- Email: [email protected]
Posttraumatic stress disorder and depressive symptoms: Joined or independent sequelae of trauma?
Journal of Psychiatric Research 54 (2014) 64e69
Contents lists available at ScienceDirect
Journal of Psychiatric Research journal homepage: www.elsevier.com/locate/psychires
Posttraumatic stress disorder and depressive symptoms: Joined or independent sequelae of trauma? Sharon Dekel a, b, *, Zahava Solomon c, Danny Horesh d, e, Tsachi Ein-Dor f a
Department of Psychiatry, Harvard Medical School, Boston, MA, USA Department of Psychiatry, Massachusetts General Hospital, Charlestown, MA, USA c Bob Shapell School of Social Work, Tel Aviv University, Tel Aviv, Israel d Department of Psychology, Bar Ilan University, Ramat Gan, Israel e Department of Psychiatry, New York University, NY, USA f School of Psychology, Interdisciplinary Center Herzliya, Herzliya, Israel b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 1 May 2013 Received in revised form 6 March 2014 Accepted 7 March 2014
Objective: The nature of co-morbidity between posttraumatic stress disorder (PTSD) and depression has been the subject of much controversy. This study addresses this issue by investigating associations between probable PTSD and depressive symptoms in a prospective, longitudinal sample of combat veterans. Method: Symptoms of PTSD and depression were assessed at 3 points of time (i.e., 1991, 2003, 2008) over a period of 17 years utilizing the PTSD Inventory and the SCL-90 (Derogatis, 1977). Two groups of combat veterans, 275 former prisoners of war (ex-POWs) and 219 matched combatants (controls), were assessed. Data were analyzed using descriptive statistics, latent variable modeling, and confirmatory factor analysis. Results: A series of c2 tests revealed that the prevalence proportions of depressive symptoms and probable PTSD were higher among ex-POWs compared to controls at all time points. The prevalence of depressive symptoms was higher than the prevalence of PTSD symptoms in both groups at the each of the times. Latent Trajectories Modeling (LTM) indicated that while ex-POWs’ PTSD symptom severity increased over time, the severity of symptoms remained stable among controls. Parallel Process Latent Growth Modeling (PLGM) revealed a positive bi-directional relationship whereby PTSD symptoms mediated the affect of captivity on depressive symptoms and depressive symptoms mediated the affect of captivity on PTSD symptoms over time. Utilizing Confirmatory Factor Analysis (CFA), a single factor model emerged for depressive and PTSD symptoms. Conclusion: The findings suggest that while depression and PTSD seem to be different long-term manifestations of traumatic stress, accounted for in part by the severity of the trauma, they both may be parts of a common general traumatic stress construct. Clinical and theoretical implications of these findings are discussed. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: PTSD Depression POWs Captivity Combat trauma Longitudinal study
Posttraumatic stress disorder (PTSD) is recognized as a distinguished psychiatric disorder precipitated by exposure to a traumatic event (DSM-V, American Psychiatric Association, 2013). However, over the years, the distinctiveness of PTSD has generated much controversy (McNally, 2003; Spitzer et al., 2007). Studies consistently reveal that PTSD is often accompanied by other disorders with the most notable being the association with depression (e.g., Campbell et al., 2007). The rates of PTSD in individuals
* Corresponding author. PTSD Research Laboratory, Massachusetts General Hospital-East, 120 Second Avenue Charlestown, MA 02129, USA. E-mail addresses: [email protected], [email protected] (S. Dekel). http://dx.doi.org/10.1016/j.jpsychires.2014.03.003 0022-3956/Ó 2014 Elsevier Ltd. All rights reserved.
reporting a lifetime history of depression range between 37% and 48% (Breslau et al., 1997; Kessler et al., 1995); and among clinical samples, as much as 68% of individuals with PTSD meet diagnostic criteria for depression at some point in time (see Keane and Kaloupek, 1997; for a review). Co-occurrence of PTSD and depression has a negative impact on prognosis including hospitalization, disability, and suicide attempts (Sher, 2005). Despite the clinical significance and the body of existing research, the mechanisms underlying the association between PTSD and depression are not fully understood. According to one school of thought, PTSD and depression embody a single, “general traumatic stress” construct (O’Donnell et al., 2004; Sher, 2005). Supporting evidence suggests that one
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
disorder may be a derivate of the other. Suffering from prior depression was fond to increase the risk of developing PTSD (Breslau et al., 1997; O’Teelo et al., 1998), while other studies documented that PTSD renders the individuals vulnerable to first onset of depression (Breslau et al., 2000; Ginzburg et al., 2010; Kessler et al., 1995). Furthermore, similar risk factors (e.g., severity of trauma, negative event appraisal) have been documented to set in motion both depression and PTSD (O’Donnell et al., 2004). A shared genetic and biological basis for PTSD and depression has been suggested (Gaffney, 2003; Davidson et al., 1998). A contrary view deems PTSD and depression independent trauma sequelae (e.g., Shalev et al., 1998). Relatively high rates of depression in the short-term following trauma (i.e., 19% among survivors of various traumas) were reported among individuals with no PTSD diagnosis (Shalev et al., 1998). Similarly, there is evidence indicating that patients with a history of trauma report more episodes of depression than those without such history (Zlotnick et al., 1997). These findings may be taken to suggest that depression is first and foremost a derivate of exposure to trauma, rather than of PTSD per-se. Furthermore, some biological studies support the notion of two independent trauma sequelae with evidence that the two disorders entail opposing dysregulations in the body’s stress response (Oquendo et al., 2003; Yehuda, 2002). As can be seen, the attempt to understand the observed association between PTSD and depression has so far yielded inconclusive results. This may be attributed, in large part, to methodological shortcomings of previous studies. Most notably, the reliance on cross-sectional designs using retrospective reports (e.g., Kessler et al., 1995; Zlotnick et al., 1997), which are subject to memory bias in reporting the course of symptoms over time (Dekel and Bonanno, 2013) rather than using longitudinal assessments (e.g., Grieger et al., 2006); and the study of homogenous trauma samples which ignores the probable affect of event severity on outcomes. Grieger et al. (2006) showed that in the very short-term following exposure PTSD and depressive symptoms tend to escalate over time, and yet, to the best of our knowledge, no study looked into the interplay between the two disorders longitudinally following trauma in the very long-term and with respect to the severity of the trauma. This prospective study examined Israeli ex-POWs and comparable combat veterans of the 1973 Yom Kippur War who had not been incarcerated at three time points: 1991, 2003, and 2008. We set to evaluate the prevalence and course of symptoms of PTSD and depression, as well as the long-term interrelationships between the symptoms. Our research aims were (1) To assess the prevalence of symptoms of PTSD and depression at three points in time and the changes in prevalence over time. (2) To assess the interrelationship of symptoms of PTSD and depression, at each assessment and across time and (3) To assess the bi-directional associations between symptoms of PTSD and depression across time. 1. Method This study uses data from a longitudinal study on the psychological effects of war (Dekel et al., 2012). A cohort of Israeli male veterans who participated in the 1973 Yom Kippur War were followed over 17 years with assessment at three time points: 1991 (T1), 2003 (T2), and 2008 (T3). Following approval from both the Israel Defense Forces (IDF) and Tel Aviv University review boards, lists of potential participants from the IDF computerized data bank were composed. We then phoned those participants and, after explaining the purpose of the study, asked them to take part in the assessments. The questionnaire packet was administered in their homes or in another location of their choice. Informed consent was obtained from all participants.
