Physiological Predictors of Response to Exposure, Relaxation, and Rescripting Therapy for Chronic Nightmares in a Randomized Clinical Trial

Participants

See Figure 1 for the flow chart indicating attrition. Fifty-eight participants met the inclusion criteria and were invited to participate in the treatment study. Eleven individuals refused and did not enter the study. Forty-seven individuals began the study and were randomly assigned to either the treatment condition (n= 24) or the delayed treatment control condition (n = 23). Of these, 40 were randomized to study arms that included psychophysiology assessment (initially the study included a treatment arm that did not include physiological assessment, but this arm was abandoned in order to focus on physiological predictors). Twelve participants (26%) did not complete the Time 2 assessment, and these dropouts did not differ across condition, χ² (1, N = 47) = 0.34, p = ns. The 18 people in the control group who completed the second initial assessment were invited to participate in the treatment, and 13 did so. A total of 30 individuals (treatment group and delayed treatment group) were treated, and 24 completed the 6-month follow-up.

Of the 47 participants who entered treatment, the majority were Caucasian (81%; treatment = 91.7%, control = 69.6%), female (75%; treatment = 66.7%, control = 82.6%), married (42.6%; treatment = 29.6%, control = 56.5%) and reported at least some college education (80%; treatment = 70.9%, control = 82.6%). Most were employed (treatment = 54.2%, control = 73.9%). Average age was 47 years (SD = 38.49; M_treatment = 38.80, M_control = 38.17), and average household income was $60,115 (SD = $70,736; M_treatment = $52079, M_control = $67,453). Participants reported a mean of 5.5 traumatic events (SD = 2.75 Range = 1-11). The most frequent types of trauma reported were unwanted sexual contact (59.6%), serious accidents (57.4%), and physical assault with a weapon (57.4%). Six individuals (13%) reported combat exposure. Although having a PTSD diagnosis was not a requirement of the study, 53.2% (n= 25) of the participants met criteria, including 54.2% (n= 13) of the treatment group and 52.2% (n= 12) of the control group, χ² (1, N= 47) = 0.02, ns.On the Clinician-Administered PTSD Scale (CAPS)20, approximately 37% reported moderate, 14% reported severe, and 25% reported extreme PTSD symptoms (M= 56.83, SD= 27.19). Participants reported experiencing nightmares an average of 18 years (SD= 12.41; range 1.5-40) and a mean of 5.92 (SD= 1.74) h sleep/night. Approximately 36% of the sample reported taking over an hour to get to sleep each night. Twenty-six percent reported that their most recurrent and distressing nightmare was exactly like their trauma; 34% reported it was similar to the trauma; 32% reported it was unrelated to their trauma; and 8% did not answer the question.

Measures

Physiological Apparatus, Skin Preparation, and Recording Parameters

A PC with dual 17-inch flat panel monitors and A/D board (National Instruments, PCI-6036E) presented scripts and acquired physiological data. Physiological signals were sampled at 250 Hz and collected/filtered using a Model 15LT Bipolar Amplifier (Grass Technologies; West Warwick, RI) with Quad AC (15A54) and Dual DC (15A12) modules. An adaptor (Grass; Model SCA1) for the 15A12 amplifier was used to measure SCL.

Miniature electrodes to measure left corrugator EMG and left lateralis frontalis EMG were placed according to the recommendations of Fridlund and Cacioppo.30 Facial EMG was amplified (×20,000) and bandpass filtered (10-300 Hz) before being stored digitally (while EMG frequencies can range from 1-1000 Hz, most of the signal power is from 10-150 Hz.31,^I Electrocardiogram (ECG) was assessed from electrodes placed on each forearm and was converted to HR by calculating the distance between R to R intervals. Facial EMG and ECG electrodes were applied by degreasing the skin, slightly abrading the skin using NuPrep gel (≤ 10 KΩ), and filling the electrodes with EC60 gel (Grass Technologies). SCL was recorded from electrodes filled with isotonic paste (EC33; Grass Technologies) that were placed on the distal volar surface of the non-dominant index and middle fingers. The signals for HR, SCL, and full-wave rectified, non-integrated facial EMG were visually inspected for artifacts in 5-sec bins. Bins that did not contain artifacts were averaged to make a 30-sec pre-script baseline period and a 30-sec imagery period, as noted in the previous section. For additional details about the physiological recording procedures, see the previously published studies by Rhudy and colleagues.3,32

Procedure

The study was approved by the University of Tulsa Institutional Review Board. Participants were recruited via flyers, email, and radio ads and were screened over the phone for inclusion criteria (experiencing a traumatic event and having nightmares at least one time per week for the previous month) and exclusion criteria (apparent psychosis or mental retardation, aged less than 18, active suicidality or recent parasuicidal behaviors, or current drug/alcohol dependence). Eligible individuals were invited to participate in the study and were randomly assigned using a random number generator to either the treatment or the wait-list control group. Each assessment point included psychological and physiological assessments that were conducted by trained upper level graduate students supervised by the papers' first two authors (JLD and JLR).

