this as a Level 1 predictor. This evaluates change over time, characterized by a strong initial change that flattens out across trials. Hierarchical linear modeling (HLM) allows the estimates of initial status and change over time for each participant to be saved and used in subsequent analyses (e.g., Griffin, 1997; Raudenbush et al., 2011). To examine the association between initial response amplitude and habituation among measures, we saved the initial response amplitude and habituation estimates for the PAI and SC response measures and computed zero-order correlations.

The final set of analyses examined the association between the physiological measures and both PTSD diagnostic status and PCL symptom severity on the day of testing. To estimate the association between PTSD diagnostic status and the physiological responses, a dummy-coded PTSD diagnostic status variable was used as a Level 2 predictor of PAI and SC response estimates, separately. These analyses were conducted in two ways. First, to examine the association between PTSD and average reactivity across all loud noise trials, a model with no Level 1 predictors (thus, the Level 1 estimate represents the average PAI or SC response across all trials) and PTSD as a Level 2 predictor was estimated. Second, to examine PTSD as a predictor of initial reactivity and change over time, a growth curve model was estimated with trial number as a Level 1 predictor and PTSD diagnosis as a Level 2 predictor of both initial response amplitude and trial number (i.e., change across trials).

HLM does not produce standardized regression equations. To depict the strength of the associations, we report three estimates of effect sizes. For within-subject associations, we include estimates of r, which represent the amount of Level 1 (within-subject) variance accounted for by the predictor variable. For all results, we report partial correlation coefficients (pr) calculated using

$$pr=\sqrt{\frac{t^2}{t^2+df}},$$ (3)

where df is degrees of freedom and t is the t-test statistic. For both r and pr, .10, .24, and .37 indicate small, medium, and large effect sizes, respectively (Kirk, 1996). We also computed the more familiar Cohen’s d metric using

$$d=\sqrt{\frac{2t}{df}},$$ (4)

with .20, .50, and .80 representing small, medium, and large effect sizes, respectively (Cohen, 1988).

As the primary purpose of the current study was to examine the utility of the SAFE skin pore count as a unique and novel measure of SNS reactivity and not to provide a rigorous test of theory specifying the role of SNS reactivity in PTSD, we adopted a more liberal hypothesis-generating style in examining the relationships between (a) SC and SAFE skin pore count reactivity, and (b) PTSD diagnosis and symptom cluster severity, in accordance with Abelson (1995). Therefore, we did not correct for multiple comparisons and set significance at p < .05 and a trend for significance at p < .10. Importantly, however, we report effect size estimates for all of the tests, which are probably more informative than null hypothesis testing in the context of the relatively small size.

Results

Characteristics of Participants in Final Sample

Complete demographic information is reported in Table 1. Thirty participants completed the loud noise task: 11 with current PTSD and 19 with trauma exposure but no current PTSD (five met criteria for remote past PTSD). Four participants were identified as nonresponders (i.e., exhibited SC response amplitude of less than 0.05 μS to the first two noises); their data were excluded from further analyses (Orr et al., 2003). Thus, the final sample analyzed for this study included 26 participants (see Table 1 for demographic information as well as diagnostic and smoking status). Among the trauma-exposed participants without current PTSD, nine took no medications and one smoked. Medications taken by the other 10 non-PTSD participants included (1) acne cream; (2) ibuprofen and steroid contraception; (3) hydrochlorthiazide; (4) naproxen, cetirizine, and loratadine; (5) aspirin, atorvastatin, hydrochlorothiazide, gabapentin, and amlodipine; (6) smoker—statin and nonbeta blocker blood pressure medication; (7) colchicine; (8) smoker—aspirin, diclofenac, tramadol, docusate sodium, and famotidine; (9) smoker—steroid contraception; and (10) fexofenadine. In the PTSD group, five participants took no medications; one of these five smoked. Medications taken by the other six PTSD participants included (1) an antibiotic; (2) albuterol; (3) atorvastatin, hydrochlorothiazide, gabapentin, and amlodipine; (4) smoker—meloxicam, diphenhydramine/mirtazapine at night for sleep; (5) ibuprofen and lisinopril; and (6) smoker—quetiapine, zolpidem, acetaminophen, and oxycodone.

Table 1.