65
A target population of 520 representative combat veterans was comprised of two groups: individuals who were captured during the war and consequently subjected to torture, harassment and humiliations (ex-POWs: n ¼ 240); and veterans who participated in the same war but were not taken captive (combats: n ¼ 280). Both groups were exposed to battlefield stressors entailing active fighting and encounters with death. Groups were matched on military background and socio-demographic status with no differences in combat exposure and pre- and post war negative life events (see Dekel et al., 2012; for details on sampling method). The sample included 164 ex-POWs who participated in the first assessment, and 144 in the second (10 could not be located/refused, 4 had died, and 6 could not participate due to mental deterioration). At the third assessment, the original group of 240 ex-POWs was contacted again. Of these 171 participated (11 could not be located, 20 had died, 11 were living abroad, 21 refused to participate and 6 could not participate due to their mental status). 185 combatants participated in T1, 143 in T2 (41 could not be located and 1 had died), and 118 in T3 (20 could not be located/refused and 5 had died). In the GLM-based analyses, missing data was handled using multiple imputation (Rubin, 1987; Schafer, 1997). In the SEM-based analyses missing data was handled with the case-wise maximum likelihood estimation. For all participants, the mean age was 53.4 (SD ¼ 4.4); years of schooling were 14.02 (SD ¼ 3.41); the majority were secular (67%) with an average income (62%) in T2. No significant differences were found between those who participated in the follow-up assessments with regard to initial PTSD level, military rank, age, and education. No difference was found between the study groups in post-war level of exposure to potentially traumatic events (see Dekel et al., 2013a; for details about the measurement). 1.1. Measures PTSD symptoms were assessed using the PTSD Inventory (Dekel et al., 2012), a self-report measure corresponding to the 17 PTSD symptoms listed in the DSM-III-R (APA, 1987) which was the standard of practice when the study commenced. Participants were asked to indicate the frequency they had each symptom in the past month anchored on their 1973 war experiences, on a 4-point scale ranging from 1 (not at all) to 4 (I usually did), with 3 (often) indicating cutoff for symptom endorsement. For each participant, we calculated the number of symptoms (intrusion, avoidance and hyper-arousal) that he endorsed in each time point. In order to conform to the DSM-IV symptom clusters, we classified individuals as endorsing “Probable PTSD” by having at least 1 intrusion, 3 avoidance and 2 hyper-arousal symptoms (“physiological reactivity to resembling events” symptom of the DSM-III-R was categorized under the intrusion cluster). The inventory has proven psychometric properties in terms of high test-retest reliability concurrent and convergent validity. In the present study, reliability values for PTSD symptoms and symptom clusters (intrusion, avoidance and hyper-arousal) were high (Cronbach’s a ranged between .78 and .90, .88 and .95, and .87 and .96, respectively, in T1, T2, and T3). Depressive symptoms were assessed using the depression subscale of the Symptom Checklist-90 (SCL-90, Derogatis, 1977). Participants were asked to indicate how frequently they experienced each symptom during the last 2 weeks on a 5-point distress scale. Based on norms for psychiatric outpatients (Derogatis, 1977) scores equal to or above .73 were considered as an indication for endorsement of depressive symptoms (Dekel et al., 2013a, 2013b, 2013c). For each participant, we calculated the average frequency of experiencing depressive symptoms in each of the time points. The SCL-90 has high concurrent validity and the specific subscales display high empirical agreement across various samples
66
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
Table 1 Rates of PTSD and depression, test statistics and relative risk for examining the differences between ex-POWs and controls. Time
Variable PTSD
Depression
Rate: ex-POWs [controls] T1(1991;
N ¼ 325)
T2(2003;
N ¼ 227)
T3
(2008; N ¼ 287)
8.7% (N ¼ 14) [3% (N ¼ 5)] 29.9% (N ¼ 35) [1.9% (N ¼ 2)] 34.7% (N ¼ 58) [2.5% (N ¼ 3)]
X2(1)
Relative risk [95% CI]
4.77*
2.87 [1.06, 7.78]
p .038
31.57***
15.86 [3.91, 64.33]
<.0001
42.59***
13.66 [4.38, 42.56]
<.0001
Rate: ex-POWs [controls]
X2(1)
Relative risk [95% CI]
26.3% (N ¼ 42) [10.9% (N ¼ 18)] 69.9% (N ¼ 86) [19.2% (N ¼ 20)] 67.1% (N ¼ 114) [18.8% (N ¼ 22)]
12.7***
2.41 [1.42, 4.00]
.0007
58.17***
3.64 [2.41, 5.48]
<.0001
64.73***
3.57 [2.41, 5.27]
<.0001
p
Note: *p < .05, ***p < .001.
(Derogatis et al., 1976). In this study, reliability values were high in all assessments (Cronbach’s a was .92, .94, .90, respectively, in T1, T2, and T3). 1.2. Data analysis We examined the link between PTSD and depressive symptoms from various complimentary perspectives. First, we examined whether ex-POWs and controls differ in the prevalence of probable PTSD and depressive symptoms in each of the time points using c2 tests of independence and relative risk analysis. Then, we examined whether and to what extent the prevalence of depressive symptoms and Probable PTSD had changed over time using McNemar tests. In these tests, we also examined the contribution of captivity (ex-POWs vs. controls) in moderating these effects. Second, we examined whether ex-POWs and controls differ in the prevalence of co-occurring PTSD and depressive symptoms in each of the time points using c2 tests of independence (including odd ratios). Third, we examined whether PTSD and depressive symptoms are two distinct psychopathological constructs or are they one construct of internalizing-related psychopathology. To this end, we employed multi-group Confirmatory Factor Analysis (CFA). Fourth, we examined the contribution of captivity (exPOWs vs. controls) to the differences in the 17-year trajectories of PTSD and depressive symptoms using Latent Trajectories Modeling. Finally, we employed parallel process latent growth modeling to examine the contribution of captivity (ex-POWs vs. controls) to the bidirectional associations between PTSD and depressive symptoms over time. We also examined the bidirectional associations between PTSD and depressive symptoms over time regardless of captivity. 2. Results 2.1. Prevalence of probable PTSD and depressive symptoms Table 1 presents the prevalence of probable PTSD and depressive symptoms, test statistics and relative risk values. The analyses revealed that the prevalence proportions of depressive
symptoms and probable PTSD were higher among the ex-POWs than controls in T1, T2 and T3. Also, the prevalence of depressive symptoms was higher than the prevalence of PTSD symptoms in both groups at the each of the time points. However, importantly, as can be seen, the relative risk of ex-POWs to suffer from PTSD symptoms was higher than depressive symptoms at T2 and T3. 2.2. Changes in prevalence of probable PTSD and depressive symptoms over time In this section, we examined separately for ex-POWs and controls the changes in the prevalence of PTSD and depressive symptoms between T1 and T2, and between T2 and T3. McNemar tests revealed that the controls’ PTSD symptoms prevalence was constant over time (ps ¼ 1). Conversely, the analyses revealed that exPOWs’ PTSD prevalence significantly increased between T1 and T2 (p < .0001); yet, it remained constant between T2 and T3 (p ¼ .13). Also, McNemar tests revealed that for both the controls and exPOWs depressive symptoms prevalence significantly increased between T1 and T2 (p < .05 for controls and p < .0001 for exPOWs); yet, it remained constant between T2 and T3 (p ¼ 1 for controls and p ¼ .84 for ex-POWs). 2.3. Co-occurrence of probable PTSD and depressive symptoms over time In this section, we examined separately for ex-POWs and controls, the prevalence of the co-occurrence of probable PTSD and depressive symptoms in T1, T2, and T3. Frequency of co-occurrence and test statistics are presented in Table 2. The analyses revealed significant differences in ex-POWs’ and controls’ prevalence of PTSD-depression co-occurrence at all time points. As can be seen, ex-POWs’ had higher co-occurrence rates. In addition, a series of McNemar tests revealed that in both groups the frequency of co-occurring depressive and PTSD- symptoms significantly increased from T1 to T2 (ps < .05) but remained unchanged between T2 and T3 (ps > .35).
