Journal of Anxiety Disorders 30 (2015) 81–87

Contents lists available at ScienceDirect

Journal of Anxiety Disorders

Between-session and within-session habituation in Prolonged Exposure Therapy for posttraumatic stress disorder: A hierarchical linear modeling approach Rebecca K. Sripada a,b,c,∗ , Sheila A.M. Rauch b,c a

Veterans Affairs Serious Mental Illness Treatment Resource and Evaluation Center (SMITREC), Ann Arbor, MI, United States Veterans Affairs Ann Arbor Health Care System, Ann Arbor, MI, United States c University of Michigan Department of Psychiatry, Ann Arbor, MI, United States b

a r t i c l e

i n f o

Article history: Received 3 October 2014 Received in revised form 1 January 2015 Accepted 4 January 2015 Available online 13 January 2015 Keywords: PTSD PE HLM Extinction Trauma-focused treatment

a b s t r a c t Prolonged Exposure Therapy is a frontline intervention for posttraumatic stress disorder, but the mechanisms underlying its efficacy are not fully understood. Previous research demonstrates that betweenand within-session habituation of fear during exposure is associated with treatment outcome, but these calculations are historically performed with summary statistics such as mean subjective units of distress (SUDS). This question could be better assessed with an analytic technique that uses all SUDS measurements available within sessions. Hierarchical linear modeling was used to investigate the impact of treatment response on SUDS nested within therapy sessions nested within 14 patients. Symptom change (t = −2.43, p = .03) and responder status (t = −2.68, p = .02) predicted slope of SUDS across sessions, but did not reliably predict slope of SUDS within-session, indicating that high responders demonstrated differential between- but not within-session habituation. Thus, individuals who show greater habituation between treatment sessions may be more likely to respond to treatment. Published by Elsevier Ltd.

1. Introduction Prolonged Exposure is a frontline intervention for posttraumatic stress disorder (PTSD) with substantial evidence supporting its efficacy for a variety of patient populations (Powers, Halpern, Ferenschak, Gillihan, & Foa, 2010; Rauch et al., 2009; Schnurr et al., 2007; Tuerk et al., 2011). The factors associated with treatment efficacy are of considerable interest and debate (e.g. Bluett, Zoellner, & Feeny, 2014). Previous research demonstrates that habituation1 of fear is a key process in exposure and is associated with PE treatment outcome. In particular, previous findings have demonstrated that between-session habituation, or a decrease in patient-reported subjective units of distress (SUDS) between sessions, is an important contributor to symptom reduction. In several PE studies, greater reductions in mean and peak SUDS between the first and

∗ Corresponding author at: 2800 Plymouth Road, Bldg. 16, Ann Arbor, MI 48109, United States. Tel.: +1 734 222 7432; fax: +1 734 222 7514. E-mail address: [email protected] (R.K. Sripada). 1 We use the term habituation in this paper to indicate anxiety reduction during exposure, while recognizing that both extinction and habituation processes are active during exposure therapy. http://dx.doi.org/10.1016/j.janxdis.2015.01.002 0887-6185/Published by Elsevier Ltd.

last imaginal exposure session of PE were associated with greater PTSD symptom reduction (Bluett et al., 2014 Gallagher & Resick, 2012; Rauch, Foa, Furr, & Filip, 2004). Greater habituation between imaginal sessions one and two has also been shown to be a marker of treatment response (van Minnen & Hagenaars, 2002). Another study demonstrated that between-session habituation was associated with reduced PTSD symptoms both at treatment endpoint and at one-month follow-up (van Minnen & Foa, 2006). Additionally, a hierarchical clustering analysis revealed that the best treatment outcomes were associated with “high engagers/habituators”– individuals who showed high SUDS during the first imaginal exposure session and gradually declining SUDS over the course of the next six sessions (Jaycox, Foa, & Morral, 1998). Outcomes in this group were superior to the outcomes of the two other classes of response: high engagers/non-habituators and low engagers/non-habituators. These studies provide promising preliminary evidence that differences in SUDS between sessions are associated with positive treatment response. However, several studies of individuals with PTSD (Bluett et al., 2014; Pitman, Orr Altman, Longpre, Poiré, et al., 1996; Pitman, Orr Altman, Longpre, Poire, et al., 1996), and other disorders (Kozak, Foa, & Steketee, 1988; Lang & Craske, 2000; Meuret, Seidel, Rosenfield, Hofmann, & Rosenfield, 2012; Rowe & Craske, 1998; Tsao & Craske, 2000) demonstrate positive treatment

