Abstract
Keywords
Posttraumatic stress disorder (PTSD), as defined in DSM-IV [1], may occur after a traumatic event which produced or threatened serious consequences and to which the response was intense fear, helplessness or horror. The traumatic event is persistently re-experienced, there is persistent avoidance of stimuli associated with the trauma or a general psychological numbing, and there are persistent symptoms of increased arousal. Diagnosis is primarily made by clinical interview but a number of self-report scales have been developed for screening, diagnosis and for measuring change; for example the Davidson Trauma Scale (DTS) [2] and a PTSD subscale derived from the Symptom Check List (SCL-90-R) [3]. There has also been interest in measuring physiological aspects of PTSD [4–6]. In general, the physiological measures have been studied to better understand the disorder but they also offer the possibility of measures not susceptible to misreporting, which is a particular problem if financial benefits follow diagnosis, as with war pensions for veterans. Frueh et al. [7] review apparent over-reporting of psychopathology in combat veterans evaluated for PTSD.
This study was designed to investigate self-report scales and physiological indicators of PTSD in a group of Vietnam War veterans with a range of PTSD symptoms, from none to multiple and severe. If the physiological measures showed reasonable reliability, validity and sensitivity to change, then they could be considered a useful addition to the self-report measures both for initial assessment and in monitoring change in clinical trials, particularly if they were relatively cheap, easy to administer and acceptable to veterans or others who had experienced trauma.
The self-report measures used in this study were the DTS [2], and the PTSD scale of the SCL-90-R [3]. The DTS was chosen as a reliable and valid scale, linked to DSM-IV criteria, with the advantage of brevity and ease of self-administration. For each of the 17 symptoms, frequency and severity are rated from 0 to 4. For comparison, the SCL-90-R was given as a well-established measure of psychiatric morbidity, from which a PTSD subscale had recently been developed.
Exaggerated startle response has been found in combat veterans and victims of other trauma [8–12]. These startle studies have used auditory stimuli, usually a burst of white noise, and have mostly found increased startle reflex [6, 13]. One problem with an auditory stimulus in combat veterans is the variable degree of hearing loss. In a study of Australian Vietnam veterans [14] 30% reported hearing loss. To avoid this problem the stimulus used in this study was electrical stimulation of the supraorbital branch of the trigeminal nerve, a stimulus used in routine neurophysiological clinical testing of the blink reflex.
Mason and colleagues have shown alterations in thyroid function with persistent and disproportionate elevations in both total and free triiodothyronine (T3) in relation to free thyroxine (T4) [15–17]. The elevated total-T3/free-T4 ratio is a simple, reliable and cheap investigation requiring only a blood test, so it appeared promising as another physiological measure of PTSD. Because alcohol abuse and dependence is strongly associated with PTSD [18], and could confound relationships with other variables, the Alcohol Use Disorder Identification Test (AUDIT) [19] was also given to the participants. The AUDIT has been widely used [20–23], including previous use in combat veterans [24].
Method
Sample
The addresses of all New Zealand Army and Air Force Vietnam War veterans living in the North Canterbury region of New Zealand were on an existing database [25]. Following approval from the Canterbury Ethics Committee, the Department of Defence sent each of the approximately 100 veterans an information sheet and letter inviting them to participate. Fifty responded but 14 who lived too far from Christchurch were not included and one withdrew when contacted by telephone, leaving 35 in the study, which was close to the planned sample size of 40. With n = 35 there is power of 80% with α 2 = 0.05 to detect correlations of 0.45 and change over time of 0.48 standard deviations for difference scores.
Measures
The questionnaires used for self-report of PTSD were the DTS [2], the SCL-90-R [26], and the AUDIT [19]. There were also questions about age, sex, source of income, service in Vietnam and prior diagnosis of PTSD. Thyroid function was measured as total-T3 and free-T4 and the ratio of total T3 to free T4 was calculated. The eye blink electrical startle response was measured from below the obicularis oculi muscle by surface electrodes positioned 1 cm below the eye, referred to an electrode 1 cm lateral to the canthus. Stimulation of the supraorbital branch of the trigeminal nerve was performed at five intensities (10, 20, 30, 40 and 50 mA) presented in ascending order. The amplitude and the area under the response curve were recorded but only results for area under the curve are reported in this study as amplitude was measured to only 1 figure in contrast to 2 or 3 figures for area, and also rose less consistently with intensity.