Table 2 Rates of co-morbidity between PTSD and depression among ex-POWs and controls. ex-POWs
T1 T2 T3
Controls
95% CI of Exp(B)
Co-morbidity
Either depression or PTSD
Neither
Co-morbidity
Either depression or PTSD
Neither
X2(2)
8.8% (14) 26.5% (31) 30.9% (50)
17% (27) 47% (55) 38.9% (63)
74.2% (118) 26.5% (31) 30.25% (49)
3% (5) 1% (1) 2.6% (3)
7.9% (13) 18.3% (19) 16.2% (19)
89.1% (147) 80.8% (84) 81.2% (95)
12.22* 69.54*** 74.67***
Note: Exp(B) ¼ odd ratio. Ns are presented in parentheses.
3.08 [1.09, 8.78] 37.13 [4.97, 277.61] 16.96 [5.14, 55.98]
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
67
Table 3 CFA loadings by study group. ex-POWs
T1 T2 T3
Controls
Avoidance
Intrusive
Hyper-arousal
Depression
Avoidance
Intrusive
Hyper-arousal
Depression
.78 .75 .77
.89 .67 .74
.89 .86 .86
.71 .75 .70
.64 .67 .78
.62 .62 .58
.71 .72 .83
.74 .81 .78
Note: scaled c2 (Yuan-Bentler; 120) ¼ 105.43, p ¼ .83, CFI ¼ 1, NNFI ¼ 1, RMSEA ¼ 0.
2.4. Interrelationships between PTSD and depressive symptoms over time To examine whether PTSD and depression are two different constructs or if they represent one general construct, we ran a Confirmatory Factor Analysis (CFA) with three latent variables: one for each of the three waves of measurement. We estimated the appropriateness of the model using EQS 6.1 SEM software (Bentler and Wu, 1995). The model’s fit was assessed by the comparative fit index (CFI), Bentler-Bonett non-normed fit index (NNFI) and the root-mean-square error of approximation (RMSEA). A model is judged as reasonably fitting the data when CFI and NNFI are higher than .95 and the RMSEA is lower than .05 (Bollen and Curran, 2006). Missing data was handled with the case-wise maximum likelihood estimation. As seen in Table 3, the analysis revealed that the multi-group CFA had excellent fit to the observed data, indicating that PTSD avoidance, intrusive, and hyper-arousal clusters as well as depressive symptom level were significantly loaded to one construct at all time points.
2.5. Trajectories of PTSD and depressive symptoms over time To examine changes in PTSD and depressive symptoms over time, we conducted a series of Latent Trajectories Modeling (LTM: see Bollen and Curran, 2006). Two latent factors were estimated: the first, defines the initial levels of PTSD and depression (i.e., intercept), and the second, explores the constant of the trajectory of change in PTSD and depression over time (i.e., linear; time was coded as 0, 12 and 17). In addition, we assessed whether the LTMs were different for ex-POWs and controls by using a multi-group SEM procedure. LTMs results are presented in Table 4. The analysis revealed that ex-POWs’ PTSD level annually increased since the initial assessment in 1991, and that controls’ PTSD level remained unchanged. The LTM estimating the ex-POWs’ depression trajectory showed poor fit to the observed data, and thus, ex-POWs’ depression level did not change linearly over time. To account for this non-linear change, we estimated an additional LTM in which we fixed only the first (coded 0) and the last (coded Table 4 LTMs results for examining changes in PTSD and depression over time. Change (value of the linear slope) ex-POWs PTSD
.13***
Depression
N/A
Scaled c2
CFI
NNFI
RMSEA
104.44** [51.16*]
.98 [1.0] .93 [1.0]
.97 [1.0] .70 [1.0]
.05 [.0] .13 [.0]
Controls .02 01**
37.71** [.19]
Note: controls’ fit indices are in brackets. CFI ¼ comparative fit index; NNFI ¼ BentlereBonett non-normed fit index; RMSEA ¼ root-mean-square error of approximation.
17) time points, while freely estimating the second time point. The model has an excellent fit to the observed data [c2 (4) ¼ 4.68, p ¼ .32, CFI ¼ 1, TLI ¼ .99, RMSEA ¼ .04.], and indicated that exPOWs’ depression level increased significantly between T1 and T2 and then remained unchanged between T2 and T3. In comparison, controls’ depression level annually increased since 1991. 2.6. Bidirectional associations between PTSD and depressive symptoms over time To examine the bidirectional association between PTSD and depression from T1, through T2 to T3, we employed parallel process latent growth modeling (Cheong et al., 2003), which is extension of LTM. Because PTSD consists of three clusters (intrusion, avoidance, and hyper-arousal), we used latent variables in SEM environment to represent the PTSD underlying phenomenon. The model was tested by MPlus 6.1 (Muthén & Muthén, 1998e2010). Specifically, we examined whether captivity (ex-POWs ¼ 1 vs. controls ¼ 0) predicts a change in PTSD, and whether this change, in turn, predicts a change in depression (i.e., whether the change in PTSD mediates the link between captivity and later depression). In this model, we also examined the opposite process by which captivity (ex-POWs ¼ 1 vs. controls ¼ 0) predicts a change in depression, which in turn predicts a change in PTSD. The model had an excellent fit for the observed data, c2 (8) ¼ 10.68, p ¼ .22, CFI ¼ 1, TLI ¼ .99, RMSEA ¼ .04. The analysis revealed that ex-POWs, on average, had more PTSD symptoms severity (b ¼ .76, b ¼ .16, p < .05) and depression (b ¼ .14, b ¼ .16, p < .05) in T1 than controls (i.e., greater intercepts). Ex-POWs, on average, also suffered from a greater increase in PTSD over time than controls (b ¼ .39, b ¼ .76, p < .0001) but not in depression (at least not directly; b ¼ .01, b ¼ .14, p ¼ .19). Captivity did influence the trajectory of depression via an effect on PTSD: Bias-corrected bootstrap analyses indicated that the increase in PTSD symptoms severity over time mediated the effect of captivity on depression, b ¼ .06, p < .0001, 99% CI [.03, .09], such that captivity was linked with greater increase in PTSD symptoms severity over time, which in turn was associated with a greater increase in depression over time. The model revealed an additional mediation process regarding the bidirectional association between PTSD and depression. Bias-corrected bootstrap analyses indicated that captivity increased the level of depression in T1 (an effect of captivity on the intercept of depression), which in turn was related with a greater increase in PTSD symptoms severity over time, b ¼ .20, p ¼ .034, 99% CI [.005, .08]. Next, we examined the parallel processes of depressive and PTSD symptoms over time without recognizing captivity. The model had an excellent fit for the observed data, c2 (5) ¼ 8.64, p ¼ .12, CFI ¼ 1, TLI ¼ .99, RMSEA ¼ .06, and revealed that higher level of depression in T1 was associated with a greater increase in PTSD symptoms severity over time, b ¼ .33, p < .001. A greater increase in depression over time was also associated with a greater increase in PTSD symptoms severity over time, b ¼ 3.19, p < .001. Conversely, PTSD symptoms severity in T1 was not associated with a change in depression over time, b ¼ .001, p ¼ .37. Rather, a greater
68
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