82

R.K. Sripada, S.A.M. Rauch / Journal of Anxiety Disorders 30 (2015) 81–87

outcomes in the absence of between-session habituation. Thus, further research is needed to determine the relationship between treatment response and between-session habituation. Another habituation pattern that may be a beneficial component of PE is within-session habituation, or decline in SUDS over the course of an individual session. One PE study demonstrated that treatment responders showed greater within-session habituation than non-responders while listening to imaginal exposure at home, though not during the therapy session itself (van Minnen & Hagenaars, 2002). Another study found that amongst 106 individuals with various anxiety disorders, greater within-session habituation during the first exposure session was associated with lower risk for treatment dropout (Norton, Hayes-Skelton, & Klenck, 2011). Similarly, during the third exposure session, greater within-session habituation was associated with lower anxiety symptomatology at therapy endpoint (Norton et al., 2011). Thus, there is some evidence to suggest that a pattern of decreasing anxiety during a session of exposure is predictive of good treatment outcome. However, several studies show no evidence for the necessity of within-session habituation (Baker et al., 2010; Bluett et al., 2014; Culver, Stoyanova, & Craske, 2012; Jaycox et al., 1998; Kozak et al., 1988 Meuret et al., 2012; Pitman, Orr Altman, Longpre, Poiré, et al., 1996; Pitman, Orr Altman, Longpre, Poire, et al., 1996; Riley et al., 1995; van Minnen & Foa, 2006; van Minnen & Hagenaars, 2002). Thus, there is mixed support for the association between within-session habituation and positive treatment response. In addition to habituation, several other markers of PE efficacy have been proposed. A recent investigation demonstrated that the perceived helpfulness of imaginal exposure homework had an indirect effect on the relationship between distress reduction and clinical outcome such that individuals with modest reductions in peak distress still experienced clinical improvement if they perceived homework to be helpful (Bluett et al., 2014). Reduced negative cognitions about the world and the self may also mediate treatment response (Foa & Rauch, 2004). Finally, distress tolerance has recently emerged as an important potential mechanism of change in exposure therapy (Craske et al., 2008; Meuret et al., 2012). Fostering distress tolerance during exposure promotes the acquisition of inhibitory learning that weakens or negates the original conditioned stimulus-unconditioned stimulus expectancies (see Myers & Davis, 2002). While all these factors may contribute to successful treatment response, habituation remains the most consistent theoretical construct related to response in exposure therapy. Though several theories have been proposed, the question of what predicts success in PE could be better assessed by using analytic techniques that capitalize on the wealth of data that is available from standard PE treatment. During imaginal exposure in the standard PE protocol, SUDS is collected every five minutes. Given such a large number of data points, past studies (Bluett et al., 2014; Jaycox et al., 1998; Rauch et al., 2004; van Minnen & Hagenaars, 2002) have used peak SUDS, mean SUDS, or another summary measure to estimate habituation, but no study to date has used the full extent of available data to model all of the variability in SUDS during imaginal exposure. The current analyses utilized all the SUDS data available in PE sessions from a small mechanistic study of PE to examine habituation both within and between sessions. Hierarchical linear modeling (HLM) is a multilevel modeling technique that accounts for the inherent nested nature of data generated by treatment studies. Unlike repeated measures ANOVA, HLM allows for different numbers of sessions between patients. In the current study, we modeled SUDS data points (level-one) nested within session (level-two), nested within patient (level-three). We then tested the effects of symptom change and treatment responder status at each level. We hypothesized that high responders

would show greater reduction in SUDS than low responders both within session and between sessions. 2. Material and methods 2.1. Sample and treatment setting The following analyses were conducted as a post-hoc examination of data collected in a translational treatment mechanisms study. For a full description of study procedures and the sample, see (Rauch et al., in press). All Veterans participated in the treatment study between January 2008 and July 2010 (N = 14; 9% Female, 23% African American, 73% White, 5% Asian). Mean age was 32.7 years (SD = 6.9). With regard to combat location, 81% served in Iraq and 33% served in Afghanistan, with some Veterans reporting deployments to both locations. As reported in the primary outcome paper, comorbidity was representative of the OEF/OIF/OND population: 57% had depression or dysthymia, 10% had alcohol abuse, and 29% met criteria for another anxiety disorder at intake. All treatment was conducted by an experienced PE provider (author SAMR) with over 15 years of experience using the protocol. To enhance generalizability, exclusion criteria were minimized and included only contraindications for PTSD treatment and factors that would interfere with the biological mechanisms investigated in the parent study. Exclusion criteria were: (1) level of self-harm risk that requires immediate, focused intervention, (2) unmanaged psychosis or bipolar disorder, (3) alcohol or substance dependence in the past 3 months, (4) working night-shifts, (5) changes to psychoactive medications in the past 4 weeks, or (6) taking medication that makes hypothalamic-pituitary-adrenal axis measures difficult to interpret. Eligible Veterans reviewed consent with author SAMR. Those who were interested signed consent and were randomly assigned to receive 10–12, 80-min sessions of PE or Present Centered therapy, an active control condition. Data were collected with approval from the Institutional Review Board (IRB) for the VA Ann Arbor Healthcare System. 2.2. Measures Clinician Administered PTSD Scale (CAPS; Blake, Weathers, Nagy, & Kaloupek, 1995). The CAPS is an interview measure of PTSD severity with excellent psychometric properties (Blake et al., 1995). Cronbach’s alpha for this sample was .95. We used a stringent classification of responders, such that patients were characterized as high responders only if they experienced a 50% or greater reduction in CAPS over the course of therapy. CAPS was administered at pre-, mid-, and post-treatment. All treatment completers presented for the post-assessment, with the exception of early terminators, who were assessed only at pre- and mid-treatment. Early terminators were assessed via the last-visit carried forward technique. Subjective Units of Distress (SUDs) are self-ratings of distress ranging from 0 (complete relaxation) to 100 (maximum distress) (Wolpe & Lazarus, 1966). SUDS were collected every 5 min during the imaginal exposure portion of the session. Imaginal exposure began at session three and occurred at every session thereafter, with the exception of sessions that included therapist-assisted in vivo exposure. Within-session habituation was operationalized as a drop in SUDS over the course of a session, and between-session habituation was operationalized as a drop in SUDS across sessions. 2.3. Intervention Each patient underwent at least two and up to ten imaginal exposure sessions, and each imaginal exposure session assessed SUDS between two and ten times. Treatment completers were defined as any patient attending at least seven sessions of PE and at