Demographic Variables and Smoking Status

Variable		Total sample		Responders		Nonresponders		PTSD−		PTSD+		Nonsmokers		Smokers
	Category	(N = 30)		(n = 26)		(n = 4)		(n = 16)		(n = 10)		(n = 19)		(n = 7)
		M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
Age		48.77	11.96	48.92	11.96	47.75	13.70	47.56	11.98	51.10	12.24	46.79	13.10	54.71	5.31
Gender		n	%	n	%	n	%	n	%	n	%	n	%	n	%
	Male	20	66.7	19	73.1	1	25.0	12	75.0	1	70.0	13	68.4	6	85.7
	Female	10	33.3	7	26.9	3	75.0	4	25.0	3	30.0	6	31.6	1	14.3
Race
	Caucasian	16	53.3	13	50.0	3	75.0	9	56.2	4	40.0	10	52.6	3	42.9
	African	12	40.0	11	42.3	1	25.0	7	43.8	4	40.0	8	42.1	3	42.9
	American Latino	2	6.7	2	7.7	0	0.0	0	0.0	2	20.0	1	5.3	1	14.3
Smoking status
	Nonsmoker	20	66.7	19	73.1	1	25.0	12	75.0	7	70.0	19	100.0	0	0.0
	Smoker	10	33.3	7	26.9	3	75.0	4	25.0	3	30.0	0	0.0	7	100.0

Note. Responder and nonresponders refer to startle responders and nonresponders as operationalized in the text. PTSD- and PTSD1 indicate PTSD diagnostic status as assessed by the Clinician Administered PTSD Scale (CAPS).

Within-Subject Associations Between PAI (FiP and FaP) and SC Responses

Changes in pore activity on the fingers and the face were clearly recorded using the SAFE thermal imaging sensors. Visual inspection of raw data from the loud noise trials supported convergent validity of FiP and FaP measures by SAFE, with each other and also with SC response. The PAI and SC responses to the initial trials were large and decreased in amplitude over successive trials, as illustrated in Figure 3. Of note, pore opening responses to each stimulus returned rapidly to baseline whereas SCL did not, as observed consistently since this SCL characteristic was first reported by Darrow (1964).

Figure 3.

Left: Thermal images of the fingers and face during the loud noise test for a mock participant. A mock participant’s image is presented to protect the privacy of actual study participants. Right: The three panels show three dashed vertical lines at 30, 80, and 122 s, representing points in time when the acoustic stimulus occurred. Onset of SC and pore responses occurred within 1.5 s of stimulus onset. The response amplitudes demonstrate diminutions across the repeated stimulus presentations, which is characteristic of response habituation. SC level in the top panel did not return to baseline between stimuli, a reliable SC phenomenon first reported by Darrow (1964).

Table 2 provides the within-subject associations among the PAI and SC measures. All three of the measures were significantly associated with each other. The associations between SC and FiP, when examined with SC response as the predictor and again with FiP response as the predictor (rs .92 and .81, respectively), were larger than the associations between SC and FaP (rs .46 and .76, respectively) and between FiP and FaP responses (rs .47 and .47, respectively).

Table 2.

Within-Subject Associations Among Measures of SNS Response

		Outcome
Predictor		1 SC	2 FiP	3 FaP
1	SC
	r	-	.81	.46
	pr	-	.89	.42
	d	-	3.94	.91
2	FiP
	r	.92	-	.47
	pr	.93	-	.58
	d	5.06	-	1.43
3	FaP
	r	.76	.47	-
	pr	.60	.53	-
	d	1.51	1.26	-

Note. All associations are statistically significant with ps < .001. For both r and pr, .10, .24, and .37 indicate small, medium, and large effect sizes, respectively (Kirk, 1996). Bivariate multilevel regression analyses were conducted to estimate the associations. Estimates of the association between each pair of variables was calculated twice, switching which variable was specified as the predictor and which was specified as the outcome with each result presented either above or below the diagonal. SC = skin conductance; FiP = finger pore activation index; FaP = face pore activation index; r = amount of the Level 1 (within-subject) variance accounted for by the predictor variable; pr = partial correlation coefficient; d = Cohen’s d with .20, .50, and .80 representing small, medium, and large effect sizes, respectively (Cohen, 1988).

Results of the growth curve analyses examining response amplitude change across loud noise trials showed that SC (b_{ln(trial number)} = −.36, SE = .04, t = −7.42, p < .001, pr = .83, d = −2.97), FiP (b_{ln(trial number)} = −5.54, SE 5 .95, t = −5.85, p < .001, pr = .76, d = −2,34), and FaP (b_{ln(trial number)} = −1.64, SE = .64, t = −2.55, p = .017, pr = .45, d = −1.02) response amplitudes decreased across trials. The response profiles were characterized by large initial decreases in amplitude that became smaller across trials.² While all three measures exhibited substantial amplitude decreases across trials, the effect size for this decrease was larger for SC (d = −2.97) and FiP (d = −2.34) responses compared to FaP response (d = −1.02). Table 3 displays the correlations for intercept (amplitude of response to the first loud noise) and slope (habituation) across PAI and SC measures and shows that estimates of initial response amplitude and habituation for SC and FiP were strongly correlated, while estimates of initial response amplitude and habituation for FaP were not associated with estimates of initial response amplitude and habituation for either SC or FiP.

