Auditory Startle Response in Firefighters Before and After Trauma Exposure

Eligible subjects were 87 successive male recruits (mean age=30.1 years, SD=5.1) to the New South Wales Fire Brigades who were receiving class-based instruction at the time of recruitment; three subjects (3%) declined to participate in the study. The subjects were initially evaluated before commencing active firefighting duties (time 1), and then attempts were made to reassess all subjects within 12 months of the initial assessment (time 2). The experimenters monitored the subjects’ trauma exposure on a weekly basis with the New South Wales Fire Brigade’s emergency calls database. In total, 71 subjects (85%) were reevaluated within 12 months of the commencement of active firefighting duty. Nonparticipation occurred because one participant had moved to a rural area, two were receiving workers’ compensation related to a physical injury, and 10 refused to be reassessed. The subjects who participated in the follow-up assessment did not differ from the nonparticipants on age or initial psychometric responses.

At follow-up, 35 subjects were reassessed between 2 and 28 days (mean=14.0, SD=8.2) after experiencing a trauma and were classified into the trauma-exposed group. Thirty-six subjects at follow-up had not been exposed to a trauma and were classified into the non-trauma-exposed group. An event was defined as a stressor if it satisfied the DSM-IV description of criterion A of the PTSD criteria. The traumatic events related to the fire brigade experienced by the subjects in the trauma-exposed group included motor vehicle accidents (N=27, 77%), fires (N=6, 17%), and attending the scene of a suicide (N=2, 26%). One subject in each group was taking antihypertensive medication at the time of both assessments.

At the initial assessment, a master’s level psychologist administered the subjects the Structured Clinical Interview for DSM-IV (SCID) (21) to look for current axis I disorders and the Clinician Administered PTSD Scale (22, 23) to index current posttraumatic stress symptoms. The subjects were also administered the Traumatic Events Questionnaire (24) to assess prior traumatic events, the trait version of the State-Trait Anxiety Inventory (25) to assess trait anxiety, the Beck Anxiety Inventory (26) to index state anxiety, the Beck Depression Inventory II (27) to index depressive symptoms, the Dissociative Experiences Scale—Taxon (28), a subscale of the Dissociative Experiences Scale (29) that measures pathological dissociation, and the Alcohol Use Disorders Identification Test (30) to assess current alcohol use.

At time 2, the subjects were readministered the SCID, the Beck Depression Inventory II, the Beck Anxiety Inventory, the trait version of the State-Trait Anxiety Inventory, and the Alcohol Use Disorders Identification Test. In addition, they were administered the Acute Stress Disorder Interview (31) to assess acute stress disorder, the Impact of Event Scale (32) to assess intrusive and avoidance symptoms, the Peritraumatic Dissociative Experiences Questionnaire (33) to index dissociative responses during and immediately after the trauma, and the Social Desirability Scale (34) to assess the subjects’ need to respond to self-report items in culturally acceptable ways. Trauma-exposed subjects were also asked to rate the severity of the event they were exposed to (0=not at all traumatic, 100=very traumatic). (The non-trauma-exposed subjects were administered the Acute Stress Disorder Interview, the Impact of Event Scale, and the Peritraumatic Dissociative Experiences Questionnaire in relation to the most stressful event they had experienced during the preceding 4-week period.)

Experimental stimuli consisted of 15 consecutive 100-dB (sound-pressure level), 500-msec, 1000-Hz pure tones, with instantaneous rise and fall times. The tones were generated by a Coulbourn audio source module (V85-05, Coulbourn Instruments, Allentown, Pa.) and a stereo amplifier and were presented binaurally through stereo headphones. A sound level and an artificial ear were used for calibration of stimulus intensity. Intertrial intervals were randomly selected by a computer and ranged from 30 to 55 seconds.

Physiological data, consisting of left orbicularis oculi EMG and skin conductance measures, were recorded employing a Coulbourn lablinc V system. EMGs were recorded through 4-mm (sensor diameter) silver/silver chloride surface electrodes filled with Microlyte electrolytic gel (Coulbourn Instruments) and attached with adhesive collars over the left orbicularis oculi, in accordance with recommendations by Fridlund and Cacioppo (35). The raw EMG signal was amplified by 50,000 and filtered to retain the 90–1000-Hz frequency range by a Coulbourn isolated bioamplifier (V75-04). The EMG signal was rectified and integrated by a Coulbourn multifunction integrator (V76-23A) by using a 10-msec time constant. Skin conductance was measured directly by a Coulbourn isolated skin conductance coupler (V71-23) by using a constant 0.5 V through 8-mm silver/silver chloride electrodes. Electrodes were filled with Microlyte gel and attached with adhesive collars to the subjects’ nondominant distal phalanges of the index and middle fingers, in accordance with published guidelines (36). The saline concentration of Microlyte gel (1.34 mol) is not near saturation point, in compliance with published recommendations (37, 38); however, it is higher than the recommended concentration for skin conductance recording (0.05 mol [36]). The Coulbourn instruments were interfaced with a computer through a Coulbourn general-purpose lablinc port (V19-16).

