The Relationship Between Posttraumatic and Depressive Symptoms During
Prolonged Exposure With and Without Cognitive Restructuring for the
Treatment of Posttraumatic Stress Disorder
Idan M. Aderka
Boston University
Seth J. Gillihan, Carmen P. McLean, and
Edna B. Foa
University of Pennsylvania Objective:In the present study, we examined the relationship between posttraumatic and depressive symptoms during prolonged exposure (PE) treatment with and without cognitive restructuring (CR) for the treatment of posttraumatic stress disorder (PTSD).Method:Female assault survivors (N5153) with PTSD were randomized to either PE alone or PE with added CR (PE/CR). During treatment, bi-weekly self-report measures of posttraumatic and depressive symptoms were administered.Results:Multilevel mediational analyses indicated that during PE, changes in posttraumatic symptoms accounted for 80.3% of changes in depressive symptoms, whereas changes in depressive symptoms accounted for 45.0% of changes in posttraumatic symptoms. During PE/CR, changes in posttraumatic symptoms accounted for 59.6% of changes in depressive symptoms, and changes in depressive symptoms accounted for 50.7% of changes in posttraumatic symptoms.Conclusions:This pattern of results suggests that PE primarily affects posttraumatic symptoms, which in turn affect depressive symptoms. In contrast, PE/CR results in a more reciprocal relationship between posttraumatic and depressive symptoms.
Keywords:posttraumatic stress disorder, depression, prolonged exposure
Posttraumatic stress disorder (PTSD) is a common and chronic psychiatric condition with prevalence rates of 6.8% in the general population (Kessler et al., 2005) and high comorbidity rates (Kessler, Sonnega, Bromet, Hughes, & Nelson, 1995). Individuals with PTSD typically report elevated depressive symptoms (e.g., Erickson, Wolfe, King, King, & Sharkansky, 2001), and comorbid major depressive disorder (MDD) is common among individuals with PTSD (e.g., Brown, Campbell, Lehman, Grisham, & Mancill, 2001).
Evidence regarding the temporal relationship between posttrau-matic and depressive symptoms has been mixed. Some studies sug-gest that posttraumatic symptoms lead to depression. For example, in the National Comorbidity Survey, 78.4% of individuals with both PTSD and MDD reported that the onset of PTSD preceded that of MDD (Kessler et al., 1995). In contrast, other studies have suggested that depressive symptoms lead to posttraumatic symptoms more con-sistently than vice versa (King, King, McArdle, Shalev, & Doron-LaMarca, 2009; Schindel-Allon, Aderka, Shahar, Stein, & Gilboa-Schechtman, 2010). Finally, some studies have found that PTSD and MDD develop simultaneously (e.g., Shalev et al., 1998) and that reciprocal relations exist between posttraumatic and depressive symp-toms (Erickson et al., 2001).
Treatments for PTSD have been found to reduce both posttrau-matic and depressive symptoms (see Harvey, Bryant, & Tarrier, 2003, for a review). Specifically, prolonged exposure (PE), a widely used treatment for PTSD, has been shown to significantly reduce posttraumatic and depressive symptoms across diverse pop-ulations and settings (Powers, Halpern, Ferenschak, Gillihan, & Foa, 2010). Research has shown that PE can alleviate posttrau-matic and depressive symptoms among adults treated in academic and community settings (Feske, 2008; Foa et al., 2005), among female assault (Foa & Rauch, 2004) and rape (Resick, Nishith, Weaver, Astin, & Feuer, 2002) survivors, among veterans (Rauch et al., 2009; Tuerk et al., 2011), and among individuals previously treated with medication (Rothbaum et al., 2006).
Despite the large body of research documenting the effects of PE on both posttraumatic and depressive symptoms, little is known about the relationship between reductions in the two during PE. There are several possible ways in which posttraumatic and depressive toms may interact during treatment. Changes in posttraumatic symp-toms may lead to subsequent changes in depressive sympsymp-toms but not vice versa. Conversely, changes in depressive symptoms may lead to subsequent changes in posttraumatic symptoms but not vice versa. Alternatively, changes in posttraumatic symptoms may lead to sub-sequent changes in depressive symptoms, and changes in depressive symptoms may lead to subsequent changes in posttraumatic symp-toms. Finally, changes in posttraumatic and depressive symptoms may be unrelated. Understanding the relationship between posttrau-matic and depressive symptoms during PE is important as it can help us better understand how PE works and can aid in the development of more effective treatments.
To our knowledge, only one published study has examined the relationship between posttraumatic and depressive symptoms in
Idan M. Aderka, Department of Psychology, Boston University; Seth J. Gillihan, Carmen P. McLean, and Edna B. Foa, Department of Psychiatry, University of Pennsylvania.
Seth J. Gillihan is now at the Department of Psychology, Haverford College.