R.K. Sripada, S.A.M. Rauch / Journal of Anxiety Disorders 30 (2015) 81–87

least one additional CAPS following intake. Treatment completers were classified as high responders or low responders, based on CAPS score change over the course of treatment. High responders exhibited at least 50% reduction in CAPS, whereas low responders exhibited less than 50% reduction. In addition, early termination due to response (i.e. prior to session seven) was not operationalized as attrition. One patient met the criterion for early termination (50% reduction in CAPS from baseline before session seven). In sum, ten patients were classified as completers and four as noncompleters. Of the ten completers, seven patients were classified as high responders and three were classified as low responders. 2.4. Data analysis Three-level Hierarchical Linear Modeling (HLM) was used to characterize patterns of within-session and between-session habituation and to investigate the potential associations between these habituation patterns and (1) symptom change, and (2) responder status. HLM is a practical strategy for analyzing effectiveness data because the method does not assume fixed time points of measurement or equal numbers of observations and missing data do not cause special problems (Raudenbush & Bryk, 2002). Specifically, in the current study, HLM was used to investigate variance in SUDS level (level-one) nested within session (Level-2) nested with patient (level-three). Time in minutes was the only predictor included at Level-1. At Level-2, the only session-level predictor was the session number. A random intercept and time slope were included for each session within a patient, enabling estimation of the variance in SUDS means (␶␲00 ) and linear changes over time (␶␲11 ) between sessions within a patient. At Level-3, patientlevel predictors included responder status (high responder or low responder) and symptom change (change in CAPS score). A random intercept and session effect were included for each patient, enabling estimation of the variance in SUDS means (␶␤00 ) and session effects (␶␤11 ) across patients. We used a sequential approach to fitting the models. In Step 1, an unconditional model was fitted and intraclass correlations were calculated in order to assess the variance attributable to each level of the data. Time-in-session (as measured in 5-min increments corresponding to the points at which SUDS measurements were acquired) was added in Step 2 as a level-one, within-session predictor. Deviance likelihood ratio tests were used to assess whether adding parameters improved model fit. In Step 3, weeksin-treatment (i.e. “session dose”) was added to the model as a level-two within-patient predictor. Four final models were then fitted. The first and second models assessed the relationship between CAPS change and habituation patterns (as measured by SUDS change). Models 3 and 4 assessed the relationship between responder status and habituation patterns. In Models 1 and 3, the non-completers were grouped with the low responders. In Models 2 and 4, we excluded the noncompleters. The variances of random errors at Level 1 and random effects at Levels 2 and 3 were estimated. Statistically significant outcomes were qualified using R2 -type effect sizes in a manner consistent with recommended guidelines (Snijders & Bosker, 1994). HLM analyses were conducted using HLM software (Raudenbush, Bryk, Cheong, & Congdon, 2004). 3. Results 3.1. Descriptive statistics Fourteen patients were included in the study. Two noncompleters were removed from the analysis because they both had only one imaginal exposure session, and thus could not be

83

assessed for between-session habituation. This left a final sample of twelve. Based on a cutoff of at least 50% CAPS reduction, seven participants were classified as high responders, three participants were classified as low responders, and two participants were classified as non-completers. The mean number of imaginal exposure sessions attended was 6.1 (SD = 3.0). Treatment with PE was effective, as evidenced by a mean reduction in CAPS of 45.1 (SD = 25.7). Greater CAPS reductions were observed in the high responder group as compared to the low responder group (t = 5.21, df = 9, p = .001, Cohen’s d = 4.2). Mean CAPS change in the high responders was 63.6 (SD = 10.8); in low responders was 29.0 (SD = 4.4); and in non-completers was 4.5 (SD = 6.4). Detailed results are reported elsewhere (Rauch et al., 2015). 3.2. HLM results Models of longitudinal SUDS scores were based on 339 SUDS measurements (level-one) nested within 72 sessions (level-two) nested within 12 patients (level-three). The unconditional model estimated variance components for level-one, level-two, and levelthree units (see Table 1). The value of  was significantly different from zero, indicating the presence of patient-level effects on outcomes. The intraclass correlation (ICC) for between-patient variability was .37, indicating that 37% of variance in SUDS could be accounted for by factors associated with the patient (R2 -between). Such factors could be related to personal characteristics (i.e. age or baseline pathology) or external factors that varied by patient (i.e. seasonal effects). The ICC for session nested within patient was .75, indicating that 75% of variance in SUDS was due to the effects of session nested within patient (R2 -within). In Step 2, time in session predicted SUDS outcome (t = 3.14, p = .003). The modeled coefficient indicated a linear increase in SUDS over the course of a session. SUDS increased 1.62 units on average every 5 min. A deviance likelihood ratio test indicated that allowing the impact of time to vary by session (by adding a random slope factor to the level-two model) improved the fit of the overall model (2 = 36.24, p < .001). AIC and BIC values also indicated improved model fit. Thus, this random slope was retained in the final model. In Step 3, weeks-in-treatment significantly predicted SUDS outcomes (t = −2.47, p = .028). The modeled coefficient indicated linear decreases in SUDS over the course of treatment (across sessions). Time-in-session remained a significant predictor of SUDS. A deviance likelihood ratio test indicated that allowing the impact of weeks-in-treatment to vary by patient (by adding a random slope factor to the level-three model) did not significantly improve the fit of the overall model (2 = 3.09, p = .21). However, given the strong theoretical justification for allowing patients to vary in their degree of between-session habituation, and given that the model comparison test is often overly conservative (West, Welch, & Galecki, 2006), this random slope was retained in the final model. 3.2.1. Model 1: CAPS change and outcomes; all participants With overall treatment effects accounted for, the next step in the model involved adding CAPS change to explore potential associations with within- and between-session SUDS change. CAPS change predicted slope of SUDS across sessions (t = −2.43, df = 12, p = .03), but did not predict overall SUDS level or slope of SUDS within a given session (see Table 2), indicating that symptom change was related to between-session habituation but not differential withinsession habituation or overall SUDS level. The direction of the interaction between CAPS reduction and session was negative, indicating CAPS reductions were associated with a decrease in SUDS between sessions. The proportion of variance among patients in the session effects at level-three was calculated via change in  ␤11 after the addition of the interaction between CAPS and session to