Table 3.

Associations Between Estimates of SC and FiP or FaP Responses to the First Loud Tone (Intercept) and Habituation Across Trials (Slope)

	SNS coefficient	1	2	3	4	5	6
1	SC intercept
2	SC slope	−0.93*
3	FiP intercept	0.83 *	−0.74*
4	FiP slope	−0.78*	0.76 *	−0.97*
5	FaP intercept	0.34	−0.23	0.34	−0.25
6	FaP slope	−0.26	0.15	−0.36	0.29	–0.96*

Note. Bold denotes associations between SCR and FiP or FaP intercepts across measures, and underline denotes associations between SCR and FiP or FaP slopes across measures. SC = skin conductance; FiP = finger pore activation index; FaP = face pore activation index; Intercept = initial response amplitude; Slope = change across trials (natural log) or habituation.

^*p < .05; however, for all significant associations, p < .00.

PTSD as a Predictor of PAI and SC Responses

As depicted in the top portion of Table 4, participants diagnosed with PTSD tended to show larger PAI and SC responses to the first loud noise, compared to those not meeting criteria for PTSD. While these effects did not reach statistical significance, they were all characterized by medium effect sizes (d = .44 to .56). As depicted in the middle portion of Table 4, participants diagnosed with PTSD also tended to show faster habituation than participants without PTSD; again, the differences were characterized by medium effect sizes. The bottom portion of Table 4 displays the estimates for responses when averaged across trials as a function of PTSD status. When averaged across all trials, individuals with PTSD tended to show larger PAI and SC responses; the differences were also associated with a medium effect size (d = .56 to .64), which did not reach statistical significance in this relatively small sample.

Table 4.

Estimates of Initial Response, Habituation, and Average Response for SNS Measures as a Function of PTSD Diagnostic Status

	SC Estimate [95% CI)	FiP Estimate [95% CI)	FaP Estimate [95% CI)
		Initial response
PTSD+	1.38 [0.94, 1.82]	26.32 [17.98, 34.65]	15.99 [6.50, 25.48]
PTSD−	0.99 [0.64, 1.33]	17.97 [9.23, 26.71]	10.55 [7.75, 13.34]
Difference	0.39 [−0.17, 0.95]	8.34 [−3.74, 20.43]	5.45 [−4.44, 15.34]
Difference d	0.57	0.56	0.44
		Habituation
PTSD+	−0.38 [−0.51, −0.25]	−6.55 [−9.28, −3.81]	−2.15 [−5.04, 0.74]
PTSD−	−0.29 [−0.41, −0.18]	24.92 [27.39, 22.45]	−1.33 [−2.32, 20.35]
Difference	−0.09 [−0.26, 0.08]	21.63 [−5.31, 2.06]	−0.82 [−3.87, 2.24]
Difference d	−0.41	−0.36	−0.22
		Average response (collapsed across all trials)
PTSD+	0.62 [0.36, 0.89]	13.25 [8.65, 17.85]	11.61 [6.68, 16.54]
PTSD−	0.40 [0.24, 0.57]	8.24 [3.91, 12.57]	7.87 [6.63, 9.12]
Difference	0.22 [−0.09, 0.53]	5.01 [−1.31, 11.33]	3.74 [−1.35, 8.82]
Difference d	0.56	0.64	0.59

Note. SNS = sympathetic autonomic nervous system; SC = skin conductance; FiP = finger pore activation index; FaP = face pore activation index; d = Cohen’s d with .20, .50, and .80 representing small, medium, and large effect sizes, respectively (Cohen, 1988).

Table 5 provides the estimates for CAPS scores (ascertained during the screening evaluation) and PCL scores (measured on the day of loud noise testing) as predictors of initial response (i.e., response to the first loud noise presentation) and habituation for the PAI and SC measures. While many of these effects fell in the medium to medium-large range, the most robust effect emerged for the reexperiencing subscale of the PCL (PCL B). Participants with higher PCL B scores exhibited significantly larger SC (p = .008; d = 1.19) and FiP (p = .023; d = .99) responses to the first loud noise, and greater subsequent habituation of SC (p = .001; d = 1.19) and FiP (p = .020; d = .99) responses. In general, the effect sizes for predicting PAI and SC measures by PTSD symptom severity scores were smaller for FaP than for SC and FiP; none of the FaP estimates reached statistical significance.