Correspondence concerning this article should be addressed to Idan M. Aderka, who is now at the Department of Psychology, University of Haifa, Mount Carmel, Haifa 31905, Israel. E-mail: [email protected]
the treatment of PTSD (Aderka, Foa, Applebaum, Shafran, & Gilboa-Schechtman, 2011). This study examined prolonged expo-sure treatment (PE) among children and adolescents, treating either previous depression and PTSD symptoms as a mediator of later PTSD or depression reduction, respectively. Aderka et al. (2011) found that changes in posttraumatic symptoms accounted for 64.1% in changes in depressive symptoms, whereas changes in depressive symptoms accounted for only 11.0% of the changes in posttraumatic symptoms. Importantly, Aderka et al. examined lagged mediation rather than contemporaneous changes, therefore establishing temporal precedence of the mediator. These findings suggest that the effects of PE follow a sequential pattern in which treatment first reduces PTSD symptoms, which leads to a reduc-tion in depressive symptoms.
The aim of the present study was to evaluate whether the pattern of influence of PE on PTSD and depressive symptoms in adults is similar to that found in youths. Thus, we examined the relationship between posttraumatic and depressive symptoms during PE in an adult sample. Based on the findings in a pediatric sample (Aderka et al., 2011), and the fact that PE was specifically developed to target posttraumatic symptoms, we hypothesized that in PE changes in posttraumatic symptoms would account for subsequent changes in depressive symptoms to a greater extent than vice versa.
The relationship between posttraumatic and depressive symp-toms also may vary across treatments. Specifically, it is possible that the addition of cognitive restructuring (CR) to PE may affect the relationship between posttraumatic and depressive symptoms. To examine this hypothesis, we explored the PTSD– depression relationship during PE with and without added CR. Due to the paucity of research on patterns of symptom change, our examination regarding potential differences between the treat-ments was exploratory, and we did not have explicit hypotheses about the PTSD– depression relationship in PE with added CR (PE/CR).
For both PE and PE/CR, we examined the four possible rela-tionships between posttraumatic and depressive symptoms over time: (a) changes in posttraumatic symptoms account for changes in depressive symptoms but not vice versa, (b) changes in depres-sive symptoms account for changes in posttraumatic symptoms but not vice versa, (c) changes in both posttraumatic and depressive symptoms account for changes in each other, and (d) changes in posttraumatic symptoms are unrelated to changes in depressive symptoms.
Method Participants
Participants were recruited through a variety of sources includ-ing flyers, newspaper advertisements, and referrals from police departments and victim advocacy workers. Eligible participants were adult women with a primary diagnosis of PTSD related to childhood sexual abuse or to a sexual or nonsexual assault that occurred at least 3 months prior to the evaluation. Exclusion criteria were as follows: being in an abusive relationship; current diagnosis of organic mental disorder, schizophrenia, or psychotic disorder; unmedicated, symptomatic bipolar disorder; current sub-stance dependence; illiteracy in English; current suicidal intent or
plan; and recent history of serious self-injurious behavior (e.g., cutting). For additional details, see Foa et al. (2005).
A total of 210 eligible women consented to participate; 20 of these women withdrew before being assigned a treatment con-dition, and 11 were removed from the study after randomiza-tion. Thus, our intention-to-treat (ITT) sample comprised 179 women. Twenty-six were assigned to waitlist (WL), 74 to PE/CR, and 79 to PE. Only individuals from the PE and PE/CR conditions (n5153) were included in the present study. A total of 90 were treated at the Center for Treatment and Study of Anxiety in Philadelphia, and 63 were treated at the Women Organized Against Rape clinic. For flow of patients through the study and reasons for removal, see Foa et al. (2005).
Participants in the ITT sample had a mean age of 31 years and were predominately Caucasian (49.2%) or African American (43.6%), single (61.6%), with at least some college education (70.0%). Sexual assault during adulthood was the most prevalent index trauma; the average time since the index trauma was 9 years. Psychiatric comorbidity was common, with 67% of the sample having at least one comorbid Axis I disorder. The most common comorbid conditions were major depressive disorder (41.2%), so-cial phobia (20.4%), and specific phobia (20.4%). For complete demographic information, see Foa et al. (2005).
Procedure
Evaluations. The Structured Clinical Interview forDSM–IV Axis I Disorders With Psychotic Screen (SCID-I; First, Spitzer, Gibbon, & Williams, 1995) was administered at pretreatment to establish diagnoses and assess exclusion criteria and comorbid conditions. Participants completed the PTSD Symptom Scale— Self-Report (PSS–SR; Foa, Riggs, Dancu, & Rothbaum, 1993) and the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) prior to each even-numbered therapy session (i.e., bi-weekly) to monitor treatment progress and to determine treatment termination; these repeated assessments across the course of treatment allow for tests of the current study hypotheses regarding the relation between posttraumatic and de-pressive symptoms over time.
Treatments. Both treatments were delivered in individual weekly sessions that lasted 90 –120 min. Participants who reported at least a 70% reduction in PSS–SR scores (n 5 27) ended treatment after Session 9; the remaining participants were offered three extension sessions (and 13 participants declined).1No dif-ferences in pre-treatment demographic or clinical measures were found between those who ended treatment at Session 9 and those who received additional sessions (Foa et al., 2005). Each session began with homework review and ended with homework assign-ment. Brief descriptions of the treatments follow; for more detailed descriptions, see Foa, Hembree, and Rothbaum (2007) and Foa et al. (2005). The mean number of therapy sessions attended was 8.14 (SD53.77).