84

R.K. Sripada, S.A.M. Rauch / Journal of Anxiety Disorders 30 (2015) 81–87

Table 1 Summary of the HLM specification steps.

Fixed effects: Level-1 Level-2 Level-3 Random effects: Level-1 Level-2 Level-3

Time Session Intercept 2  ␲00  ␲11  ␤00  ␤11

Deviance Parameters AIC BIC

Step 1 (unconditional model)

Step 2 (time-in-session added)

ˇ

SE

t

ˇ 1.62

SE 0.51

t 3.14**

55.76 Estimate 139.27 215.40

4.47 SD 11.80 14.68

12.49*** 2

14.51

66.78***

4.79 SD 9.98 16.10 3.06 15.67

11.35*** 2

210.69

54.33 Estimate 99.55 259.06 9.35 245.44

460.70***

2920.99 4 2928.99 2944.46

Step 3 (weeks-in-treatment added)

536.73*** 171.40*** 80.76***

2874.67 7 2888.67 2915.72

ˇ 1.52 −3.63 60.63 Estimate 99.29 171.37 9.05 175.30 12.86 2869.24 10 2889.24 2927.50

SE 0.52 1.47 4.45 SD 9.96 13.09 3.01 13.24 3.59

t 2.92** −2.47* 13.63*** 2 237.49*** 161.37*** 25.40** 45.30***

AIC, Akaike information criterion; BIC, Bayesian information criterion. * p < .05. ** p < .01. *** p < .001.

the model. CAPS change accounted for 99% of the variance in slope of SUDS between sessions.

df = 8, p = .005) and slope of SUDS within-session (t = −2.47, df = 8, p = .017), but did not predict overall SUDS level.

3.2.2. Model 2: CAPS change and outcomes; completers only We then ran a model that included only treatment completers. CAPS change predicted slope of SUDS across sessions (t = −3.83,

3.2.3. Model 3: responder status and outcomes; all participants Responder status predicted slope of SUDS across sessions (t = −2.68, df = 12, p = .02), but did not predict overall SUDS level or slope of SUDS within a given session. Compared to low

Table 2 Summary of the full HLM models. All subjects Impact of symptom change Fixed effects: Level-1 Level-2 Level-3 Random effects: Level-1 Level-2 Level-3

Time Time × CAPS Session Session × CAPS Intercept CAPS 2  ␲00  ␲11  ␤00  ␤11

Deviance Parameters AIC BIC Impact of responder status Fixed effects: Level-1 Level-2 Level-3 Random effects: Level-1 Level-2 Level-3 Deviance Parameters AIC BIC

Time Time × Responder Session Session × Responder Intercept Responder 2  ␲00  ␲11  ␤00  ␤11

Model 1 ˇ 1.77 −0.01 2.77 −0.09 65.84 −0.18 Estimate 99.13 196.77 9.13 172.52 0.04 2863.51 13 2889.51 2939.25 Model 3 ˇ 1.75 −0.36 0.47 −4.37 62.82 −7.09 Estimate 99.00 188.79 9.14 161.34 0.66 2863.05 13 2889.05 2938.79

AIC, Akaike information criterion; BIC, Bayesian Information Criterion. * p < .05. ** p < .01. *** p < .001.