Table 5.

Estimates for PTSD Continuous Symptom Severity Measures as Predictors of SNS Response to the First Loud Noise and Habituation of SNS Responses Across Trials

SNS response	PTSD measure	Estimate	b [95% CI]	p	pr	d
SC	CAPS Total
		Initial	0.004 [20.004, 0.013]	.310	.21	0.42
		Habituation	20.002 [20.005, 0.001]	.169	.28	20.58
	CAPS B
		Initial	0.018 [−0.009, 0.045]	.191	.27	0.55
		Habituation	−0.007 [−0.016, 0.001]	.108	.32	−0.68
	CAPS C
		Initial	0.013 [−0.007, 0.033]	.211	.25	0.53
		Habituation	−0.005 [−0.011, 0.002]	.168	.28	−0.58
	CAPS D
		Initial	0.003 [−0.024, 0.03]	.821	.05	0.09
		Habituation	−0.004 [−0.013, 0.005]	.353	.19	−0.39
	PCL Total
		Initial	0.013 [−0.001, 0.027]	.074	.36	0.76
		Habituation	−0.005 [−0.01, 20.001]	.035	.41	−0.91
	PCL B
		Initial	0.049 [0.016, 0.082]	.008	.51	1.19
		Habituation	−0.021 [−0.032, −0.01]	.001	.61	−1.53
	PCL C
		Initial	0.03 [0, 0.07]	.092	.34	0.72
		Habituation	−0.01 [−0.02, 0]	.073	.36	−0.76
	PCL D
		Initial	0.028 [−0.018, 0.074]	.245	.24	0.49
		Habituation	−0.013 [−0.028, 0.003]	.115	.32	−0.67
FiP	CAPS Total
		Initial	0.08 [−0.14, 0.3]	.492	.14	0.28
		Habituation	−0.03 [−0.1, 0.04]	.436	.16	−0.32
	CAPS B
		Initial	0.27 [−0.38, 0.93]	.418	.17	0.34
		Habituation	−0.1 [−0.33, 0.12]	.367	.18	−0.38
	CAPS C
		Initial	0.23 [−0.29, 0.75]	.387	.18	0.36
		Habituation	−0.07 [−0.23, 0.1]	.428	.16	−0.33
	CAPS D
		Initial	−0.01 [−0.47, 0.45]	.974	.01	−0.01
		Habituation	0.05 [−0.07, 0.16]	.407	.17	0.34
	PCL Total
		Initial	0.29 [−0.11, 0.69]	.161	.28	0.59
		Habituation	−0.09 [20.22, 0.04]	.200	.26	−0.54
	PCL B
		Initial	1.16 [0.22, 2.11]	.023	.45	0.99
		Habituation	−0.4 [−0.71, −0.08]	.020	.46	−1.02
	PCL C
		Initial	0.56 [−0.43, 1.56]	.274	.22	0.46
		Habituation	−0.14 [−0.46, 0.17]	.381	.18	−0.36
	PCL D
		Initial	0.71 [−0.5, 1.92]	.256	.23	0.48
		Habituation	−0.2 [−0.6, 0.19]	.312	.21	−0.42
FaP	CAPS Total
		Initial	0.01 [−0.13, 0.15]	.869	.03	0.07
		Habituation	0.01 [−0.02, 0.05]	.472	.15	0.30
	CAPS B
		Initial	0 [−0.37, 0.37]	.995	.00	0.00
		Habituation	0.05 [−0.05, 0.14]	.348	.19	0.39
	CAPS C
		Initial	0.07 [−0.28, 0.42]	.694	.08	0.16
		Habituation	0.02 [−0.07, 0.11]	.640	.10	0.19
	CAPS D
		Initial	−0.01 [−0.47, 0.45]	.974	.01	−0.01
		Habituation	0.05 [−0.07, 0.16]	.407	.17	0.34
	PCL Total
		Initial	0.14 [−0.2, 0.48]	.418	.17	0.34
		Habituation	−0.02 [−0.12, 0.08]	.719	.07	−0.15
	PCL B
		Initial	0.06 [−0.49, 0.62]	.818	.05	0.10
		Habituation	0.02 [−0.17, 0.2]	.848	.04	0.08
	PCL C
		Initial	0.46 [−0.57, 1.49]	.385	.18	0.36
		Habituation	−0.05 [−0.33, 0.23]	.705	.08	−0.16
	PCL D
		Initial	0.58 [−0.59, 1.76]	.337	.20	0.40
		Habituation	−0.12 [−0.49, 0.26]	.552	.12	−0.25