All procedures were conducted in a 3×2.5-m temperature-controlled room and connected by wires to an adjoining room in which the experimental apparatus was located. The subjects were seated in a comfortable armchair and monitored by an unobtrusive video camera. After completion of the written informed consent procedures and the initial assessment, the psychophysiological recording equipment was attached. The subjects were asked to sit quietly for a 5-minute resting period, during which time the physiological measures were sampled continuously at the rate of 2 Hz. After the baseline period, the subjects’ basal tonic level was balanced with a voltmeter. The subjects were asked to rate their current mood (0=very happy, 12=very sad) and anxiety level (0=very calm, 12=very anxious) and were then given the following instructions, based on those used by Shalev et al. (4): “Shortly you are going to hear a series of sounds. Please sit quietly and listen to the sounds as they come. Keep your eyes open throughout the entire procedure, which will last approximately 15 minutes.” Physiological analogue signals were digitized at 1 kHz throughout the startle procedure.

Peak magnitude of the orbicularis oculi EMG response to the acoustic startle probe for each trial was calculated by subtracting the average EMG level for the 20 msec after tone onset from the maximum level within 21 to 200 msec of tone onset. Trials during which a voluntary or spontaneous blink occurred within 20 msec of the stimulus onset were omitted from further analysis because of evidence that blink reflexes typically do not occur within 20 msec (39, 40). In total, 3% of eye blinks were rejected because of excessive noise during the baseline period (e.g., spontaneous blinks or unusually high amounts of integrated EMG during baseline) or because the onset of EMG eye blink response began less than 20 msec after tone onset. Rejected trials were treated as missing data and were assigned a mean response score based on the trial preceding and following the missing value. There were no more than three missing EMG values within either assessment for any participant. Trials with no perceptible eye blink reflex (10% of all trials) were assigned a magnitude of zero and were included in the analysis, following a protocol outlined by previous investigators (40, 41). A skin conductance response score was calculated for each trial by subtracting the average skin conductance level for the 1 second immediately preceding tone onset from the maximum level within 1 to 4 seconds after tone onset (4, 5). To reduce the variance associated with unusually large responses, square root transformations were performed on the EMG and skin conductance response scores before analysis (4, 5).

Habituation was assessed by using two methods (4, 5). First, a measure of relative habituation was computed from the slope of the regression equation Y=bX+a for trials 2–15, where Y is the square root of the response score and X is the log of the trial number. The first trial was excluded from computation of the habituation curve because its skin conductance response can be smaller than the skin conductance response to the second trial (5). Second, the number of trials taken to reach a criterion of two successive nonresponse trials was computed for each participant for EMG and skin conductance responses as a measure of absolute habituation. A nonresponse trial was defined for EMG as a response score ≤3.0 μV and for skin conductance as a response score ≤0.05 μS (cutoff values were based on untransformed data) (5). (There were no perceptible eye blinks below 3.0 μV with the hardware settings employed in this study.) Trials to nonresponse scores ranged from 0 (nonresponses to the first two trials) to 14 (criterion of two successive nonresponse trials not met).

Group mean demographic and psychometric data at times 1 and 2, with analyses of variance and analysis of covariance (ANCOVA), are presented in Table 1. Planned multiple comparisons (that adopted an adjusted alpha of 0.01) at time 1 indicated that the trauma-exposed and non-trauma-exposed groups did not differ in terms of age, education, trauma history, or scores on the Clinician-Administered PTSD Scale, the trait version of the State-Trait Anxiety Inventory, the Dissociative Experiences Scale—Taxon, the Beck Depression Inventory II, or the Alcohol Use Disorders Identification Test. As can be seen, the trauma-exposed group reported higher Beck Anxiety Inventory scores at time 1 than the non-trauma-exposed group, and the mean number of days between time 1 and time 2 assessments was significantly greater in the non-trauma-exposed than the trauma-exposed group. Accordingly, an ANCOVA that entered assessment interval as a covariate was employed to analyze the time 2 psychometric data. Analyses indicated that the trauma-exposed group had higher scores than the non-trauma-exposed group on the Impact of Event Scale intrusion score, the Impact of Event Scale total score, the Peritraumatic Dissociative Experiences Questionnaire score, and the Acute Stress Disorder Interview score. ANCOVAs performed on time 2 Beck Anxiety Inventory scores, with control for both the group difference in assessment interval and time 1 Beck Anxiety Inventory scores, revealed no difference between the trauma-exposed and the non-trauma-exposed groups. At time 2, no subjects met criteria for acute stress disorder.

Group mean resting physiological levels at time 1 and 2 are presented in Table 2. To assess the relationship between baseline physiological measures at times 1 and 2, a series of two-by-two (group-by-assessment) repeated-measures ANCOVAs were conducted separately for each dependent variable (skin conductance and EMG), with control for the difference between groups in assessment interval. There were no significant group or assessment main effects or group-by-assessment interactions for either physiological variable.