PE. Session 1 included presentation of the treatment rationale and program overview, information gathering, and breathing re-training. In Session 2, the therapist provided education about common reactions to trauma and introduced in vivo exposure to a
1Due to this feature of the design, our sample size at Session 10 was 62 and at Session 12 was 20.
collaboratively developed hierarchy of previously avoided situa-tions. Throughout the treatment, participants were assigned home-work to approach items on the hierarchy in a systematic, gradual fashion. Imaginal exposure began in Session 3. In this procedure, participants were asked to revisit in imagination their most dis-tressing traumatic memory and recount it aloud in present tense, repeating the recounting as necessary to allow total revisiting of 45 min. Following imaginal exposure, therapists briefly discussed the patients’ experiences (processing). Participants were instructed to listen daily to an audio recording that was made of the recounting. Sessions 4 –9 (or 12) were conducted in a similar fashion. In the final session, participants summarized what they had learned in treatment and discussed their progress. Therapists and participants also discussed future plans and what to do if the participant’s symptoms increased.
PE/CR. The procedures used in the PE/CR treatment were identical to PE alone with two exceptions. First, in Session 3 therapists presented the idea that posttraumatic symptoms are maintained in part by trauma-related thoughts and beliefs. Partic-ipants were taught to identify and challenge erroneous and unhelp-ful beliefs and were instructed to record and challenge these beliefs for homework using a daily diary. Imaginal exposure was intro-duced in Session 4. Second, instead of the discussion about the anxiety experienced during the imaginal and in vivo exposure, patients in the PE/CR group received about 25 min of formal CR using the Socratic method developed by Beck (Beck & Emory, 1985). Participants in PE/CR were given the same amount of exposure homework as those in PE, and they also practiced CR using their diaries.
Treatment adherence. Weekly supervision meetings were used to monitor adherence to the treatment protocol. In addition, using adherence manuals, 141 therapy sessions (11.5% of 1,227 sessions) were randomly selected, and videotapes were rated for fidelity to the treatment manual. Ten raters trained to conduct the adherence ratings reviewed session videotapes, rated each essential component as present or absent, and monitored for protocol vio-lations. Of these sessions, 29 (21%) were rated independently by two raters. Interrater reliability was .88; therapists on average completed 97% of the components prescribed in the protocol.
Measures
The SCID-I (First et al., 1995) is a semistructured interview used to assess major Axis I disorders as well as to screen for the presence of psychotic symptoms. In the current study, it was used to assess exclusion criteria and comorbid conditions as well. Spe-cifically, for PTSD diagnosis, the SCID-I has shown excellent inter-rater reliability (Lobbestael, Leurgans, & Arntz, 2011) and has been found to be superior to other diagnostic instruments in screening for trauma (Elhai, Franklin, & Gray, 2008).
The PSS–SR (Foa et al., 1993) is a self-report version of the PTSD Symptom Scale—Interview (PSS–I; Foa, Riggs, Dancu, & Rothbaum, 1993). It includes 17 items that measure the frequency and severity of PTSD symptoms in the past week, according to Diagnostic and Statistical Manual of Mental Disorders(4th ed., text rev.;DSM–IV–TR; American Psychiatric Association, 2000) criteria. In the present study, the PSS–SR was internally consistent (Cron-bach’sa 5.91) and stable over a period of 1 month (r5.74).
The BDI (Beck et al., 1961) is a 21-item self-report inventory measuring cognitive, affective, and physiological symptoms of depression in the past week. In the present study, the BDI was internally consistent (Cronbach’sa 5.93).
Analytic Strategy
Our data formed a multilevel structure. The lower level, or Level 1 data, consisted of the repeated measures that were col-lected during treatment sessions (i.e., posttraumatic and depressive symptoms). The Level 1 data were nested within Level 2 units (i.e., participants). This data structure is appropriate for hierarchi-cal linear modeling (HLM) techniques (Raudenbush, 2001). HLM allows the number of observations to vary between participants and handles missing data effectively.
To examine mediation of change, we used the mixed models module of SPSS Version 18. We followed the procedures reported by Kenny, Korchmaros, and Bolger (2003), which have been previously applied to mediational analysis of anxiety and depres-sion during treatment (Aderka et al., 2011; Moscovitch, Hofmann, Suvak, & In-Albon, 2005). As described by Kenny et al., four statistical criteria must be present in order to establish mediation: (1) The predictor variable must be significantly related to the outcome variable; (2) the predictor variable must be significantly related to the mediator; (3) when the outcome is regressed simul-taneously on the predictor and mediator, the mediator must be significantly related to the outcome; and (4) the relationship be-tween the predictor and the outcome must be significantly atten-uated when the mediator is included. In order to perform a more rigorous test for mediation, we “lagged” the mediator variable in all analyses. Thus, we examined whether changes in the mediator variable at timetaccounted for changes in the outcome variable at timet11. Due to multiple comparisons, we set significance levels at .01 in all analyses.