increase in PTSD over time was also associated with a greater increase in depression over time, b ¼ .13, p < .001. 3. Discussion This study set out to examine the long-term prevalence, course, and interrelations of symptoms of PTSD and depression among exPOWs and comparable controls. The prevalence of depressive symptoms and probable PTSD, as well as their co-occurrence, was higher among ex-POWs than controls at all three waves of measurement. Results also indicate that the prevalence of depressive symptoms was higher than that of probable PTSD among both study groups, at all three assessments. However, while the prevalence of depressive symptoms increased over time in both study groups, PTSD symptoms increased only among ex-POWs, but not among controls. Likewise, captivity contributed to the bidirectional association between symptoms of PTSD and depression over time. Importantly, in accord with the notion of a unified traumatic stress construct, our findings show that symptoms of depression and PTSD were inter-related across times. The finding that ex-POWs endorsed higher depressive symptom severity and probable PTSD is consistent with previous research, as well as earlier investigations of this sample conducted by our group (e.g., Dekel et al., 2013a, 2013b, 2013c). Several explanations may exist. First, captivity is a particularly severe traumatic experience, which entails hardships such as torture and humiliation that are inter-personal, deliberately inflicted and repeatedly experienced. Second, in this sample, the ex-POWs were exposed to a ‘double trauma’ pertaining to both captivity and combat exposure. Our finding that ex-POWs also reported more co-occurrence of depressive and PTSD symptoms suggests that their pathology entails a more severe type of posttraumatic disorder (Dekel et al., 2013a, 2013b, 2013c) or complex PTSD (Herman, 1992). While depression has been conceived as a derivate of PTSD (Kessler et al., 1995), our findings suggest that in relation to sever trauma, such as the experience of captivity, symptoms of depression may subsequently worsen symptoms of PTSD. This is consistent with some prospective short-term assessments showing that depression fuels PTSD (Erikson et al., 2001; King et al., 2009). Dysphoria symptoms, entailing a sense of helplessness and loss of mastery, may prevent seeking social support, self-disclosure, and other self-efficacious behaviors that buffer against the development of symptoms of PTSD (King et al., 2009). Moreover, sleep and concentration impairments seen in depression may trigger a hypervigilance response (King et al., 2009). The contribution of our findings is in underscoring that the severity of event affects the nature of the bi-directional relationships between PTSD and depressive symptoms. The findings suggest that severe traumatic exposure not only yields higher levels of psychopathology at a given time point, but also entails enduring long-term associations between different psychopathological conditions. Interestingly, our findings show that the prevalence of depressive symptoms increased among both study groups, while that of probable PTSD increased only among ex-POWs. While PTSD is the most common psychiatric sequela of trauma in the initial period following exposure (e.g., Shalev et al., 1998), a different pattern appears evident in the long-term, with depressive symptoms emerging as the dominant clinical response. This may be explained by Seyle’s (1956) conceptualization of a general adaptation syndrome in response to an external stressor. First, the individual exhibits an alarmed and the fight-or flight response. As the stressor persists, the individual enters a resistance state and attempts to deploy means of coping. In the final state, exhaustion, resources are depleted and normal functioning is impaired. When this phase is
prolonged, it may manifest in long-term pathogenic outcomes and the development of depressive symptoms. The findings that depressive symptoms may worsen over time among individuals exposed to relatively moderate traumatic events, similarly to survivors of extreme and prolonged trauma, suggests that unlike PTSD (e.g., Solomon and Dekel, 2005), the enduring nature of depressive symptoms is possibly attributed primarily to post-event factors such as level of social support rather than to the level of exposure and other peritraumatic factors. Finally, our findings revealed a single underlying dimension for depressive and PTSD symptoms in all three waves of measurement. While we cannot fully negate an illusionary unified post-trauma construct merely due to symptom overlap (Franklin and Zimmerman, 2001; Southwick et al., 1991); in a recent study, a single factor emerged even after removing shared symptoms and symptoms of PTSD that are depression-related (Elhai et al., 2011). This may suggest that PTSD and depression share a common diathesis, possibly indicative of a single general traumatic response (Breslau et al., 2000; O’Donnell et al., 2004). This study has several methodological limitations. First, it does not allow inferences regarding causal relations between depressive and PTSD symptoms following trauma. Second, in spite of the strength of longitudinal designs, there are inherent limitations. Not all participants took part in all assessments, although the samples did not differ with respect to socio-demographic variables. Since we focused on the long-term phase, we may have overlooked fine fluctuations in depressive and PTSD symptoms given short interval assessments in the immediate phase and earlier exposure to trauma. Finally, our use of self-report measures, although well validated, rather than having obtained clinical diagnosis, leaves open the possibility of overestimating symptom prevalence. Future studies may benefit from the inclusion of clinical assessments, as well as from the examination of additional factors associated with co-occurring PTSD-depression, such as psychiatric history, personality traits, and cognitive variables. Nevertheless, the present study makes an important contribution to the literature regarding the co-occurrence of PTSDdepression. The repeated evidence in support of a tight interplay between symptoms of PTSD and depression assessed in three time points following the war in a veterans sample may indicate that PTSD symptoms are not the only direct outcome of trauma exposure. Instead, survivors of trauma may experience a general PTSDdepression response to include symptoms of re-experiencing, avoidance, and arousal along with negative alteration in mood, according, in part, with recent criteria revisions (DSM-V) for PTSD diagnosis. In practical terms, our findings suggest that mental health professionals may benefit from a close monitoring of depressive symptoms, as those may be a clinical marker of ensuing symptoms of PTSD. Role of the funding source Part of this work has been funded by the NARSAD Young Investigator Grant sponsored by Dr. Dylan Tauber which was awarded to Sharon Dekel. Contributors Dr. Sharon Dekel initiated and generated the study paradigm with respect to the sample, aims, measurements, and design, and prepared the manuscript. Prof. Zahava Solomon supervised the research project. Dr. Tsachi Ein-Dor conducted the statistical analysis. Dr. Danny Horesh contributed to the writing of the manuscript.