Group mean physiological responses to the 15 tone trials at times 1 and 2 are presented in Figure 1 and Figure 2. Group mean physiological responses averaged across the 15 trials, slopes, and trials to nonresponse at times 1 and 2 are presented in Table 2. To assess the relationship between physiological responses at times 1 and 2, a series of two-by-two (group-by-assessment) repeated-measures ANCOVAs were conducted on mean response, response slope, and number of trials to nonresponse for EMG and skin conductance, with control for the difference between groups in assessment interval. Analyses revealed no significant group or assessment main effects or group-by-assessment interactions for any physiological variable.

A series of stepwise multiple regression analyses were performed to identify the time 1 predictors of the time 2 psychophysiology measures. Pearson’s product-moment correlations, shown in Table 3, were calculated for the trauma-exposed group between selected time 2 psychophysiology-dependent variables and time 1 psychometric and psychophysiological variables to determine a subset of potential predictors for the regressions. To predict each time 2 psychophysiological measure, the psychometric and physiological time 1 variable that correlated the highest with the dependent variable was entered with assessment interval into the regression equation. This approach for restricting the battery of potential predictors for regression analyses has been employed in previous prospective research (42).

To predict time 2 baseline skin conductance level in the trauma-exposed group, time 1 resting skin conductance level, Dissociative Experiences Scale—Taxon score, and assessment interval were entered into the regression. Resting skin conductance level explained 12% of the variance in time 2 skin conductance level (β=0.38, SE=0.13, t=2.4, df=32, p<0.05), and Dissociative Experiences Scale—Taxon score accounted for a further 10% (β=–0.36, SE=0.02, t=–2.3, df=31, p<0.05). To predict time 2 skin conductance mean response, the time 1 variables entered into the regression equation were resting skin conductance level, Beck Depression Inventory II score, and the interval between assessments. Of these variables, resting skin conductance level accounted for 22% of the variance (β=0.50, SE=0.01, t=3.2, df=32, p<0.005). The subset of time 1 variables entered into the regression equation to predict time 2 number of skin conductance trials to nonresponse was trials to skin conductance nonresponse, years of education, and the interval between assessments. Time 1 number of trials to skin conductance nonresponse was the only significant predictor and explained 14% of the variance (β=0.41, SE=0.14, t=2.6, df=32, p<0.05).

To predict time 2 mean EMG response in the trauma-exposed group, the time 1 variables entered into the regression equation were mean EMG response, Alcohol Use Disorders Identification Test score, and the assessment interval. Of these variables, mean EMG response was the only significant predictor, explaining 61% of the variance (β=0.79, SE=0.11, t=7.3, df=32, p<0.001). To predict time 2 number of EMG trials to nonresponse, the time 1 variables entered into the regression were the number of EMG trials to nonresponse, prestartle anxiety rating, and the assessment interval. Time 1 number of trials to EMG nonresponse was the only significant predictor and accounted for 16% of the variance (β=0.43, SE=0.16, t=2.7, df=32, p<0.05).

An identical series of multiple regression analyses was performed for the non-trauma-exposed group, using the same predictors employed for the trauma-exposed group analyses. Time 1 number of trials to skin conductance nonresponse was the only significant predictor of time 2 trials to skin conductance nonresponse, explaining 18% of the variance (β=0.45, SE=0.15, t=3.0, df=34, p<0.01). The only significant predictor of time 2 mean EMG response was time 1 mean EMG response, which accounted for 44% of the variance (β=0.68, SE=0.13, t=5.4, df=34, p<0.001). In the non-trauma-exposed group, there were no significant predictors of time 2 resting skin conductance level, mean skin conductance response, or trials to EMG nonresponse from the time 1 variables entered into the regression equations.

Because of the absence of acute stress disorder cases, a dimensional approach was taken to assess the relationship between pretrauma variables and posttrauma psychopathology. A stepwise multiple regression analysis was performed to identify the time 1 predictors of time 2 Impact of Event Scale total score in the trauma-exposed group. Table 3 shows the Pearson product-moment correlations between the Impact of Event Scale score and the time 1 psychometric and psychophysiological variables in the trauma-exposed group. A subset of potential predictors was chosen for the regression, entering the psychometric and physiological time 1 variable that correlated the highest with time 2 Impact of Event Scale total score. The time 1 variables entered into the regression were mean skin conductance response, the score on the Alcohol Use Disorders Identification Test, and the assessment interval. Of these variables, the Alcohol Use Disorders Identification Test score accounted for 23% of the variance (β=0.48, SE=0.18, t=3.6, df=33, p<0.005), and mean skin conductance response accounted for a further 18% of the variance (β=0.42, SE=2.0, t=3.2, df=32, p<0.005). (With resting skin conductance level entered into the regression to control for baseline arousal, the Alcohol Use Disorders Identification Test score and skin conductance response for 15 trials remained the only significant predictors of Impact of Event Scale total score.) With the same variables entered into a stepwise multiple regression analysis for the non-trauma-exposed group as a validity check, there were no significant predictors of time 2 Impact of Event Scale score.