Completers Only SE 1.12 0.02 1.86 0.04 8.15 0.17 SD 9.96 14.03 3.02 13.13 0.21

t 1.60 −0.24 1.49 −2.43* 8.08*** −1.05 2

SE 0.81 1.06 1.16 1.63 6.28 8.72 SD 9.95 13.74 3.02 12.70 0.81

t 2.16* −0.34 0.40 −2.68* 10.00*** −0.81 2

243.93*** 160.45*** 23.70** 25.60**

241.54*** 161.55*** 25.35** 25.34**

Model 2 ˇ 5.47 −0.07 4.87 −0.14 54.85 0.02 Estimate 99.68 134.59 7.93 152.41 0.10 2518.52 13 2544.52 2593.18 Model 4 ˇ 3.28 −1.89 1.19 −5.42 55.69 0.59 Estimate 99.59 132.79 9.05 134.20 0.85 2522.85 13 2548.85 2597.51

SE 1.51 0.03 1.74 0.04 14.40 0.26 SD 9.98 11.60 2.82 12.35 0.32

t 3.63*** −2.47* 2.80* −3.83** 3.81** 0.09 2

SE 0.97 1.18 1.08 1.50 8.03 9.69 SD 9.98 11.62 3.01 11.58 0.92

t 3.39*** −1.60 1.10 −3.61** 6.94*** 0.60 2

227.85*** 135.25*** 28.77*** 12.82

226.86*** 148.56*** 28.70*** 13.96

R.K. Sripada, S.A.M. Rauch / Journal of Anxiety Disorders 30 (2015) 81–87 100

Low Responder High Responder

SUDS

75

50

25

0

0

1

2

3

4

5

6

7

8

9

Session Number Fig. 1. Between-session habituation by responder status. High responders exhibited greater between-session habituation than low responders (ˇ = 5.42, SE = 1.50). For illustration purposes, SUDS here is averaged over the course of the imaginal exposure session. SUDS = Subjective Units of Distress Scale.

responders, high responders exhibited an additional 4.37 (SE = 1.63) point SUDS decrease, on average, between sessions. Responder status accounted for 95% of the variance in slope of SUDS between sessions. 3.2.4. Model 4: responder status and outcomes; completers only Responder status predicted slope of SUDS across sessions (t = −3.61, df = 8, p = .007), but did not predict overall SUDS level or slope of SUDS within a given session. Compared to low responders, high responders exhibited an additional 5.42 (SE = 1.50) point SUDS decrease, on average, between sessions (see Fig. 1). 4. Discussion In this study, we analyzed subjective distress ratings collected within imaginal exposure sessions to demonstrate significant within- and between-session change over the course of PE. We took advantage of the frequent data collection inherent in PE in order to examine habituation in a greater level of detail than has been conducted in any previous study. Our analysis revealed that the direction of the between-session (weeks-in-treatment) effect on SUDS was negative, indicating habituation, but that the direction of the within-session effect was positive, indicating increased engagement over the course of an individual session. This runs counter to common conceptualizations of within-session habituation of anxiety (Foa & Kozak, 1986). However, it is consistent with the notion that successful fear extinction requires fear activation during each exposure session (Foa, Riggs, Massie, & Yarczower, 1995). Of note, the observed within-session rise was small; on average, patients demonstrated an increase of 1.75 SUDS units every 5 min. This finding demonstrates that the typical pattern of anxiety generated by PE may be slightly increased and then sustained SUDS over the course of a session, as opposed to decreased SUDS. We also demonstrated that high treatment responders showed greater between-session habituation than low responders, but not differential within-session habituation (in 3 of 4 models) nor any difference in overall SUDS level. This result is consistent with evidence suggesting that between-session habituation is an important indicator of treatment response in PE (Bluett et al., 2014; Gallagher & Resick, 2012; Jaycox et al., 1998; Rauch et al., 2004; van Minnen & Foa, 2006; van Minnen & Hagenaars, 2002). The advantage of the current study is that data were analyzed from every available session, as opposed to the first-two-session approach or first-andlast session approach used by previous studies. In addition, we