Procedure
The main set of assessments took place in April and May 1998. An interviewer met with each participant and administered all the questionnaires and arranged for the startle test and blood tests within two to three working days, often on the same day. The DTS was given to the participants for them to fill in 1 week later, in order to measure test–retest reliability. If no questionnaire arrived within a week of the due date, the interviewer contacted the veteran by telephone to clarify
any questions and encourage completion (only six veterans required this phone call, of whom two never returned the questionnaire). The mean time from initial assessment until the repetition of the DTS was 9.4 days (SD = 4.6 days). Follow-up occurred in October and November 1998, about 6 months after the initial assessment. Because of the results from the initial assessment the thyroid function measures were omitted and the startle response test was revised by dropping the higher intensity stimulus levels which many participants found intolerable. The AUDIT and the SCL-90-R were administered only at initial assessment as they were included solely for comparison with the other measures.
Statistical methods
Pearson product moment correlations, scatterplots and t-tests were used. Data was entered in EPI-INFO, Version 6 and analysed in that package and in SAS Version 8.2. There was a small amount of imputation using the integer closest to the mean values for missing items in the 34-item DTS, which the participants filled out themselves. The number of imputed items was 7/1190 for the initial assessment, 11/1122 for the second (which was filled out at home) and 0/1088 at follow-up. A number of veterans left severity blank when they reported zero frequency for a symptom; these blanks were scored as zeros. One item was imputed in the AUDIT.
Results
All the participants were male with a mean age of 53.4 years (SD = 5.1, range = 46–69). Their service in Vietnam ranged from 5 months to 48 months with a mean of 12.6 months (SD = 7.7). More than 90% had completed their Vietnam service in the four years between 1968 and 1971, the period of New Zealand's major commitment to that war. Fifteen reported a diagnosis of PTSD. For 14 participants this diagnosis was confirmed from clinical records by one of the authors (ADM) who had assessed 20 of the participants (10/14 were current patients and 2/14 had been examined within 3 months prior to the start of the study). One veteran reporting a diagnosis of PTSD had not been seen by ADM. Of the 35 veterans given the main set of assessments, 33 filled out the DTS again one to two weeks later, and 32 were available again at follow-up, after about six months.
Initial assessment
The means for the main set of assessments, plus the repeat of the DTS one to two weeks later, are given in Table 1. The self-report measures show that the sample did contain veterans with a wide range of PTSD severity, as was desired to enable the detection of correlations. Fourteen of the 15 men who reported a PTSD diagnosis were above the recommended cut-off on the DTS at initial assessment and the one below the cut-off was above it 8 days later and at follow-up. Four veterans who did not report a diagnosis of PTSD did score above the cut-off on the DTS at initial assessment. Two of these had been diagnosed with PTSD by one of the authors (AMD), one was subsequently seen and assessed as having PTSD, and one was not seen. More veterans were above the recommended cut-off on the DTS than on the SCL-90-R PTSD scale (51% versus 37%) but the two scales were highly correlated (Table 2, r = 0.89) even although the DTS is overtly about response to trauma whereas the SCL-90-R questions are a general assessment of psychological symptoms.