Results
Treatment Effects on Posttraumatic and Depressive Symptoms
To examine whether the two treatment conditions had differen-tial effects on posttraumatic symptoms, we conducted a repeated-measures analysis of variance (ANOVA). The independent vari-ables were Treatment (a within-subjects, two-level variable: pre vs. post) and Condition (a between-subjects, two-level variable: PE vs. PE/CR). The dependent variable was posttraumatic symp-toms (PSS–I scores). Results indicated a significant main effect for treatment, F(1, 106)5504.04,p,.001, partialh25.83, such that participants’ posttraumatic symptoms were significantly lower at post-treatment (M 5 10.15, SD 5 9.04) compared to pre-treatment (M531.72,SD57.50). The main effects for Condition and the Treatment 3 Condition interaction were both non-significant,F(1, 106)50.26,p5.61, partialh25.00,ns;F(1, 106)52.68,p5.11, partialh25.03, respectively. These results indicate that PE and PE/CR reduced posttraumatic symptoms to a similar extent.
We used a similar ANOVA model to examine reductions in depressive symptoms. Results indicated a significant main effect for treatment,F(1, 100)5257.25,p,.001, partialh25.72, such
that participants’ depressive symptoms were significantly lower at post-treatment (M57.99,SD59.39) compared to pre-treatment (M523.72,SD58.99). The main effects for Condition and the Treatment3Condition interaction were both non-significant,F(1, 100)50.14,p5.71, partialh25.00;F(1, 100)52.23,p5.14, partial h2 5 .02, respectively, indicating that PE and PE/CR reduced depressive symptoms to a similar extent. A comprehen-sive examination of the differences between conditions including an intent-to-treat analysis can be found in Foa et al. (2005).
Multilevel-Mediational Analyses
Rates of change. Time was modeled using measurement number (i.e., 1, 2, 3, 4, 5, 6). Prior to running analyses, we compared a model with a linear time variable, a non-linear, natural log time variable, and a quadratic time variable. As these models were non-nested, we compared them using the Akaike information criterion (AIC; Akaike, 1987) and the deviance statistic. Results indicated that a linear model was the best-fitting model. We also examined a model with both a linear time variable and a quadratic variable, and a model with both a linear time variable and a non-linear, natural log variable. No significant non-linear effects were found (all ps..05), suggesting that changes in both post-traumatic and depressive symptoms occur at a similar rate through-out treatment. Thus, we used a linear time variable in all analyses. Effect of time. In the PE condition, our time variable ex-plained 52.9% of the changes in posttraumatic symptoms and 35.7% of the changes in depressive symptoms during treatment (rPTSD2 5.529;rDEP2 5.357).2In the PE/CR condition, this variable explained 51.7% of the changes in posttraumatic symptoms and 38.1% of the changes in depressive symptoms during treatment (rPTSD2 5.517;rDEP2 5.381). We found that in the PE condition, the effect size of time on posttraumatic symptoms was 3.29, and the effect size of time on depressive symptoms was 1.64.3 In the PE/CR condition, the effect size of time on posttraumatic symp-toms was 2.00, and the effect size of time on depressive sympsymp-toms was 1.45. These effect sizes are highly similar to those reported by Foa et al. (2005).
PE condition. We employed a model in which time was the predictor, posttraumatic symptoms were the mediator, and depres-sive symptoms were the outcome. This model tested the extent to which posttraumatic symptoms mediated the effect of time on depressive symptoms. In these analyses, posttraumatic symptoms were lagged so that posttraumatic symptoms at time tpredicted depressive symptoms at timet11. Consistent with the guidelines of Kenny et al. (2003), we first regressed depressive symptoms on time (Path C in Figure 1 and Table 1). Results indicated that depressive symptoms significantly decreased during treatment (b 5–.32,SE5 .04,t5– 8.06,p,.001). We then regressed posttraumatic symptoms on time (Path A in Figure 1 and Table 1). Results indicated that posttraumatic symptoms also significantly decreased during treatment (b 5–.48,SE5.05,t5–9.17,p, .001). Next, we regressed depressive symptoms on both time and posttraumatic symptoms simultaneously. The effect of posttrau-matic symptoms (Path B in Figure 1 and Table 1) was significant indicating that reductions in posttraumatic symptoms significantly predicted subsequent reductions in depressive symptoms (b 5.05, SE5.01,t57.22,p,.001). The effect of time (Path C=in Figure 1 and Table 1) was non-significant (b 5–.06,SE5.04,t5–1.54,
p5.13), indicating that time did not have an additional affect on depressive symptoms above the effect of posttraumatic symptoms. Thus, posttraumatic symptoms fully mediated the relationship between time and depressive symptoms. As can be seen in Figure 1 and Table 1, the original pathway between time and depressive symptoms (Path C) was reduced from – 0.32 to – 0.06 when the mediator (posttraumatic symptoms) was entered (Path C=).
2We examined explained variance for the time variable following Kreft and de Leeuw’s (1998) recommendations for HLM analyses. Specifically, the formula used was as follows: [(null model error2model with time error)/null model error].
3We computed effect sizes based on Feingold’s (2009) recommendations for multilevel models. Specifically, we used the formula [d5B(time)3
number of time periods/raw preSD], which has the advantage of being in the same metric commonly used in clinical trials.