85

examined every SUDS measurement within every session, for a total of over 300 SUDS measurements. Thus, our data provide convincing evidence that SUDS changes between sessions are meaningful and signify a favorable response to treatment. Our findings do not provide strong support for the predictive power of within-session habituation. However, several previous studies suggest that within-session habituation is beneficial to treatment response. Norton et al. (2011) demonstrated that amongst 106 individuals with anxiety disorders, within-session habituation was associated with reduced likelihood of dropout and lower anxiety symptomatology at treatment endpoint. van Minnen and Hagenaars (2002) reported that treatment responders showed greater within-session habituation than non-responders while listening to their imaginal exposure tapes at home, though they did not exhibit differential within-session habituation during the therapy session itself. Of note, methodological differences may contribute to this pattern of results. Specifically, habituation upon listening to a tape of imaginal exposure may be different than the process that occurs during a new imaginal exposure session. During imaginal exposure, details shift and change and generation of the story is ongoing, versus listening to a tape with a static representation of the memory account. This static representation makes habituation more likely to occur during homework practice than it would during a “new” memory account as it is being produced in a therapy session. Our findings echo those of studies demonstrating that within-session habituation does not determine treatment response (Jaycox et al., 1998; van Minnen & Foa, 2006). For instance, van Minnen and Foa (2006) demonstrated that shorter imaginal exposure sessions were equally as effective as longer sessions, despite the fact that they produced less within-session habituation. Furthermore, the authors demonstrated that level of within-session habituation during the first imaginal exposure was not associated with treatment outcome. Similar findings have been reported in studies of OCD (Kozak et al., 1988). Thus, the collected findings on within-session habituation in PE do not present a compelling case for its necessity for good treatment outcomes. Instead, the current study is more supportive of a pattern of sustained SUDS within each session of imaginal exposure. Our findings are partially consistent with Emotion Processing Theory, which posits that incorporating incompatible information into one’s fear structure results in modifications of the fear structure that reduce PTSD symptoms. Reduced anxiety (over time) in the face of fear cues, due to a lack of expected aversive outcomes, exemplifies such incompatible information. Within- and between-session anxiety reductions represent two such pieces of information that may be incorporated, resulting in PTSD symptom reduction (Foa & Kozak, 1986). While our results are consistent with the assertion of the need for between-session SUDS reduction, they do not support the assertion that within-session reductions are also needed for symptom reduction. Of note, Foa proposed that within-session habituation may be most important for patients who believe that anxiety will not diminish unless one escapes the fear-associated context (van Minnen & Foa, 2006). Thus, with patients for whom this belief is less central to the maintenance of PTSD, within-session habituation may not play a central role in recovery. Further, in PE, recording of SUDS is restricted to the imaginal exposure portion of the session. This practice suits the goals of PE, as the therapist is focused on maintaining optimal engagement in session and SUDS can be used for this purpose. However, when investigating the full scope of SUDS change over the course of the session, little is known about what happens when the imaginal exposure portion of the session is over and the emotional processing portion begins. Since patients are not as fully engaged with the trauma memory during processing as during imaginal exposure, it is likely that SUDS declines during this time for most patients. Furthermore, since therapists are unlikely to allow patients to leave

86

R.K. Sripada, S.A.M. Rauch / Journal of Anxiety Disorders 30 (2015) 81–87

their offices with very high levels of distress and activation, it is even more likely that reduced SUDS occurs after imaginal exposure but before the session ends. Future studies could include SUDS recordings during this period of the session in order to address these concerns.

5. Limitations There are several limitations to this study that merit discussion. First, our sample was small, thus our results require replication. Second, we did not test fear activation (either subjective or psychophysiological), which has been demonstrated by some studies to be a critical aspect of successful exposure therapy (Foa et al., 1995; Norton et al., 2011; Pitman, Orr Altman, Longpre, Poiré, et al., 1996). However, the lack of main effect of responder status on SUDS level in our study demonstrates that overall, SUDS levels did not predict treatment response. Since reported SUDS varied from 0 to 100, (M = 55.75, SD = 22.3) there was no evidence of restricted range. Higher levels of SUDS were not related to better response, thus our results do not support the notion that fear level alone is predictive of outcome. Rather, characteristic patterns of decreasing fear across sessions are associated with better outcome. In future studies, a larger sample would enable implementation of a more complex model (e.g. testing a quadratic effect of time-withinsession to investigate whether SUDS reliably increases and then decreases during an imaginal exposure session). This type of model could better test the fear activation hypothesis. We did not test whether symptom change is mediated by habituation, thus conclusions cannot be drawn about causal relationships between these constructs. We did find evidence that between-session habituation (but not within-session habituation) is associated with symptom change, which can provide important information to treatment providers about whether or not patients are exhibiting the patterns of habituation that are characteristic of high responders. However, claims that habituation is the underlying mechanism of symptom change would be speculative. Additionally, it is possible that change in symptoms occurred prior to, or simultaneous with, between-session habituation, and that symptom change is responsible for this habituation. Time-lagged studies that assess symptom severity at each visit and that test the predictive power of both variables (symptom severity and habituation) are needed to investigate this possibility. Another limitation of our study was that we did not assess physiological habituation of distress, which may have differed from perceived distress, particularly from the linguistic representation of perceived distress that is captured by SUDS level. Future work incorporating measurements of skin conductance, heart rate variability, or EEG would help to clarify whether physiological habituation exhibits the same patterns as self-reported habituation. Finally, we were unable to distinguish between the processes of habituation and extinction that occurred over the course of treatment. Habituation refers to a decrease in responding to a stimulus after repeated presentations of the stimulus (Groves & Thompson, 1970), whereas extinction refers to a weakening of the learned association between a conditioned stimulus and an unconditioned stimulus, once the conditioned stimulus is no longer reinforced (Mowrer, 1960). In our discussion, in order to be consistent with the exposure therapy literature, we have used the term habituation to indicate anxiety reduction during exposure. However, we recognize that both extinction and habituation processes are active during exposure therapy, and that both are necessary to symptom improvement. Laboratory studies that explicitly test processes of extinction (by removing the aversive stimulus after conditioning has taken place) versus habituation (by retaining the aversive stimulus) can help disentangle the role of these processes in exposure therapy.

6. Conclusions In conclusion, the current study provides robust support for the importance of between-session habituation in PTSD treatment outcome. In addition, within-session habituation does not appear to be related to response. Instead, a pattern of slight increase in SUDS within a session is apparent in the overall sample, though not related to treatment response. As such, practitioners’ efforts to optimize treatment should focus on those patients who are not showing reductions in SUDS across sessions. Indeed, discussion of within-session patterns of SUDS might include the fact that no change or even an increase in SUDS is not related to overall treatment response. Such patterns may represent engagement with the memory over the course of the session and suggest possible good outcomes for specific patients. Tracking SUDS changes across sessions can help practitioners monitor patient progress and can provide data to better guide decisions about implementing treatment modifications or enhancements.