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
Conflict of interest The authors of the study have no conflict of interest. Acknowledgment We thank Dr. Talya Greene who assisted in the grammatical editing of the manuscript. References American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 3rd ed., revised. Washington, DC: American Psychiatric Press; 1987. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Washington, DC: American Psychiatric Press; 2013. Bentler PM, Wu EJC. EQS for Windows user’s guide. Encino, CA: Multivariate Software, Inc; 1995. Bollen KA, Curran PJ. Latent curve models. A structural equation perspective. Hoboken, NJ: Wiley; 2006. Breslau N, Davis CG, Peterson EL, Schultz LR. Psychiatric sequelae of post-traumatic stress disorder in women. Arch Gen Psychiatry 1997;54:81e7. Breslau N, Davis GC, Peterson EL, Schultz LR. A second look at comorbidity in victims of trauma: the post-traumatic stress disorder-major depression connection. Biol Psychiatry 2000;48:902e9. Campbell DG, Felker BL, Liu CF, Yano E, Kirchner JE, Chan D, et al. Prevalence of depressionePTSD comorbidity: Implications for clinical practice guidelines and primary care-based interventions. J Gen Intern Med 2007;22:711e8. Cheong J, MacKinnon DP, Khoo ST. Investigation of mediational processes using parallel process latent growth curve modeling. Struct Equ Model 2003;10:238e62. Davidson JR, Tupler LA, Wilson WH, Connor KM. A family study of chronic posttraumatic stress disorder following rape trauma. J Psychiatric Res 1998;32: 301e9. Dekel S, Bonanno G. Changes in trauma memory and patterns of posttraumatic stress. Psychol Trauma Theory Res Pract Policy 2013;5:26e34. Dekel S, Ein-Dor T, Solomon Z. Posttraumatic growth and posttraumatic distress: a longitudinal study. Psychol Trauma Theory Res Pract Policy 2012;4:94e101. Dekel S, Mandl C, Solomon Z. Is the holocaust implicated in PTG in second generation holocaust survivors? A prospective study. J Trauma Stress 2013a;26:1e4. Dekel S, Peleg T, Solomon Z. The relationship of PTSD to negative cognitions: a 17year longitudinal study. Psychiatry Interpers Biol Process 2013b;76:241e55. Dekel S, Solomon Z, Rozenstreich E. Secondary salutogenic effects in veterans whose parents were holocaust survivor? J Psychiatric Res 2013c;47:266e71. Derogatis LR. SCL-90: Administration, scoring, and procedure manual e I. Baltimore, MD: Johns Hopkins; 1977. Derogatis LR, Rickles K, Rock AF. SCL-90 and the MMPI: a step in the validation of a new self-report scale. Br J Psychiatry 1976;128:280e9. Elhai JD, Carvalho LF, Miguel FK, Palmieri PA, Primi R, Frueh B. Testing whether posttraumatic stress disorder and major depressive disorder are similar or unique constructs. J Anxiety Disord 2011;25:404e41. Erickson DJ, Wolfe J, King DW, King LA, Sharkansky EJ. Post-traumatic stress disorder and depression symptomatology in a sample of Gulf War veterans: a prospective analysis. J Consult Clin Psychol 2001;69:41e9.
69
Franklin CL, Zimmerman M. Post-traumatic stress disorder and major depressive disorder: Investigating the role of overlapping symptoms in diagnostic comorbidity. J Nerv Ment Dis 2001;189:548e51. Gaffney M. Factor Analysis of treatment response in post-traumatic stress disorder. J Trauma Stress 2003;16:77e80. Ginzburg K, Ein-Dor T, Solomon Z. Comorbidity of post-traumatic stress disorder, anxiety and depression: a 20-year longitudinal study of war veterans. J Affect Disord 2010;23:249e57. Grieger TA, Cozza SJ, Ursano RJ, Hoge C, Martinez PE, Engel CC, et al. Posttraumatic stress disorder and depression in battle-injured soldiers. Am J Psychiatry 2006;163:1777e83. Herman JL. Complex PTSD: a syndrome in survivors of prolonged and repeated trauma. J Trauma Stress 1992;5:377e91. Keane TM, Kaloupek DG. Comorbidity psychiatric disorders in PTSD: implications for research. Ann New York Acad Sci 1997;821:24e34. Kessler RC, Sonnega A, Bromet E, Hughes M, Nelson C. Post-traumatic stress disorder in the National Comorbidity Survey. Arch Gen Psychiatry 1995;52:1048e 60. King DW, King LA, McArdle JJ, Shalev AY, Doron-Lamarca S. Sequential temporal dependencies in associations between symptoms of depression and posttraumatic stress disorder: an application of bivariate latent difference score structural equation modeling. Multivar Behav Res 2009;44:437e64. McNally RJ. Progress and controversy in the study of post-traumatic stress disorder. Annu Rev Psychol 2003;54:229e52. Muthén LK, Muthén BO. Mplus user’s guide. 6th ed. Los Angeles, CA: Muthén & Muthén; 1998e2010. O’Toole BI, Marshall RP, Schureck RJ, Dobson M. Post-traumatic stress disorder and comorbidity in Australian Vietnam veterans: risk factors, chronicity and combat. Aust New Zeal J Psychiatry 1998;32:32e42. Oquendo MA, Echavarria G, Galfalvy HC, Grunebaum MF, Burke A, Barrera A. Lower cortisol levels in depressed patients with comorbid post-traumatic stress disorder. Neuropsychopharmacology 2003;28:591e8. O’Donnell M, Creamer MC, Pattison P. Post-traumatic stress disorder and depression following trauma: understanding comorbidity. Am J Psychiatry 2004;161: 1390e6. Rubin DB. Multiple imputation for nonresponse in surveys. New York: John Wiley & Sons, Inc; 1987. Schafer JL. Analysis of incomplete multivariate data. New York: Chapman and Hall; 1997. Selye H. Annual report of stress. New York: McGraw-Hill; 1956. Shalev AY, Freedman SA, Peri T, Brandes D, Sahar T, Orr SP, et al. Prospective study of post-traumatic stress disorder and depression following trauma. Am J Psychiatry 1998;155:630e7. Sher L. The concept of post-traumatic mood disorder. Med Hypotheses 2005;65: 205e10. Solomon Z, Dekel R. Posttraumatic stress disorder among Israeli ex-prisoners of war18 and 30 years after release. J Clin Psychiatry 2005;66:1031e7. Southwick SM, Yehuda R, Giller EL. Characterization of depression in war-related post-traumatic stress disorder. Am J Psychiatry 1991;148:179e83. Spitzer RL, First MB, Wakefield JC. Saving PTSD from itself in DSM-IV. J Anxiety Disord 2007;21:233e41. Yehuda R. Current status of cortisol findings in post-traumatic stress disorder. Psychiatric Clin North Am 2002;25:341e68. Zlotnick C, Warshaw M, Shea MT, Keller MB. Trauma and chronic depression among patients with anxiety disorders. J Consult Clin Psychol 1997;65:333e6.