Note. Bold denotes statistically significant relationships. SNS 5 sympathetic nervous system; SC 5 skin conductance; FiP = finger pore activation index; FaP = face pore activation index; pr = partial correlation coefficient with 10, .24, and .37 indicating small, medium, and large effect sizes respectively (Kirk, 1996); d = Cohen’s d with .20, .50, and .80 representing small, medium, and large effect sizes, respectively (Cohen, 1988).

All of the analyses above included responses to all loud noise trials (i.e., trials 1–15). These analyses were repeated with the response to the first loud noise removed (i.e., loud noise trials 2–15). The pattern of results was similar across the two sets of analyses. However, with the first trial removed, the associations were attenuated. For example, the r value for the bivariate within-subject associations with SC response as the predictor and FiP response as the outcome was .92 when including all loud noise trials, but reduced to .79 with the first trial excluded. Similarly, the association between PCL B symptoms and initial SC and FiP response amplitudes were statistically significant with effect sizes of d = 1.18 and .99 (respectively), when all trials were included. With the first trial removed, the same estimates were no longer statistically significant, with effect sizes of d = .61 and .57, respectively. This suggests that the first trial provides important information. For brevity, we report the results below with all loud noise trials included. Results of analyses that exclude the first trial can be obtained from the first author.

Effects of Smoking on PAI and SC Responses to the Loud Noise Task

Seven (27%) of the 26 participants with valid physiological data were smokers. Three of the seven smokers (42.9%) were diagnosed with PTSD; four were PTSD negative. A chi-square test indicated that the proportion of smokers among participants with PTSD (3/10, 33.0%), and those without PTSD (4/16, 25.0%), did not differ significantly, χ²(df = 1, n = 26) = .08, p = .780).³

The top one third portion of Table 6 displays the mean PAI and SC response amplitudes to the loud noises across all trials as a function of smoking status. Smokers displayed significantly smaller responses for both SC (p = .004, d = 1.32) and FiP (p = .003, d = 1.38); the difference was not statistically significant for FaP (p = .124, d = .65). The bottom two thirds of Table 6 displays the estimates of the growth curve analysis examining initial response amplitude and habituation as a function of smoking status. Although smokers tended to exhibit smaller initial responses than nonsmokers (ds ranging from −.64 to −.78), the differences were not statistically significant for any of the measurements (all ps > .06). Smokers and nonsmokers showed similar patterns of habituation for the SC and PAI measures (all ps > .11). Although we did not have sufficient power to adequately test the PTSD × Smoking Status interactions, Figure 4 demonstrates similar effects of smoking for individuals with and without PTSD (smoking effect = −9.65 for PTSD and −8.13 for no PTSD).

Table 6.

Estimates of Initial Response, Habituation, and Average Response for SNS Measures as a Function of Smoking Status

Estimate		SCR Estimate [95% CI]	FiP Estimate [95% CI]	FaP Estimate [95% CI]
Mean response across trials
	Smoker	0.28 [0.18, 0.37]	4.73 [−1.6, 11.06]	7.68 [6.39, 8.98]
	Nonsmoker	0.63 [0.43, 0.82]	13.15 [9.31, 16.99]	10.27 [7.32, 13.22]
	Difference	−0.35 [−0.57, −0.14]	−8.42 [−13.36, 23.48]	−2.59 [−5.81, 0.63]
	d Difference	−1.32	−1.38	−0.65
Initial response to first loud noise
	Smoker	0.85 [0.49, 1.21]	13.66 [5.28, 22.04]	8.26 [5.19, 11.33]
	Nonsmoker	1.25 [0.9, 1.6]	23.94 [16.02, 31.86]	14.25 [8.85, 19.66]
	Difference	−0.4 [−0.9, 0.1]	−10.28 [−21.81, 1.25]	−5.99 [−12.21, 0.23]
	d Difference	−0.64	−0.72	−0.78
Habituation
	Smoker	−0.31 [−0.46, −0.16]	−4.8 [−8.24, −1.37]	−0.31 [−1.82, 1.2]
	Nonsmoker	−0.33 [−0.44, −0.23]	−5.81 [−8.03, −3.59]	−2.14 [−3.74, −0.54]
	Difference	0.03 [−0.16, 0.21]	1.01 [−3.08, 5.1]	1.83 [−0.37, 4.03]
	d Difference	0.11	0.20	0.67

Note. SNS = sympathetic nervous system; SC = skin conductance; FiP = finger pore activation index; FaP = face pore activation index; d = Cohen’s d with .20, .50, and .80 representing small, medium, and large effect sizes, respectively (Cohen, 1988).