Role of the funding source This material is based upon work supported by the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Clinical Sciences Research and Development (VA-CDA-2-010-06F) including a Career Development Award (Rauch: PI). Writing of this manuscript was supported by the Department of Veterans Affairs Office of Academic Affiliations, Advanced Fellowship Program in Mental Illness Research and Treatment, the Mental Health Service of the Veterans Affairs Ann Arbor Health Care System, and the Veterans Affairs Serious Mental Illness Treatment Resource and Evaluation Center (SMITREC). The authors would like to thank Dr. Brady West in the University of Michigan Program in Survey Methodology and the Center for Statistical Consultation and Research for support in statistical analysis for this paper.

References Baker, A., Mystkowski, J., Culver, N., Yi, R., Mortazavi, A., & Craske, M. G. (2010). Does habituation matter? Emotional processing theory and exposure therapy for acrophobia. Behaviour Research and Therapy, 48(11), 1139–1143. http://dx.doi.org/10.1016/j.brat.2010.07.009. S0005-7967(10)00161-0 [pii] Blake, D. D., Weathers, F. W., Nagy, L. M., & Kaloupek, D. G. (1995). The development of a Clinician-Administered PTSD Scale. Journal of Traumatic Stress, 8(1), 75–90. Bluett, E. J., Zoellner, L. A., & Feeny, N. C. (2014). Does change in distress matter? Mechanisms of change in prolonged exposure for PTSD. Journal of Behavior Therapy and Experimental Psychiatry, 45(1), 97–104. http://dx.doi.org/10.1016/j.jbtep.2013.09.003. S0005-7916(13)00066-9 [pii] Craske, M. G., Kircanski, K., Zelikowsky, M., Mystkowski, J., Chowdhury, N., & Baker, A. (2008). Optimizing inhibitory learning during exposure therapy. Behaviour Research and Therapy, 46(1), 5–27. http://dx.doi.org/10.1016/j.brat.2007.10.003. S0005-7967(07)00205-7 [pii] Culver, N. C., Stoyanova, M., & Craske, M. G. (2012). Emotional variability and sustained arousal during exposure. Journal of Behavior Therapy and Experimental Psychiatry, 43(2), 787–793. http://dx.doi.org/10.1016/j.jbtep.2011.10.009. S0005-7916(11)00116-9 [pii] Foa, E. B., & Kozak, M. J. (1986). Emotional processing of fear: exposure to corrective information. Psychological Bulletin, 99(1), 20–35. Foa, E. B., & Rauch, S. A. (2004). Cognitive changes during prolonged exposure versus prolonged exposure plus cognitive restructuring in female assault survivors with posttraumatic stress disorder. Journal of Consulting and Clinical Psychology, 72(5), 879–884. http://dx.doi.org/10.1037/0022-006X.72.5.879, 2004-19094-015 [pii] Foa, E. B., Riggs, D. S., Massie, E. D., & Yarczower, M. (1995). The impact of fear activation and anger on the efficacy of exposure treatment for posttraumatic stress disorder. Behavior Therapy, 26(3), 487–499. http://dx.doi.org/10.1016/s0005-7894(05)80096-6 Gallagher, M. W., & Resick, P. A. (2012). Mechanisms of change in cognitive processing therapy and prolonged exposure therapy for PTSD: preliminary evidence for the differential effects of hopelessness and habituation. Cognitive Therapy and Research, 36(6) http://dx.doi.org/10.1007/s10608-011-9423-6 Groves, P. M., & Thompson, R. F. (1970). Habituation: a dual-process theory. Psychological Review, 77(5), 419–450.