PE PE/CR Time (Session) Depressive Symptoms Posttraumatic Symptoms B 0.03*** A -0.39*** Time (Session) Depressive Symptoms Time (Session) Depressive Symptoms Posttraumatic Symptoms B 0.05*** A -0.48*** Time (Session) Depressive Symptoms C -0.32*** C’ -0.06 C -0.25*** C’ -0.10**
Figure 1. Lagged mediational models. Numbers are standardized regres-sion coefficients. Paths indicate Level 1 regresregres-sion analyses. Posttraumatic symptoms were lagged so that posttraumatic symptoms at timetpredicted depressive symptoms at timet11. PE5prolonged exposure; PE/CR5
prolonged exposure with added cognitive restructuring. ppp ,
We also examined mediation using the program Prodclin (MacK-innon, Fritz, Williams, & Lockwood, 2007). This program provides a powerful test for mediation that can adjust for covariance between Paths A and B. We found clear evidence for mediation as the confi-dence interval (CI) for the indirect effect was [–1.89, – 0.68] and did not include zero. Of the total effect of time on depressive symptoms, 80.3% was mediated by posttraumatic symptoms.4
We examined the reverse model in which depressive symptoms mediated the effects of time on posttraumatic symptoms. In these analyses, depressive symptoms were lagged so that depressive symptoms at timetpredicted posttraumatic symptoms at timet1 1. Table 2 and Figure 2 present the results of these analyses. In the reverse model, the effect of time on posttraumatic symptoms (Path C=) remained significant after regressing posttraumatic symptoms on time and depressive symptoms simultaneously, which indicates that depressive symptoms only partially mediate the effect of time on posttraumatic symptoms. The mediational effect of depressive symptoms was significant (CI [–2.00, – 0.72]) and accounted for 45.0% of the total effect of time on posttraumatic symptoms. In sum, changes in posttraumatic symptoms accounted for subsequent changes in depressive symptoms to a greater extent than vice versa. This pattern of results suggests that PE may work primarily by reducing posttraumatic symptoms, which in turn reduce depres-sive symptoms.
PE/CR condition. We repeated the above analyses for the PE/CR condition. Results for the mediational model are pre-sented in Table 1 and Figure 1, and results for the reverse model are presented in Table 2 and Figure 2. We found evidence for partial mediation in both the mediational model (95% CI [–2.35, – 0.84]) and the reverse model (95% CI [–2.59, – 0.93]). Changes in posttraumatic symptoms mediated 59.6% of the changes in depressive symptoms, and changes in depressive symptoms mediated 50.7% of the changes in posttraumatic symptoms.
Discussion
The present study examined the relationship between posttrau-matic and depressive symptoms during PE and PE/CR for PTSD. We found that during PE, changes in posttraumatic symptoms fully mediated the effect of treatment on depressive symptoms and
accounted for 80.3% of changes in depressive symptoms, whereas changes in depressive symptoms partially mediated the effect of treatment on posttraumatic symptoms, and accounted for 45.0% of changes in posttraumatic symptoms. This pattern of results sug-gests that the primary effect of PE is on posttraumatic symptoms which in turn affect depressive symptoms. In contrast, for PE/CR we found that posttraumatic symptoms partially mediated the effect of treatment on depressive symptoms, and vice versa; changes in posttraumatic symptoms accounted for 59.6% of changes in depressive symptoms, and changes in depressive symp-toms accounted for 50.7% of changes in posttraumatic sympsymp-toms. Thus, during PE/CR posttraumatic and depressive symptoms af-fected each other to a similar extent.
Our findings for PE are consistent with a previous study that examined PE among children and adolescents (Aderka et al., 2011). In that study, posttraumatic symptoms fully mediated the effects of PE on depressive symptoms, whereas depressive symp-toms only partially mediated the effects of PE on posttraumatic symptoms. Thus, in both pediatric and adult populations, PE appears to reduce posttraumatic symptoms which in turn reduce depressive symptoms. However, it seems that changes in posttrau-matic symptoms and depressive symptoms are more closely linked in adult populations than in pediatric ones. In the present study, 80.3% of changes in depressive symptoms were accounted for by posttraumatic symptoms, compared to 64.1% among youths, and 45.0% of changes in posttraumatic symptoms were accounted for by changes in depressive symptoms, compared to 11.0% among youths (Aderka et al., 2011). This discrepancy may be the result of developmental differences, or differences in types of trauma ex-perienced and time since the trauma. More research is necessary to
4We calculated percent mediation according to the guidelines of Kenny et al. (2003), which have been previously applied to mediational analyses of anxiety and depression (Aderka et al., 2011; Moscovitch et al., 2005). According to Kenny et al., the total effect in lower level mediation models is the sum of the direct effect (C=), the indirect effect (AB), and the covariance between A and B (the covariance between the ordinary-least-squares estimates for Paths A and B). Thus, the formula used to calculate percent mediation was as follows: 100 3{[C=1AB1 Cov(AB)]2
C=}/[C=1AB1Cov(AB)], or 1003(total effect2direct effect)/(total effect).
Table 1
Summary of Multilevel Regression Analyses for Mediational Models
Step Path Predictor variable Outcome variable B SE B b SEb t p
PE
1 C Time Depression 23.03 0.38 2.32 .04 28.06 ,.001
2 A Time Posttraumatic symptomsa
25.44 0.59 2.48 .05 29.17 ,.001 3 B Posttraumatic symptomsa Depression 0.43 0.06 .05 .01 7.22 ,.001
C= Time Depression 20.57 0.37 2.06 .04 21.54 .13
PE/CR
1 C Time Depression 22.85 0.47 2.25 .04 26.13 ,.001
2 A Time Posttraumatic symptomsa 24.46 0.63 2.39 .05 27.07 ,.001 3 B Posttraumatic symptomsa Depression 0.36 0.07 .03 .01 5.31 ,.001
C= Time Depression 21.09 0.38 2.10 .03 22.90 ,.01
Note. PE5prolonged exposure; PE/CR5prolonged exposure with added cognitive restructuring.
replicate these differential findings for youths and adults in PE and to understand their causes.