Figure 4.

Responses to loud noise stimuli as a function of smoking status and PTSD diagnosis, measured by (a) the finger pore activation index (FiP), and (b) skin conductance (SC), were greater for the group diagnosed with PTSD than for the trauma-exposed participants without PTSD for both smokers and nonsmokers. Responses were smaller in the smokers for both the PTSD and trauma-exposed participants.

Discussion

The current study confirms work by Krzywicki et al. (2014), demonstrating that counts of open skin sweat pores can serve as a valid surrogate measure of SNS reactivity. Using breathing manipulation as a stimulus, Krzywicki et al. (2014) showed that finger and face PAI responses were significantly correlated with SC responses. In the current study—thermal imaging of finger skin pore reactivity—FiP showed excellent construct validity, when compared to SC reactivity, as a measure of SNS activity in trauma-exposed samples of men and women with and without PTSD. Overall, finger pore reactivity measured by SAFE showed the highest convergent validity with a traditional measure of SC reactivity. This relationship was observed for finger pore and SC initial response amplitudes to a loud noise stimulus, as well as habituation of these responses over a series of 15 stimulus presentations.

At first glance, other noncontact methods used to assess physiological activity via detection of infrared bandwidths may appear similar to SAFE (e.g., Di Giancinto et al., 2014; Engert et al., 2014; Pavlidis et al., 2012; Shastri et al., 2009, 2012); however, closer examination reveals that they are not. The method of Shastri et al. (2009, 2012) appears most comparable. These investigators imaged the skin with high-resolution thermal sensors, and then applied wavelet analysis or a contour-based adaptation of black top-hat transformation frame by frame, to quantify transient perspiration on the skin during cognitive challenges. Their findings were consistent with the idea that quantification of perspiration on the skin is a valid method for assessing increased stress. In contrast, our custom-designed software computes thermal gradient profiles of very small regions with very high resolution (<0.25 milliKelvin [mK]) in two dimensions in order to determine whether a sweat pore in the region is open or closed. The number of open sweat pores within a defined region of skin is then integrated over time to yield waveforms (Figure 3) that serve as surrogate signals of SNS reactivity (Krzywicki et al., 2014).

Based on research to date, our pore-counting method appears to be more sensitive than most other noncontact methods that probe skin temperature. For example, Engert et al. (2014) compared thermal imprints, resolved to 20 mK over large segments of the face, to heart rate, heart rate variability, and levels of alpha-amylase and cortisol in 15 individuals participating in a cold pressor test and the Trier Social Stress Test. They were unable to establish statistically significant correlations among these measures of stress. Pavlidis et al. (2012) was able to quantify regional facial blood flow (i.e., instantaneous warming around the eyes) as a surrogate of SNS reactivity.

As depicted in Figure 3 of the current study, the SAFE poreopening responses returned rapidly to baseline, while SCL changes did not. The pore data reflect a binary mechanical response (open vs. closed); thus, it is not surprising that this measure quickly returns to baseline. SCL, however, is a more complicated measure influenced by a range of individually variable factors including anxiety, novelty of the laboratory experience, and general arousal, as well as the total amount of sweat produced, permeability of the dermis, and rates of sweat absorption and evaporation. Of note, the change in SCL between peaks over the three trials depicted in Figure 3 is approximately 0.2 μS, which is actually a rather small and common shift in baseline SCL consistent with previous reports (Darrow, 1964).