After the initial assessment (Time 1), the control group was not contacted during the period that the treatment group received the 3-week intervention, with the exception of a phone call to schedule a time for the re-assessment. One week following the completion of treatment, both groups were reassessed (Time 2) and the control group was offered the treatment. Follow-up assessments occurred at 3 months (Time 3) and 6 months (Time 4) post-treatment. Although attempts were made to keep assessors blinded to treatment group assignment, in some instances assessors were un-blinded at the one-week psychological post assessment for several reasons, including participant disclosure of treatment condition, potentially influencing scoring of the structured interview of PTSD symptoms at the one-week post-treatment assessment point. However, similar results were found for PTSD scores in the successfully blinded previous RCT,7 results were consistent with the one-week physiological assessments which were all blinded; and as reported above, interrater reliability PTSD diagnosis was excellent (κ = 0.87; 97.3 % agreement). At each assessment point, participants were provided compensation.

Intervention

Exposure, relaxation, and rescripting therapy (ERRT) is a cognitive behavioral treatment for trauma-related nightmares and is described in detail elsewhere.7,28 ERRT is conducted for 2 h once a week for 3 consecutive weeks and may be used in either an individual or group format. Treatment consists of psychoeducation about trauma, PTSD, and nightmares, relaxation training, modification of sleep habits, written and verbal exposure to the nightmare, rescription of the nightmare based on trauma-related themes (i.e., power, trust, intimacy, esteem, safety),33 and imagery rehearsal of the rescripted dream each night prior to going to sleep.

Data Analytic Strategy

Data were screened for any potential deviations for normality. Most variables were reasonably normal (skewness statistic divided by the standard error of skewness < 1.96). Not surprisingly, the variables that deviated the most from normal were the physiological reactivity variables. To determine whether non-normality (outliers in particular) caused problems for the models, we carefully examined scatterplots between each of the physiological variables and the psychological outcome variables. Using this approach, we did not find that any outliers had particularly strong statistical influence (i.e., outliers in X and Y space) that might be driving the relationships we observed in our analyses. This bolstered our confidence in proceeding with our statistical analyses.

Initially, ANOVA and χ² tests were used for univariate analyses to assess differences between groups at baseline assessment (Time 1). Treatment response on psychological variables was assessed using multilevel models (SPSS MIXED procedure). Group (immediate treatment vs. delayed treatment) was entered as a nominal between-subjects variable and Time (pre-treatment vs. post-treatment) was entered as a nominal within-subjects variable. Intent-to-treat analyses were conducted on all participants who were randomized. To provide a conservative estimate of initial treatment response for the intent-to-treat analyses, initial baseline scores were carried forward for those who dropped out of the study. Completer analyses were then conducted by selecting only those participants who completed either the post-treatment assessment (treatment group) or second baseline assessment (control group). To assess the magnitude of treatment response for the intent-to-treat and completer analyses, Cohen's d effect sizes were calculated for both groups using baseline and post treatment pooled standard deviations. The overall treatment effect reported is the Cohen's d_treatment minus the Cohen's d_control.

Data from immediate treatment and delayed treatment groups were analyzed for pre-treatment differences. The only difference found was nightmare severity (see Results section), so this variable was controlled for in all analyses, including the follow-up analyses where the groups were combined (N = 47). These data were analyzed using multilevel models (SPSS MIXED procedure) that entered Time (pre-treatment, post-treatment, 3 months, and 6 months) as a nominal within-subjects variable. End state functioning was assessed utilizing only those individuals who completed the 6-month follow-up assessment (n= 24). Effect sizes were calculated with Cohen's d.