Contents lists available at ScienceDirect
Journal of Psychiatric Research journal homepage: www.elsevier.com/locate/psychires
Posttraumatic stress disorder and depressive symptoms: Joined or independent sequelae of trauma? Sharon Dekel a, b, *, Zahava Solomon c, Danny Horesh d, e, Tsachi Ein-Dor f a
Department of Psychiatry, Harvard Medical School, Boston, MA, USA Department of Psychiatry, Massachusetts General Hospital, Charlestown, MA, USA c Bob Shapell School of Social Work, Tel Aviv University, Tel Aviv, Israel d Department of Psychology, Bar Ilan University, Ramat Gan, Israel e Department of Psychiatry, New York University, NY, USA f School of Psychology, Interdisciplinary Center Herzliya, Herzliya, Israel b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 1 May 2013 Received in revised form 6 March 2014 Accepted 7 March 2014
Objective: The nature of co-morbidity between posttraumatic stress disorder (PTSD) and depression has been the subject of much controversy. This study addresses this issue by investigating associations between probable PTSD and depressive symptoms in a prospective, longitudinal sample of combat veterans. Method: Symptoms of PTSD and depression were assessed at 3 points of time (i.e., 1991, 2003, 2008) over a period of 17 years utilizing the PTSD Inventory and the SCL-90 (Derogatis, 1977). Two groups of combat veterans, 275 former prisoners of war (ex-POWs) and 219 matched combatants (controls), were assessed. Data were analyzed using descriptive statistics, latent variable modeling, and confirmatory factor analysis. Results: A series of c2 tests revealed that the prevalence proportions of depressive symptoms and probable PTSD were higher among ex-POWs compared to controls at all time points. The prevalence of depressive symptoms was higher than the prevalence of PTSD symptoms in both groups at the each of the times. Latent Trajectories Modeling (LTM) indicated that while ex-POWs’ PTSD symptom severity increased over time, the severity of symptoms remained stable among controls. Parallel Process Latent Growth Modeling (PLGM) revealed a positive bi-directional relationship whereby PTSD symptoms mediated the affect of captivity on depressive symptoms and depressive symptoms mediated the affect of captivity on PTSD symptoms over time. Utilizing Confirmatory Factor Analysis (CFA), a single factor model emerged for depressive and PTSD symptoms. Conclusion: The findings suggest that while depression and PTSD seem to be different long-term manifestations of traumatic stress, accounted for in part by the severity of the trauma, they both may be parts of a common general traumatic stress construct. Clinical and theoretical implications of these findings are discussed. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: PTSD Depression POWs Captivity Combat trauma Longitudinal study
Posttraumatic stress disorder (PTSD) is recognized as a distinguished psychiatric disorder precipitated by exposure to a traumatic event (DSM-V, American Psychiatric Association, 2013). However, over the years, the distinctiveness of PTSD has generated much controversy (McNally, 2003; Spitzer et al., 2007). Studies consistently reveal that PTSD is often accompanied by other disorders with the most notable being the association with depression (e.g., Campbell et al., 2007). The rates of PTSD in individuals
* Corresponding author. PTSD Research Laboratory, Massachusetts General Hospital-East, 120 Second Avenue Charlestown, MA 02129, USA. E-mail addresses: [email protected], [email protected] (S. Dekel). http://dx.doi.org/10.1016/j.jpsychires.2014.03.003 0022-3956/Ó 2014 Elsevier Ltd. All rights reserved.
reporting a lifetime history of depression range between 37% and 48% (Breslau et al., 1997; Kessler et al., 1995); and among clinical samples, as much as 68% of individuals with PTSD meet diagnostic criteria for depression at some point in time (see Keane and Kaloupek, 1997; for a review). Co-occurrence of PTSD and depression has a negative impact on prognosis including hospitalization, disability, and suicide attempts (Sher, 2005). Despite the clinical significance and the body of existing research, the mechanisms underlying the association between PTSD and depression are not fully understood. According to one school of thought, PTSD and depression embody a single, “general traumatic stress” construct (O’Donnell et al., 2004; Sher, 2005). Supporting evidence suggests that one
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
disorder may be a derivate of the other. Suffering from prior depression was fond to increase the risk of developing PTSD (Breslau et al., 1997; O’Teelo et al., 1998), while other studies documented that PTSD renders the individuals vulnerable to first onset of depression (Breslau et al., 2000; Ginzburg et al., 2010; Kessler et al., 1995). Furthermore, similar risk factors (e.g., severity of trauma, negative event appraisal) have been documented to set in motion both depression and PTSD (O’Donnell et al., 2004). A shared genetic and biological basis for PTSD and depression has been suggested (Gaffney, 2003; Davidson et al., 1998). A contrary view deems PTSD and depression independent trauma sequelae (e.g., Shalev et al., 1998). Relatively high rates of depression in the short-term following trauma (i.e., 19% among survivors of various traumas) were reported among individuals with no PTSD diagnosis (Shalev et al., 1998). Similarly, there is evidence indicating that patients with a history of trauma report more episodes of depression than those without such history (Zlotnick et al., 1997). These findings may be taken to suggest that depression is first and foremost a derivate of exposure to trauma, rather than of PTSD per-se. Furthermore, some biological studies support the notion of two independent trauma sequelae with evidence that the two disorders entail opposing dysregulations in the body’s stress response (Oquendo et al., 2003; Yehuda, 2002). As can be seen, the attempt to understand the observed association between PTSD and depression has so far yielded inconclusive results. This may be attributed, in large part, to methodological shortcomings of previous studies. Most notably, the reliance on cross-sectional designs using retrospective reports (e.g., Kessler et al., 1995; Zlotnick et al., 1997), which are subject to memory bias in reporting the course of symptoms over time (Dekel and Bonanno, 2013) rather than using longitudinal assessments (e.g., Grieger et al., 2006); and the study of homogenous trauma samples which ignores the probable affect of event severity on outcomes. Grieger et al. (2006) showed that in the very short-term following exposure PTSD and depressive symptoms tend to escalate over time, and yet, to the best of our knowledge, no study looked into the interplay between the two disorders longitudinally following trauma in the very long-term and with respect to the severity of the trauma. This prospective study examined Israeli ex-POWs and comparable combat veterans of the 1973 Yom Kippur War who had not been incarcerated at three time points: 1991, 2003, and 2008. We set to evaluate the prevalence and course of symptoms of PTSD and depression, as well as the long-term interrelationships between the symptoms. Our research aims were (1) To assess the prevalence of symptoms of PTSD and depression at three points in time and the changes in prevalence over time. (2) To assess the interrelationship of symptoms of PTSD and depression, at each assessment and across time and (3) To assess the bi-directional associations between symptoms of PTSD and depression across time. 1. Method This study uses data from a longitudinal study on the psychological effects of war (Dekel et al., 2012). A cohort of Israeli male veterans who participated in the 1973 Yom Kippur War were followed over 17 years with assessment at three time points: 1991 (T1), 2003 (T2), and 2008 (T3). Following approval from both the Israel Defense Forces (IDF) and Tel Aviv University review boards, lists of potential participants from the IDF computerized data bank were composed. We then phoned those participants and, after explaining the purpose of the study, asked them to take part in the assessments. The questionnaire packet was administered in their homes or in another location of their choice. Informed consent was obtained from all participants.