R.K. Sripada, S.A.M. Rauch / Journal of Anxiety Disorders 30 (2015) 81–87 Jaycox, L. H., Foa, E. B., & Morral, A. R. (1998). Influence of emotional engagement and habituation on exposure therapy for PTSD. Journal of Consulting and Clinical Psychology, 66(1), 185–192. Kozak, M. J., Foa, E. B., & Steketee, G. (1988). Process and outcome of exposure treatment with obsessive-compulsives: psychophysiological indicators of emotional processing. Behavior Therapy, 19(2), 157–169. http://dx.doi.org/10.1016/s0005-7894(88)80039-x Lang, A. J., & Craske, M. G. (2000). Manipulations of exposure-based therapy to reduce return of fear: a replication. Behaviour Research and Therapy, 38(1), 1–12. S00057967(99)00031-5 [pii]. Meuret, A. E., Seidel, A., Rosenfield, B., Hofmann, S. G., & Rosenfield, D. (2012). Does fear reactivity during exposure predict panic symptom reduction? Journal of Consulting and Clinical Psychology, 80(5), 773–785. http://dx.doi.org/10.1037/a0028032, 2012-09125-001 [pii] Mowrer, O. H. (1960). Learning theory and behavior. New York: John Wiley and Sons. Myers, K. M., & Davis, M. (2002). Behavioral and neural analysis of extinction. Neuron, 36(4), 567–584. S0896627302010644 [pii]. Norton, P. J., Hayes-Skelton, S. A., & Klenck, S. C. (2011). What happens in session does not stay in session: changes within exposures predict subsequent improvement and dropout. Journal of Anxiety Disorders, 25(5), 654–660. http://dx.doi.org/10.1016/j.janxdis.2011.02.006. S0887-6185(11)00029-6 [pii] Pitman, R. K., Orr, S. P., Altman, B., Longpre, R. E., Poiré, R. E., & Macklin, M. L. (1996). Emotional processing during eye movement desensitization and reprocessing therapy of vietnam veterans with chronic posttraumatic stress disorder. Comprehensive Psychiatry, 37(6), 419–429. http://dx.doi.org/10.1016/s0010-440x(96)90025-5 Pitman, R. K., Orr, S. P., Altman, B., Longpre, R. E., Poire, R. E., Macklin, M. L., et al. (1996). Emotional processing and outcome of imaginal flooding therapy in Vietnam veterans with chronic posttraumatic stress disorder. Comprehensive Psychiatry, 37(6), 409–418. Powers, M. B., Halpern, J. M., Ferenschak, M. P., Gillihan, S. J., & Foa, E. B. (2010). A meta-analytic review of prolonged exposure for posttraumatic stress disorder. Clinical Psychology Review, 30(6), 635–641. http://dx.doi.org/10.1016/j.cpr.2010.04.007. S0272-7358(10)00070-X [pii] Rauch, S. A. M., Defever, E., Favorite, T., Duroe, A., Garrity, C., Martis, B., et al. (2009). Prolonged exposure for PTSD in a Veterans Health Administration PTSD clinic. Journal of Traumatic Stress, 22(1), 60–64. http://dx.doi.org/10.1002/jts.20380 Rauch, S. A. M., Foa, E. B., Furr, J. M., & Filip, J. C. (2004). Imagery vividness and perceived anxious arousal in prolonged exposure treatment for PTSD. Journal of Traumatic Stress, 17(6), 461–465. http://dx.doi.org/10.1007/s10960-004-5794-8

87

Rauch, S. A. M., King, A. P., Abelson, J. L., Tuerk, P. W., Smith, E., Rothbaum, B. O.,. Liberzon, I. (in press). Biological and symptom changes in posttraumatic stress disorder treatment: a randomized clinical trial. Depression and Anxiety. 10.1002/da.22331. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: applications and data analysis methods (2nd ed.). Newbury Park, CA: Sage. Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., & Congdon, R. T. J. (2004). HLM 6: Hierarchical Linear and Nonlinear Modeling. Lincolnwood, IL: Scientific Software International. Riley, W. T., McCormick, M. G. F., Simon, E. M., Stack, K., Pushkin, Y., Overstreet, M. M., et al. (1995). Effects of alprazolam dose on the induction and habituation processes during behavioral panic induction treatment. Journal of Anxiety Disorders, 9(3), 217–227. http://dx.doi.org/10.1016/0887-6185(95)00003-7 Rowe, M. K., & Craske, M. G. (1998). Effects of varied-stimulus exposure training on fear reduction and return of fear. Behaviour Research and Therapy, 36(7–8), 719–734. S0005-7967(97)10017-1 [pii]. Schnurr, P. P., Friedman, M. J., Engel, C. C., Foa, E. B., Shea, M. T., Chow, B. K., et al. (2007). Cognitive behavioral therapy for posttraumatic stress disorder in women: a randomized controlled trial. JAMA, 297(8), 820–830. http://dx.doi.org/10.1001/jama.297.8.820, 297/8/820 [pii] Snijders, T., & Bosker, R. (1994). Modeling variance in two-level models. Sociological Methods and Research, 22, 342–363. Tsao, J. C. I., & Craske, M. G. (2000). Timing of treatment and return of fear: effects of massed, uniform-, and expanding-spaced exposure schedules. Behavior Therapy, 31(3), 479–497. http://dx.doi.org/10.1016/s0005-7894(00)80026-x Tuerk, P. W., Yoder, M., Grubaugh, A., Myrick, H., Hamner, M., & Acierno, R. (2011). Prolonged exposure therapy for combat-related posttraumatic stress disorder: an examination of treatment effectiveness for veterans of the wars in Afghanistan and Iraq. Journal of Anxiety Disorders, 25(3), 397–403. http://dx.doi.org/10.1016/j.janxdis.2010.11.002. S0887-6185(10)00220-3 [pii] van Minnen, A., & Foa, E. B. (2006). The effect of imaginal exposure length on outcome of treatment for PTSD. Journal of Traumatic Stress, 19(4), 427–438. http://dx.doi.org/10.1002/jts.20146 van Minnen, A., & Hagenaars, M. (2002). Fear activation and habituation patterns as early process predictors of response to prolonged exposure treatment in PTSD. Journal of Traumatic Stress, 15(5), 359–367. http://dx.doi.org/10.1023/A:1020177023209 West, B. T., Welch, K. B., & Galecki, A. T. (2006). Linear mixed models: a practical guide using statistical software. Boca Raton, FL: Chapman and Hall/CRC. Wolpe, & Lazarus, A. A. (1966). Behavior therapy techniques. New York, NY: Pergamon.