The pattern of symptom change observed during PE is consis-tent with several theoretical accounts of the relationship between anxiety and depression. For instance, the helplessness– hopelessness theory (Alloy, Kelly, Mineka, & Clements, 1990) posits that feelings of helplessness are primarily related to anxiety, whereas feelings of hopelessness are primarily related to depres-sion. According to this theory, experiencing helplessness repeat-edly or for long periods of time can lead to hopelessness (Alloy et al., 1990). Thus, as individuals begin to better cope with their symptoms during treatment, their helplessness is reduced, which can result in subsequent reductions in hopelessness. In support of this potential mechanism, hopelessness has been found to mediate outcome in both PE and cognitive treatment of PTSD (Gallagher & Resick, 2012). Our findings are also consistent with the anhedonia model of depression, which stresses the causal role of anhedonia in the development of depression (Loas, 1996). Thus, reductions in trauma-related fear and avoidance may lead to the increased ex-perience of pleasurable life events, thus reducing anhedonia and subsequent depression.
We observed a sequential relationship between posttraumatic and depressive symptoms in PE (posttraumatic symptoms¡ de-pressive symptoms), and a more reciprocal relationship in PE/CR (posttraumatic symptoms ↔ depressive symptoms). A possible explanation for this difference between treatments is that PE and PE/CR may affect different posttraumatic and depressive factors. The imaginal and in vivo exposure in PE may affect unique PTSD factors (e.g., fear, guilt), and this reduction may affect shared PTSD/depression factors (e.g., negative beliefs of the self and the future) and unique depression factors (e.g., depressed mood). For example, in vivo exposures are very effective at reducing the distress that a person experiences in feared situations. This reduc-tion in distress may change the way an individual perceives him-self (e.g., “I am capable of dealing with these situations”), which may in turn lead to improved mood as the person becomes more hopeful. In contrast, the more language-focused, cognitive tech-niques in PE/CR may more directly target shared PTSD/depression factors (e.g., negative cognitions), which subsequently lead to reductions in unique PTSD and unique depression factors. For example, challenging one’s assumptions that he or she is “weak”
for having PTSD may increase one’s willingness to complete in vivo exposure exercises, and can lead to a reduction in trauma-related fear. Concurrently, challenging these assumptions can also increase self-efficacy and reduce self-blame, which can result in improved mood. While testing for these possible pathways is beyond the scope of the present article, they represent potentially fertile ground for future study.
Importantly, both PE and PE/CR produced similar improve-ments in symptoms (Foa et al., 2005) and similar reductions in negative cognitions (Foa & Rauch, 2004). Thus, the two treatments evidence differing patterns of symptom change but result in similar degrees of change. This finding has the potential to inform clinical research, as it suggests that adding an additional active ingredient of treatment may affect the pattern of change but not the outcome. This is consistent with many studies that fail to find differences in outcome between cognitive therapy and prolonged exposure (e.g., Foa et al., 2005; Resick et al., 2002), as well as a recent study suggesting that the two types of treatment may work through different mechanisms of change (Gallagher & Resick, 2012). Fu-ture studies should examine differences in patterns of change in addition to outcome, as this may increase our understanding of how different treatments work.
Limitations and Future Directions
The present study has several limitations. First, our sample included only female assault victims. Thus, it cannot be known based on the present data whether the same patterns of inter-relations exist among males or among individuals experiencing other traumas. Future studies can examine the relationship between posttraumatic and depressive symptoms in these pop-ulations. Second, the present study did not directly assess potential mediators between posttraumatic and depressive symptoms, such as hopelessness or anhedonia, that may play a role in the relationship between them; future studies can assess these mediators during treatment. Third, we examined media-tion throughout the course of treatment. However, it is possible that different mediational relationships exist in different phases of treatment (e.g., early vs. late). Future studies can examine whether the relationships found in the present study change in different phases of treatment.
Table 2
Summary of Multilevel Regression Analyses for Reverse Mediational Models
Step Path Predictor variable Outcome variable B SE B b SEb t p
PE
1 C Time Posttraumatic symptoms 24.74 0.43 2.41 .04 211.02 ,.001
2 A Time Depressiona 22.95 0.47 2.34 .05 26.24 ,.001
3 B Depressiona Posttraumatic symptoms 0.56 0.08 .05 .01 6.88
,.001 C= Time Posttraumatic symptoms 22.02 0.42 2.18 .04 24.88 ,.001
PE/CR
1 C Time Posttraumatic symptoms 24.04 0.50 2.34 .04 28.09 ,.001
2 A Time Depressiona 22.98 0.61 2.26 .05 24.87 ,.001
3 B Depressiona Posttraumatic symptoms 0.52 0.08 .04 .01 6.88 ,.001 C= Time Posttraumatic symptoms 21.56 0.41 2.13 .03 23.85 ,.001
Note. PE5prolonged exposure; PE/CR5prolonged exposure with added cognitive restructuring.