Initial assessment plus the repeat of the Davidson Trauma Scale (DTS) about one to two weeks later (n = 35)
Correlations † between initial assessment variables plus the repeat of the Davidson Trauma Scale (DTS) about one to two weeks later
Indeed, the correlations between the three self-report measures of PTSD in Table 2 are all very high. The test–retest correlation between the initial DTS and its repeat one to two weeks later could hardly be higher. Nonetheless, such a high correlation does not preclude a mean shift. The mean increased by 8.2 (SD = 15.1), which was significant (t = 3.12, df = 32, p = 0.004). There are several possible reasons for this shift upwards in the frequency and severity of PTSD symptoms. One is that the experience of filling it out may have made veterans more conscious of the extent to which they were experiencing symptoms so that they reported more accurately the second time. Also, they may have responded differently with an interviewer in the room, in contrast to filling in the questions in their own time. Alternatively, the observed change may have been a real one due to the increasing impact of the imminent Vietnam Remembered March, the first ‘homecoming’ march for New Zealand veterans, which took place in June 1998, a few weeks later. Participants had all been invited and the interviewer reported that many said that they were discussing the march with other veterans.
To varying degrees the veterans disliked the blood test for thyroid function and found the electrical startle response difficult or distressing, although all but two did continue to the highest of the five intensity levels at the initial assessment. As the intention was to find measures which would be acceptable for screening or repeated use in clinical trials, it was decided to use data from only the lowest two levels (10 and 20 mA), as these were reasonably well-tolerated. Because the distribution of the area under the curve was skewed, it was expressed as log 10. For those who did complete all five levels the correlation between the log of the total area under the curve for five levels and that from the lowest two levels was 0.87.
As Table 2 shows, the physiological measures did not correlate highly with each other or with the self-report measures of PTSD. Separating the left side and right side log startle responses provided an internal consistency measure of reliability of 0.43. As hyperarousal is only one of the symptom clusters in the DSM-IV criteria it could be that startle was elevated but not the overall severity of PTSD. Within the DTS, items 12–16 relate to hyperarousal (Cluster D), with item 16 asking ‘Have you been jumpy or easily startled?’ Therefore, scores for these items were compared with startle responses. The correlation between log startle and the total score for hyperarousal symptoms was 0.21, which was very close to that between log startle and the total DTS score (0.18). The scores on item 16 were very skewed with 18/35 reporting no symptom. Item 16 scores correlated only 0.11 with log startle and there was no significant difference in log startle between those with or without the symptom (means = 1.61, 1.51; SDs = 0.29, 0.25; t = 0.96, df = 32, p = 0.35). Any relationship between electrical startle, as measured in this study, and self-reported hyperarousal is weak.
In the above analyses only four trials were used, one on each side of the head at each of two intensity levels. The Spearman–Brown formula indicates that if twice as many trials had been given then the test–retest reliability would have increased from 0.43 to 0.60 [27]. However, the standard correction for attenuation [27] shows that even unobservable error free ‘true scores’ for both the DTS and the electrical startle would have produced at most a correlation of 0.28, using the first DTS score, and 0.41 using the repeat DTS score. Therefore, the lack of relationship between the measures cannot be attributed primarily to measurement error.
The thyroid function measures were within the normal range for total T3 and one participant was below the normal range for free T4 (normal range 55 to 160 nmol/L, value 48). Correlations with other measures were low (– 0.23 to 0.26).
Table 1 shows that on the AUDIT nearly half of the veterans reported hazardous drinking in the last year. Nonetheless the correlations with self-report measures of PTSD were low (0.29 to 0.34). Partialling out the effect of alcohol consumption, as measured by AUDIT scores, only slightly reduced the already low positive correlations between the physiological tests and the self-report tests by between 0.01 to 0.06 and increased the negative correlations to 0.00 to 0.03. What little relationship was observed between these two sets of measures was not markedly confounded by the degree of hazardous alcohol consumption.
Follow-up
The main purpose of the follow-up in November 1998 was to see if self-reported PTSD symptoms were alleviated or exacerbated after the Vietnam Remembered March held in June. The electrical startle was also repeated, using only the two lowest levels, in spite of its poor performance in the main set of assessments. It was thought that factors other than PTSD might affect the absolute level of the startle responses but that it might change along with PTSD severity and hence be of use in clinical trials, although not in screening. Thirty-two of the 35 veterans participated again at follow-up, one of whom had missing data on one startle intensity level at baseline.