65
A target population of 520 representative combat veterans was comprised of two groups: individuals who were captured during the war and consequently subjected to torture, harassment and humiliations (ex-POWs: n ¼ 240); and veterans who participated in the same war but were not taken captive (combats: n ¼ 280). Both groups were exposed to battlefield stressors entailing active fighting and encounters with death. Groups were matched on military background and socio-demographic status with no differences in combat exposure and pre- and post war negative life events (see Dekel et al., 2012; for details on sampling method). The sample included 164 ex-POWs who participated in the first assessment, and 144 in the second (10 could not be located/refused, 4 had died, and 6 could not participate due to mental deterioration). At the third assessment, the original group of 240 ex-POWs was contacted again. Of these 171 participated (11 could not be located, 20 had died, 11 were living abroad, 21 refused to participate and 6 could not participate due to their mental status). 185 combatants participated in T1, 143 in T2 (41 could not be located and 1 had died), and 118 in T3 (20 could not be located/refused and 5 had died). In the GLM-based analyses, missing data was handled using multiple imputation (Rubin, 1987; Schafer, 1997). In the SEM-based analyses missing data was handled with the case-wise maximum likelihood estimation. For all participants, the mean age was 53.4 (SD ¼ 4.4); years of schooling were 14.02 (SD ¼ 3.41); the majority were secular (67%) with an average income (62%) in T2. No significant differences were found between those who participated in the follow-up assessments with regard to initial PTSD level, military rank, age, and education. No difference was found between the study groups in post-war level of exposure to potentially traumatic events (see Dekel et al., 2013a; for details about the measurement). 1.1. Measures PTSD symptoms were assessed using the PTSD Inventory (Dekel et al., 2012), a self-report measure corresponding to the 17 PTSD symptoms listed in the DSM-III-R (APA, 1987) which was the standard of practice when the study commenced. Participants were asked to indicate the frequency they had each symptom in the past month anchored on their 1973 war experiences, on a 4-point scale ranging from 1 (not at all) to 4 (I usually did), with 3 (often) indicating cutoff for symptom endorsement. For each participant, we calculated the number of symptoms (intrusion, avoidance and hyper-arousal) that he endorsed in each time point. In order to conform to the DSM-IV symptom clusters, we classified individuals as endorsing “Probable PTSD” by having at least 1 intrusion, 3 avoidance and 2 hyper-arousal symptoms (“physiological reactivity to resembling events” symptom of the DSM-III-R was categorized under the intrusion cluster). The inventory has proven psychometric properties in terms of high test-retest reliability concurrent and convergent validity. In the present study, reliability values for PTSD symptoms and symptom clusters (intrusion, avoidance and hyper-arousal) were high (Cronbach’s a ranged between .78 and .90, .88 and .95, and .87 and .96, respectively, in T1, T2, and T3). Depressive symptoms were assessed using the depression subscale of the Symptom Checklist-90 (SCL-90, Derogatis, 1977). Participants were asked to indicate how frequently they experienced each symptom during the last 2 weeks on a 5-point distress scale. Based on norms for psychiatric outpatients (Derogatis, 1977) scores equal to or above .73 were considered as an indication for endorsement of depressive symptoms (Dekel et al., 2013a, 2013b, 2013c). For each participant, we calculated the average frequency of experiencing depressive symptoms in each of the time points. The SCL-90 has high concurrent validity and the specific subscales display high empirical agreement across various samples
66
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
Table 1 Rates of PTSD and depression, test statistics and relative risk for examining the differences between ex-POWs and controls. Time
Variable PTSD
Depression
Rate: ex-POWs [controls] T1(1991;
N ¼ 325)
T2(2003;
N ¼ 227)
T3
(2008; N ¼ 287)
8.7% (N ¼ 14) [3% (N ¼ 5)] 29.9% (N ¼ 35) [1.9% (N ¼ 2)] 34.7% (N ¼ 58) [2.5% (N ¼ 3)]
X2(1)
Relative risk [95% CI]
4.77*
2.87 [1.06, 7.78]
p .038
31.57***
15.86 [3.91, 64.33]
<.0001
42.59***
13.66 [4.38, 42.56]
<.0001
Rate: ex-POWs [controls]
X2(1)
Relative risk [95% CI]
26.3% (N ¼ 42) [10.9% (N ¼ 18)] 69.9% (N ¼ 86) [19.2% (N ¼ 20)] 67.1% (N ¼ 114) [18.8% (N ¼ 22)]
12.7***
2.41 [1.42, 4.00]
.0007
58.17***
3.64 [2.41, 5.48]
<.0001
64.73***
3.57 [2.41, 5.27]
<.0001
p
Note: *p < .05, ***p < .001.
(Derogatis et al., 1976). In this study, reliability values were high in all assessments (Cronbach’s a was .92, .94, .90, respectively, in T1, T2, and T3). 1.2. Data analysis We examined the link between PTSD and depressive symptoms from various complimentary perspectives. First, we examined whether ex-POWs and controls differ in the prevalence of probable PTSD and depressive symptoms in each of the time points using c2 tests of independence and relative risk analysis. Then, we examined whether and to what extent the prevalence of depressive symptoms and Probable PTSD had changed over time using McNemar tests. In these tests, we also examined the contribution of captivity (ex-POWs vs. controls) in moderating these effects. Second, we examined whether ex-POWs and controls differ in the prevalence of co-occurring PTSD and depressive symptoms in each of the time points using c2 tests of independence (including odd ratios). Third, we examined whether PTSD and depressive symptoms are two distinct psychopathological constructs or are they one construct of internalizing-related psychopathology. To this end, we employed multi-group Confirmatory Factor Analysis (CFA). Fourth, we examined the contribution of captivity (exPOWs vs. controls) to the differences in the 17-year trajectories of PTSD and depressive symptoms using Latent Trajectories Modeling. Finally, we employed parallel process latent growth modeling to examine the contribution of captivity (ex-POWs vs. controls) to the bidirectional associations between PTSD and depressive symptoms over time. We also examined the bidirectional associations between PTSD and depressive symptoms over time regardless of captivity. 2. Results 2.1. Prevalence of probable PTSD and depressive symptoms Table 1 presents the prevalence of probable PTSD and depressive symptoms, test statistics and relative risk values. The analyses revealed that the prevalence proportions of depressive
symptoms and probable PTSD were higher among the ex-POWs than controls in T1, T2 and T3. Also, the prevalence of depressive symptoms was higher than the prevalence of PTSD symptoms in both groups at the each of the time points. However, importantly, as can be seen, the relative risk of ex-POWs to suffer from PTSD symptoms was higher than depressive symptoms at T2 and T3. 2.2. Changes in prevalence of probable PTSD and depressive symptoms over time In this section, we examined separately for ex-POWs and controls the changes in the prevalence of PTSD and depressive symptoms between T1 and T2, and between T2 and T3. McNemar tests revealed that the controls’ PTSD symptoms prevalence was constant over time (ps ¼ 1). Conversely, the analyses revealed that exPOWs’ PTSD prevalence significantly increased between T1 and T2 (p < .0001); yet, it remained constant between T2 and T3 (p ¼ .13). Also, McNemar tests revealed that for both the controls and exPOWs depressive symptoms prevalence significantly increased between T1 and T2 (p < .05 for controls and p < .0001 for exPOWs); yet, it remained constant between T2 and T3 (p ¼ 1 for controls and p ¼ .84 for ex-POWs). 2.3. Co-occurrence of probable PTSD and depressive symptoms over time In this section, we examined separately for ex-POWs and controls, the prevalence of the co-occurrence of probable PTSD and depressive symptoms in T1, T2, and T3. Frequency of co-occurrence and test statistics are presented in Table 2. The analyses revealed significant differences in ex-POWs’ and controls’ prevalence of PTSD-depression co-occurrence at all time points. As can be seen, ex-POWs’ had higher co-occurrence rates. In addition, a series of McNemar tests revealed that in both groups the frequency of co-occurring depressive and PTSD- symptoms significantly increased from T1 to T2 (ps < .05) but remained unchanged between T2 and T3 (ps > .35).