Despite these limitations, the current study represents the first longitudinal examination of the inter-relationship between posttraumatic and depressive symptoms during treatments for PTSD among adults. These findings highlight the importance of examining patterns of symptom change in psychotherapy re-search and suggest that despite similar outcomes, the techniques used in PE and PE/CR may result in different patterns of change in posttraumatic and depressive symptoms.
References
Aderka, I. M., Foa, E. B., Applebaum, E., Shafran, N., & Gilboa-Schechtman, E. (2011). Direction of influence between posttraumatic and depressive symptoms during prolonged exposure therapy among children and adoles-cents. Journal of Consulting and Clinical Psychology, 79, 421– 425. doi:10.1037/a0023318
Akaike, H. (1987). Factor analysis and AIC.Psychometrika, 52,317–332. doi:10.1007/bf02294359
PE
PE/CR
Time
(Session)
Posttraumatic
Symptoms
Depressive
Symptoms
B 0.04*** A -0.26*** C’ -0.13***Time
(Session)
Posttraumatic
Symptoms
Time
(Session)
Posttraumatic
Symptoms
Depressive
Symptoms
B 0.05*** A -0.34***Time
(Session)
Posttraumatic
Symptoms
C -0.41*** C’ -0.18*** C -0.34***Figure 2. Reverse lagged mediational models. Numbers are standardized regression coefficients. Paths indicate Level 1 regression analyses. Depressive symptoms were lagged so that depressive symptoms at timetpredicted posttraumatic symptoms at timet11. PE5prolonged exposure; PE/CR5prolonged exposure with added cognitive restructuring. pppp,.001.
Alloy, L. B., Kelly, K. A., Mineka, S., & Clements, C. M. (1990). Comorbidity of anxiety and depressive disorders: A helplessness– hopelessness perspective. In J. D. Maser & C. R. Cloninger (Eds.),
Comorbidity of mood and anxiety disorders(pp. 499 –544). Washington, DC: American Psychiatric Association.
American Psychiatric Association. (2000).Diagnostic and statistical man-ual of mental disorders(4th ed., text rev.). Washington, DC: Author. Beck, A. T., & Emery, G. (1985). Anxiety disorders and phobias: A
cognitive perspective. New York, NY: Basic Books.
Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression.Archives of General Psychiatry, 4,561–571. doi:10.1001/archpsyc.1961.01710120031004
Brown, T. A., Campbell, L. A., Lehman, C. L., Grisham, J. R., & Mancill, R. B. (2001). Current and lifetime comorbidity of theDSM–IVanxiety and mood disorders in a large clinical sample. Journal of Abnormal Psychology, 110,585–599. doi:10.1037/0021-843X.110.4.585 Elhai, J. D., Franklin, C. L., & Gray, M. J. (2008). The SCID PTSD
module’s trauma screen: Validity with two samples in detecting trauma history.Depression and Anxiety, 25,737–741. doi:10.1002/da.20318 Erickson, D. J., Wolfe, J., King, D. W., King, L. A., & Sharkansky, E. J.
(2001). Posttraumatic stress disorder and depression symptomatology in a sample of Gulf War veterans: A prospective analysis.Journal of Consulting and Clinical Psychology, 69,41– 49. doi:10.1037/0022-006X.69.1.41 Feingold, A. (2009). Effect sizes for growth-modeling analysis for
con-trolled clinical trials in the same metric as for classical analysis. Psy-chological Methods, 14,43–53. doi:10.1037/a0014699
Feske, U. (2008). Treating low-income and minority women with posttraumatic stress disorder: A pilot study comparing prolonged exposure and treatment as usual conducted by community therapists.Journal of Interpersonal Vio-lence, 23,1027–1040. doi:10.1177/0886260507313967
First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. (1995).
Structured Clinical Interview for DSM–IV Axis I Disorders—Patient Edition (SCID–I/P, Version 2). New York: New York State Psychiatric Institute, Biometrics Research Department.
Foa, E. B., Hembree, E. A., Cahill, S. P., Rauch, S. A. M., Riggs, D. S., Feeny, N. C., & Yadin, E. (2005). Randomized trial of prolonged exposure for posttraumatic stress disorder with and without cognitive restructuring: Out-come at academic and community clinics.Journal of Consulting and Clin-ical Psychology, 73,953–964. doi:10.1037/0022-006X.73.5.953 Foa, E. B., Hembree, E. A., & Rothbaum, B. O. (2007). Prolonged
exposure therapy for PTSD: Emotional processing of traumatic experi-ences: Therapist guide. New York, NY: Oxford University Press. Foa, E. B., & Rauch, S. A. M. (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, 72,879–884. doi:10.1037/0022-006X.72.5.879 Foa, E. B., Riggs, D. S., Dancu, C. V., & Rothbaum, B. O. (1993).