The current relatively small study found that SC and sweat pore responses to the loud noise stimuli did not significantly differentiate trauma-exposed individuals with a diagnosis of PTSD from those without PTSD, although the effect size of the group difference suggested that SNS reactivity was increased in the PTSD group. PTSD is a complex diagnostic construct composed of relatively independent clusters of symptoms, which were initially defined by committee consensus, as per DSM-IV, but later refined by confirmatory factor analyses (CFAs). Thus, it is noteworthy that SC and SAFE-measured finger pore reactivity predicted overall severity of the DSM-IV PTSD reexperiencing symptom cluster (which includes emotional and physiological reactivity to trauma reminders, as well as trauma-related intrusive thoughts, nightmares, and flashbacks) self-assessed on the day of testing. In fact, the largest correlation was with self-rated severity of physiological reexperiencing. The finding of strong and significant correlations between both SC and finger pore reactivity to the first noise presentation and PTSD reexperiencing symptoms aligns with the recent emphasis being placed on dimensional approaches to psychopathology (i.e., National Institute of Mental Health Research Domain Criteria [NIMH RDoC]). Interestingly, finger pore and SC reactivity to the loud noise stimuli did not show a relationship with DSM-IV PTSD hyperarousal symptoms, which include exaggerated startle, as one might have expected. This suggests that the heightened SNS reactivity generated by the loud noise stimuli is not related to general hyperarousal as defined by DSM-IV, which also includes symptoms of hypervigilance, irritability, difficulty sleeping, and trouble concentrating. To examine this further, we conducted follow-up analyses using a four-factor CFA model of PTSD defined by Simms, Watson, and Doebbelling (2002), which includes (1) an identical reexperiencing factor (all DSM-IV criteria B symptoms), (2) a strategic avoidance factor (the two criterion C symptoms representing active avoidance of trauma-related thoughts, feelings, and situations), (3) a more discrete hyperarousal factor (hypervigilance and exaggerated startle), and (4) a dysphoria factor (DSM-IV symptoms C3–C7 and D1–D3), thought to represent general distress common to many mood and anxiety disorders (i.e., not specific to PTSD). Like the original analyses using the DSM-IV three-factor model, only the PTSD reexperiencing cluster predicted SNS activity. Thus, both SC and finger sweat pore reactivity to the first loud noise appears to objectively quantify an underlying biological dimension (i.e., PTSD endophenotype) that maps closely to a subjectively experienced and CFA-confirmed coherent PTSD symptom cluster.

Studies by Orr, Lasko, Shalev, and Pitman (1995), Orr et al. (2003), and Shalev et al. (1992), but not Carson et al. (2007) and Metzger et al. (1999), reported slower habituation (as measured by slope) of SC responses to a series of 95 dB pure tones in individuals with current PTSD. For these studies, habituation of SC responses was calculated from the second through the fifteenth trials. SAFE H₂, our equivalent of the habituation parameter used by the three aforementioned studies, produced a negative result when comparing the PTSD and non-PTSD groups, which is consistent with the negative findings reported by Carson et al. (2007) and Metzger et al. (1999). In contrast, SAFE H₁, computed for FiP and SC responses over all 15 trials, significantly predicted PTSD DSM-IV-defined reexperiencing symptoms as assessed by the PCL, and indicated faster rather than slower habituation in the PTSD group. This raises the possibility that, including the first trial in the measure of habituation, as done for SAFE H₁, may lead to different results and conclusions about potential differences in SNS reactivity and habituation between individuals with and without PTSD. The response to the first trial, which is typically large, may have a disproportionate effect on the measure of habituation. However, because of the relatively small sample size in the current study, it is important to emphasize the tentative nature of the PTSD findings and note that they have been reported primarily to complement the strong within-subject associations between the SC and SAFE measures, thus highlighting the potential utility of SAFE as an alternative measure of SNS reactivity. In addition, these findings can help guide future research examining the role of SNS activity in PTSD.

Conceptually, SAFE measurement of facial pore reactivity is very promising, as the camera is located at a distance from the individual and can be unobtrusive, or even hidden. However, in our study, facial pore reactivity was substantially less well correlated with SC reactivity than SAFE measurement of finger pore reactivity. Several factors may account for this observation. It is possible that the physiology of face sweat pore activity is different from that in the fingers. Technological limitations of the SAFE camera also likely contributed to the discrepancies. The fingers were maintained at a fixed distance from the sensor, keeping the finger pore images consistently in focus. In contrast, study participants often move their heads. Although the face sensor has automatic focus readjustment, it may not adjust quickly enough to compensate for changes in depth of focus when a participant’s head moves substantially. Therefore, although this proof-of-concept study demonstrated that SAFE FiP noncontact methodology behaves similarly to SC in capturing SNS activity, further technical work will be needed to compensate for head movement during FaP measurement to realize its promise.