At follow-up the mean DTS scores had returned to their original level (mean difference 0.3, SD = 27.0, t = 0.06, df = 31, p = 0.95). The standard deviation of the differences indicates some divergent time trends across individuals but nonetheless the correlation between initial and follow-up scores was 0.73. In contrast the mean log startle response at follow-up was slightly higher than at initial assessment (mean difference 0.10, SD = 0.28, t = 2.04, df = 30, p = 0.05) which could indicate sensitization to the procedure in some individuals. The correlation between initial and follow-up log startle values was 0.39. The correlation between the changes in the DTS scores and the changes in the log startle values was −0.14.
Discussion
The DTS showed high test–retest reliability (0.93) in this study as was also found in the original report (0.86) [2]. It had not previously been compared with the SCL-90-R PTSD scale but given the agreement of both scales with other self-report and clinician rated measures of PTSD [2, 3], the high correlation found between them was not surprising. However, the SCL-90-R PTSD scale appears to have a slightly higher cutoff for PTSD than the DTS, at least in this group of Vietnam veterans.
Mason and Wang [15–17] have consistently found elevated total T3 in veterans with PTSD and no increase in free T4. The total T3 effect observed has been large (d ≈ 1.2 [17]) and correlations with self-report and clinical measures of PTSD have been positive although not large. In contrast, we found that correlations of total T3 with self-report were slightly negative, until the small effect of hazardous alcohol consumption was partialled out when they became effectively zero. The differences between the studies may be due to characteristics of the groups included or the specific self-report measures used. It would be informative to have measures of alcohol consumption in future studies of thyroid function. Even from the results of Wang and Mason [17] the thyroid function correlations were not large enough to indicate that they could be useful for assessment or clinical trials.
The electrical startle, as used in this study, also does not appear useful for assessment or clinical trials. The veterans who participated in this study agreed to be tested in this way but did not like it. Electrical startle would need to be a clear indicator of diagnosis or a guide to treatment to enter into routine clinical practice and it appears to be neither of these. Even repetition of the electrical startle trials to improve the reliability does not seem likely to reveal much relationship between startle and self-report. Reviews of the literature for acoustic startle have shown moderate effect sizes (Cohen's d = 0.47 [5], 0.67 [6]) although not all studies have shown this [12]. Grillon et al. [13] found that acoustic startle responses were facilitated by darkness, and that elevated sensitivity to darkness was associated with combat exposure, regardless of whether or not veterans had PTSD. The absolute levels of acoustic startle were higher under light and dark conditions for veterans with PTSD than for veterans without PTSD or for civilians. Investigations of the startle response will further clarify what it is appropriate to measure and under what conditions. Nonetheless, such work may continue to be of scientific interest rather than providing a clinical tool. As noted by Orr et al. [5], ‘it is clear that not all individuals with PTSD show an exaggerated startle response’.
One limitation of the study reported in this paper is the small sample size, a problem shared by many of the other studies in the literature. A second limitation is that the sample may be unrepresentative, since it came from an approximately 50% response rate. Nonetheless, it was based on a database of veterans and hence is likely to be more representative than most of the clinical samples reported elsewhere.
Conclusions
Self-report is at least consistent. Thyroid function and electrical startle response appear to be of research interest only at this time, as part of attempts to understand the pathophysiology of PTSD. Perhaps auditory startle is more relevant than electrical startle to PTSD, particularly for combat veterans. Of alternative psychophysiological measures investigated in other studies, one of the most promising involves the use of cardiovascular reactivity to script-driven imagery, both from the trauma event and from other stressful events [4, 5]. The cardiovascular response to trauma-related stimuli is reported to have high specificity and sensitivity and to be difficult to fake, thus enabling evaluation of the over-reporting which is of such concern to Frueh et al. [7]. It may be fruitful in future research to concentrate on responses related to the trauma, rather than looking merely at general measures of hyperarousal.
Footnotes
Acknowledgements
Thanks are due to the New Zealand War Pensions Board Medical Research Trust who funded this research, Verna Braiden and to the New Zealand Vietnam War veterans and their families. Dr Wells received salary support from the Health Research Council of New Zealand to provide statistical services.