Reliability and validity of a brief instrument for assessing post-traumatic stress disorder.Journal of Traumatic Stress, 6,459–473. doi:10.1002/jts.2490060405 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, 750 –755. doi: 10.1007/s10608-011-9423-6
Harvey, A. G., Bryant, R. A., & Tarrier, N. (2003). Cognitive behaviour therapy for posttraumatic stress disorder.Clinical Psychology Review, 23,501–522. doi:10.1016/S0272-7358(03)00035-7
Kenny, D. A., Korchmaros, J. D., & Bolger, N. (2003). Lower level mediation in multilevel models. Psychological Methods, 8,115–128. doi:10.1037/1082-989X.8.2.115
Kessler, R. C., Berglund, P., Demler, O., Jin, R., Merikangas, K. R., & Walters, E. E. (2005). Lifetime prevalence and age-of-onset distributions of
DSM–IVdisorders in the National Comorbidity Survey Replication.Archives of General Psychiatry, 62,593– 602. doi:10.1001/archpsyc.62.6.593
Kessler, R. C., Sonnega, A., Bromet, E., Hughes, M., & Nelson, C. B. (1995). Posttraumatic stress disorder in the National Comorbidity Sur-vey. Archives of General Psychiatry, 52, 1048 –1060. doi:10.1001/ archpsyc.1995.03950240066012
King, D. W., King, L. A., McArdle, J. J., Shalev, A. Y., & Doron-LaMarca, S. (2009). Sequential temporal dependencies in associations between symp-toms of depression and posttraumatic stress disorder: An application of bivariate latent difference score structural equation modeling.Multivariate Behavioral Research, 44,437– 464. doi:10.1080/00273170903103308 Kreft, I., & de Leeuw, J. (1998).Introducing multilevel modeling.
Thou-sand Oaks, CA: Sage.
Loas, G. (1996). Vulnerability to depression: A model centered on anhe-donia. Journal of Affective Disorders, 41,39 –53. doi:10.1016/0165-0327(96)00065-1
Lobbestael, J., Leurgans, M., & Arntz, A. (2011). Inter-rater reliability of the Structured Clinical Interview forDSM–IVAxis I Disorders (SCID I) and Axis II Disorders (SCID II).Clinical Psychology & Psychotherapy, 18,75–79. doi:10.1002/cpp.693
MacKinnon, D. P., Fritz, M. S., Williams, J., & Lockwood, C. M. (2007). Distribution of the product confidence limits for the indirect effect: Program PRODLIN. Behavior Research Methods, 39,384 –389. doi: 10.3758/BF03193007
Moscovitch, D. A., Hofmann, S. G., Suvak, M. K., & In-Albon, T. (2005). Mediation of changes in anxiety and depression during treatment of social phobia.Journal of Consulting and Clinical Psychology, 73,945– 952. doi:10.1037/0022-006X.73.5.945
Powers, M. B., Halpern, J. M., Ferenschak, M. P., Gillihan, S. J., & Foa, E. B. (2010). A meta-analytic review of prolonged exposure for post-traumatic stress disorder. Clinical Psychology Review, 30, 635– 641. doi:10.1016/j.cpr.2010.04.007
Rauch, S. A. M., Defever, E., Favorite, T., Duroe, A., Garrity, C., Martis, B., & Liberzon, I. (2009). Prolonged exposure for PTSD in a Veterans Health Administration PTSD clinic.Journal of Traumatic Stress, 22,
60 – 64. doi:10.1002/jts.20380
Raudenbush, S. W. (2001). Comparing personal trajectories and drawing causal inferences from longitudinal data.Annual Review of Psychology, 52,501–525. doi:10.1146/annurev.psych.52.1.501
Resick, P. A., Nishith, P., Weaver, T. L., Astin, M. C., & Feuer, C. A. (2002). A comparison of cognitive-processing therapy with prolonged exposure and a waiting condition for the treatment of chronic posttrau-matic stress disorder in female rape victims.Journal of Consulting and Clinical Psychology, 70,867– 879. doi:10.1037/0022-006X.70.4.867 Rothbaum, B. O., Cahill, S. P., Foa, E. B., Davidson, J. R. T., Compton, J.,
Connor, K. M., . . . Hahn, C.-G. (2006). Augmentation of sertraline with prolonged exposure in the treatment of posttraumatic stress disorder.
Journal of Traumatic Stress, 19,625– 638. doi:10.1002/jts.20170 Schindel-Allon, I., Aderka, I. M., Shahar, G., Stein, M., &
Gilboa-Schechtman, E. (2010). Longitudinal associations between post-traumatic distress and depressive symptoms following a post-traumatic event: A test of three models. Psychological Medicine, 40,1669 –1678. doi: 10.1017/S0033291709992248
Shalev, A. Y., Freedman, S., Peri, T., Brandes, D., Sahar, T., Orr, S. P., & Pitman, R. K. (1998). Prospective study of posttraumatic stress disorder and depression following trauma.The American Journal of Psychiatry, 155,630–637. Tuerk, P. W., Yoder, M., Grubaugh, A., Myrick, H., Hamner, M., &
Acierno, R. (2011). Prolonged exposure therapy for combat-related posttrau-matic stress disorder: An examination of treatment effectiveness for veterans of the wars in Afghanistan and Iraq.Journal of Anxiety Disorders, 25,397– 403. doi:10.1016/j.janxdis.2010.11.002
Received July 3, 2011 Revision received October 31, 2012