The link between smoking and PTSD has been well documented. In a systematic review of 45 clinical studies, smoking rates were 40%–86% higher in their PTSD samples, and there were consistent positive relationships between PTSD diagnosis and smoking or nicotine dependence (Fu et al., 2007). In comparing trauma-exposed participants with PTSD to those without PTSD, the relative increase in the PTSD smoking rate diminishes (e.g., Beckham et al., 1997; Koenen et al., 2005; Shalev et al., 1990). In the current study, for both the PTSD and non-PTSD groups, recent smoking among the dependent smokers had significant dampening effects on SC and FiP response amplitude and habituation across trials. One might speculate that this “calming” effect explains why individuals exposed to trauma may be more likely to smoke (Japuntich et al., 2015) or why individuals with PTSD may be more likely to become nicotine dependent and less able to quit. As previously reviewed (Rasmusson et al., 2006), this calming effect of smoking among PTSD patients may be related to smoking-induced release of neurobiological factors such as neuropeptide Y and the GABAergic neuroactive steroid, allopregnanolone, which otherwise have been shown to be deficient in PTSD. The results of the current study also suggest that it may be important to account or control for smoking generally, and to control for time from last nicotine use in psychophysiological studies aimed at discriminating individuals with disorders characterized by dysregulation of the sympathetic nervous system or the hypothalamic-pituitary-adrenal axis. It is possible that variable smoking rates among study participants with and without PTSD in past psychophysiology studies have contributed to the variability with which measures of SNS reactivity have discriminated PTSD and control groups (e.g., Metzger et al., 1999; Orr et al., 2003).

The findings of the current study generally support the potential utility of SAFE technology in the assessment of SNS reactivity in PTSD or other populations in which SNS reactivity may be of interest. Situations and specific events precipitating PTSD fall into broad trauma categories such as exposure to combat, motor vehicle accidents, and domestic violence. Individuals exposed to these situations respond in individualized patterns. Associated with these traumatic experiences are specific cues that can trigger an involuntary neurophysiological response in the PTSD sufferer. Currently, diagnosis requires observation and appropriate attribution of conditioned reactions to causative traumatic events using a diagnostic interview performed by a clinician. These interviews may be subject to bias for a number of reasons. A patient may underreport traumatic events or related PTSD symptoms due to emotional discomfort or fear of stigma. Clinicians also may make diagnostic errors by failing to differentiate PTSD from similar diagnoses (e.g., depression or other anxiety disorders), or the patient may overreport symptoms in order to receive treatment or financial compensation or mitigate responsibility for certain actions.

Monitoring of SNS reactions may provide additional diagnostic or illness time course information, such as the progression of emotional and physiological habituation during a therapy session or extinction in response to a course of psychotherapy or pharmacotherapy. However, the current standard psychophysiological tests typically require physical contact to place sensors. In PTSD subjects, for whom sensitivity to novelty and perceived threat may be heightened, this is undesirable as it may increase discomfort or distress and alter baseline measures. Currently, leading candidates for noncontact surrogate measures of SNS activity include measures of saccadic eye movement and brain imaging. To the extent that measures are sensitive, specific, and cost effective, they may find useful application in evaluation of PTSD or in fields where their noncontact nature is a significant advantage. We believe that SAFE enjoys a comparative advantage, in part because it is both noncontact and relatively unobtrusive. The current study demonstrated that thermal imaging of finger pore reactivity is very accurate and convenient for participants. Technological advances that could increase the accuracy of thermal imaging of face pore reactivity and allow miniaturization of the camera may also enhance its utility.

Conclusions

The current study extends findings of Krzywicki et al. (2014) by demonstrating that thermal imaging of finger sweat pore openings correlates with established skin conductance measures of SNS reactivity in trauma-exposed individuals with and without PTSD. Major advantages of SAFE are that it does not require sensor contact with the person being monitored and the thermal imaging camera may be placed unobtrusively at a distance or concealed. In the current study, both SC reactivity and thermal imaging of finger pore openings to an initial brief loud white noise stimulus correlated strongly with self-rated PTSD reexperiencing symptoms. In addition, the study suggests that recent smoking to satisfaction among chronic smokers with and without PTSD is associated with diminished arousal as measured by both SC and SAFE technologies, supporting the importance of controlling for smoking or nicotine use more broadly in psychophysiological studies. Future work in a fully powered study will be needed to replicate or expand on these observations.

For most of the past two decades, the gold standard for diagnosing PTSD has been symptom quantification on a PTSD symptom rating scale such as CAPS or PCL. More recently, there has been a pivot to examination of the neurobiology underlying PTSD symptoms that may be shared with any number of other DSM-defined categorical disorders. SAFE monitoring of sweat pore reactivity may provide another avenue to the overarching goal of understanding the biological basis of PTSD and similar conditions in which dysregulation of the SNS is implicated.