Multilevel models examined whether changes in physiological reactivity to nightmare imagery were associated with treatment response on the psychological variables. These analyses included all 40 participants who completed at least one physiological assessment, although one person was excluded because of equipment problems. Because these analyses were largely exploratory, we limited the number of analyses to only those 7 variables that demonstrated significant change in both intent-to-treat and follow-up analyses (see Results section) and health symptoms (PILL) because previously published results found that health symptoms may be related to physiological reactivity to nightmare imagery.3 Therefore, 8 multilevel models were conducted; each included the 4 physiological reactions to nightmare imagery as time-varying covariates. These models controlled for medications that could potentially affect physiological responding (e.g., those that act on cholinergic or adrenergic receptors). No medication variable was a significant predictor in any model, except in the prediction of depression symptoms (not surprisingly, medications that agonize α- or β-adrenergic receptors were more likely to be taken by those more depressed). Thus, to simplify the presentation of the results, medication covariates are not reported in the table of results. Significance of individual physiological predictors in the multilevel models was tested using fixed effects tests (i.e., F tests). The significance of the overall multilevel model was tested using a χ² difference test that compared the 7 predictor models with the no predictor “null” model. Variance explained by each multilevel model was calculated from η².34 An autoregressive covariance structure (AR1) was used to model the error structure for repeated measurements in all multilevel models. Follow-up tests for all analyses were conducted using Bonferroni adjusted mean comparisons to control for family-wise type I error rate. Significance level was set at p < 0.05 (2-tailed).

Baseline Analyses

The treatment group (M = 3.29, SD= 0.62) reported greater severity of nightmares at baseline than the control group (M = 2.91, SD= 0.65), F_1,47 = 4.13, p < 0.05, so nightmare severity was entered as a covariate in subsequent analyses. The completer group (M= 14.89, SD= 2.49) reported more education than the dropout group (M= 13.17, SD= 2.69), F_1,47 = 4.08, p < 0.05. No other differences were found between groups.

Clinical Significance

Clinical significance was assessed for the 24 participants who completed the 6-month assessment. For nightmare frequency we determined that an absence of nightmares in the past week based on the TRNS constituted a good response to the treatment7; 71% achieved this, reflecting a 58% decrease in number of any nightmares in the past week from baseline to 6-month follow-up. It is important to note that we assessed anynightmares, not just the nightmare targeted in treatment. Approximately 63% of participants had a reduction in CAPS score > 10 points, which is generally considered to be a clinically significant change.35,36 Further, 75% (N =9) of those diagnosed with PTSD at baseline lost their diagnosis. For sleep quality, 9 (37.5%) participants achieved a score ≤ 5 on the PSQI.

We also examined changes in quality of life at the 6-month assessment via the PTCSS (Table 3).28 The majority of participants reported that each area assessed was somewhat or much better following treatment. Further, while participants reported some differences in how helpful they found the treatment techniques, the component with the highest percentage of people stating that it was very helpful was writing out the nightmare (78.3%).

Treatment Compliance

Treatment compliance was assessed in 2 ways. First, the percentage of between session homework assignments completed was determined. Individuals who completed treatment and at least one follow-up assessment averaged 78.40% (SD= 23.86) compliance with weekly homework. Second, participants were asked at the 6-month assessment how closely they followed the techniques learned; nearly 83% of the sample reported “mostly” or “completely” following the techniques learned in the study.

Associations between Treatment Response and Physiological Reactivity to Nightmare Imagery

Results from these multilevel models are presented in Table 4. The χ² difference tests indicated all 8 models were an improvement over the intercept only models. Decreases in HR reactivity over time were associated with improvements in sleep quality and hours of sleep. Decreases in SCL reactivity over time were associated with improvements in nightmare severity, PTSD symptoms in the past month (CAPS), sleep quality, and fear of sleep. Facial EMG was generally unrelated to psychological outcomes, except that decreases in corrugator EMG over time were associated with improvements in self-reported health. No physiological variable was significantly associated with nights with nightmares per week or depression. It is noteworthy that the variance explained by those models with significant predictors tended to be large according to η² values, ranging from 17% variance explained (nightmare severity) to 37% (sleep quality).

Limitations

One limitation of this study was the inconsistent blinding for the one-week post assessment and the potential resulting bias of the score of CAPS assessment. However, the fact that the results of this study mirror those of the previous RCT7 and the psychological indices were consistent with the blinded physiological assessments tempers these concerns. The trial included a small sample size comprised primarily of Caucasian women which may limit the generalizability of the results. The 26% dropout rate in this study matches that found in the initial RCT and is fairly typical for trauma-related therapies.40 While the rate in the current study is not exceptional, more research is needed on determining why participants drop out of treatment studies and developing additional strategies to try to increase retention. The use of a wait-list control group does not allow us to speculate about the relative efficacy of ERRT compared to other treatment approaches.41 Future studies will need to conduct direct comparisons to other viable interventions. Additionally, while the study used gold standard interview tools and physiological assessments, future studies should consider including physiological indices of sleep functioning (i.e., polysomnography or actigraphy) and an established measure of clinical significance.

Overall, current results support previous research suggesting that ERRT is an efficacious treatment for trauma-exposed individuals with chronic nightmares and advances the knowledge base in the treatment of nightmares by including an evaluation of the association of physiological and psychological changes over time.