Table 2 Rates of co-morbidity between PTSD and depression among ex-POWs and controls. ex-POWs
T1 T2 T3
Controls
95% CI of Exp(B)
Co-morbidity
Either depression or PTSD
Neither
Co-morbidity
Either depression or PTSD
Neither
X2(2)
8.8% (14) 26.5% (31) 30.9% (50)
17% (27) 47% (55) 38.9% (63)
74.2% (118) 26.5% (31) 30.25% (49)
3% (5) 1% (1) 2.6% (3)
7.9% (13) 18.3% (19) 16.2% (19)
89.1% (147) 80.8% (84) 81.2% (95)
12.22* 69.54*** 74.67***
Note: Exp(B) ¼ odd ratio. Ns are presented in parentheses.
3.08 [1.09, 8.78] 37.13 [4.97, 277.61] 16.96 [5.14, 55.98]
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
67
Table 3 CFA loadings by study group. ex-POWs
T1 T2 T3
Controls
Avoidance
Intrusive
Hyper-arousal
Depression
Avoidance
Intrusive
Hyper-arousal
Depression
.78 .75 .77
.89 .67 .74
.89 .86 .86
.71 .75 .70
.64 .67 .78
.62 .62 .58
.71 .72 .83
.74 .81 .78
Note: scaled c2 (Yuan-Bentler; 120) ¼ 105.43, p ¼ .83, CFI ¼ 1, NNFI ¼ 1, RMSEA ¼ 0.
2.4. Interrelationships between PTSD and depressive symptoms over time To examine whether PTSD and depression are two different constructs or if they represent one general construct, we ran a Confirmatory Factor Analysis (CFA) with three latent variables: one for each of the three waves of measurement. We estimated the appropriateness of the model using EQS 6.1 SEM software (Bentler and Wu, 1995). The model’s fit was assessed by the comparative fit index (CFI), Bentler-Bonett non-normed fit index (NNFI) and the root-mean-square error of approximation (RMSEA). A model is judged as reasonably fitting the data when CFI and NNFI are higher than .95 and the RMSEA is lower than .05 (Bollen and Curran, 2006). Missing data was handled with the case-wise maximum likelihood estimation. As seen in Table 3, the analysis revealed that the multi-group CFA had excellent fit to the observed data, indicating that PTSD avoidance, intrusive, and hyper-arousal clusters as well as depressive symptom level were significantly loaded to one construct at all time points.
2.5. Trajectories of PTSD and depressive symptoms over time To examine changes in PTSD and depressive symptoms over time, we conducted a series of Latent Trajectories Modeling (LTM: see Bollen and Curran, 2006). Two latent factors were estimated: the first, defines the initial levels of PTSD and depression (i.e., intercept), and the second, explores the constant of the trajectory of change in PTSD and depression over time (i.e., linear; time was coded as 0, 12 and 17). In addition, we assessed whether the LTMs were different for ex-POWs and controls by using a multi-group SEM procedure. LTMs results are presented in Table 4. The analysis revealed that ex-POWs’ PTSD level annually increased since the initial assessment in 1991, and that controls’ PTSD level remained unchanged. The LTM estimating the ex-POWs’ depression trajectory showed poor fit to the observed data, and thus, ex-POWs’ depression level did not change linearly over time. To account for this non-linear change, we estimated an additional LTM in which we fixed only the first (coded 0) and the last (coded Table 4 LTMs results for examining changes in PTSD and depression over time. Change (value of the linear slope) ex-POWs PTSD
.13***
Depression
N/A
Scaled c2
CFI
NNFI
RMSEA
104.44** [51.16*]
.98 [1.0] .93 [1.0]
.97 [1.0] .70 [1.0]
.05 [.0] .13 [.0]
Controls .02 01**
37.71** [.19]
Note: controls’ fit indices are in brackets. CFI ¼ comparative fit index; NNFI ¼ BentlereBonett non-normed fit index; RMSEA ¼ root-mean-square error of approximation.
17) time points, while freely estimating the second time point. The model has an excellent fit to the observed data [c2 (4) ¼ 4.68, p ¼ .32, CFI ¼ 1, TLI ¼ .99, RMSEA ¼ .04.], and indicated that exPOWs’ depression level increased significantly between T1 and T2 and then remained unchanged between T2 and T3. In comparison, controls’ depression level annually increased since 1991. 2.6. Bidirectional associations between PTSD and depressive symptoms over time To examine the bidirectional association between PTSD and depression from T1, through T2 to T3, we employed parallel process latent growth modeling (Cheong et al., 2003), which is extension of LTM. Because PTSD consists of three clusters (intrusion, avoidance, and hyper-arousal), we used latent variables in SEM environment to represent the PTSD underlying phenomenon. The model was tested by MPlus 6.1 (Muthén & Muthén, 1998e2010). Specifically, we examined whether captivity (ex-POWs ¼ 1 vs. controls ¼ 0) predicts a change in PTSD, and whether this change, in turn, predicts a change in depression (i.e., whether the change in PTSD mediates the link between captivity and later depression). In this model, we also examined the opposite process by which captivity (ex-POWs ¼ 1 vs. controls ¼ 0) predicts a change in depression, which in turn predicts a change in PTSD. The model had an excellent fit for the observed data, c2 (8) ¼ 10.68, p ¼ .22, CFI ¼ 1, TLI ¼ .99, RMSEA ¼ .04. The analysis revealed that ex-POWs, on average, had more PTSD symptoms severity (b ¼ .76, b ¼ .16, p < .05) and depression (b ¼ .14, b ¼ .16, p < .05) in T1 than controls (i.e., greater intercepts). Ex-POWs, on average, also suffered from a greater increase in PTSD over time than controls (b ¼ .39, b ¼ .76, p < .0001) but not in depression (at least not directly; b ¼ .01, b ¼ .14, p ¼ .19). Captivity did influence the trajectory of depression via an effect on PTSD: Bias-corrected bootstrap analyses indicated that the increase in PTSD symptoms severity over time mediated the effect of captivity on depression, b ¼ .06, p < .0001, 99% CI [.03, .09], such that captivity was linked with greater increase in PTSD symptoms severity over time, which in turn was associated with a greater increase in depression over time. The model revealed an additional mediation process regarding the bidirectional association between PTSD and depression. Bias-corrected bootstrap analyses indicated that captivity increased the level of depression in T1 (an effect of captivity on the intercept of depression), which in turn was related with a greater increase in PTSD symptoms severity over time, b ¼ .20, p ¼ .034, 99% CI [.005, .08]. Next, we examined the parallel processes of depressive and PTSD symptoms over time without recognizing captivity. The model had an excellent fit for the observed data, c2 (5) ¼ 8.64, p ¼ .12, CFI ¼ 1, TLI ¼ .99, RMSEA ¼ .06, and revealed that higher level of depression in T1 was associated with a greater increase in PTSD symptoms severity over time, b ¼ .33, p < .001. A greater increase in depression over time was also associated with a greater increase in PTSD symptoms severity over time, b ¼ 3.19, p < .001. Conversely, PTSD symptoms severity in T1 was not associated with a change in depression over time, b ¼ .001, p ¼ .37. Rather, a greater
68
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
increase in PTSD over time was also associated with a greater increase in depression over time, b ¼ .13, p < .001. 3. Discussion This study set out to examine the long-term prevalence, course, and interrelations of symptoms of PTSD and depression among exPOWs and comparable controls. The prevalence of depressive symptoms and probable PTSD, as well as their co-occurrence, was higher among ex-POWs than controls at all three waves of measurement. Results also indicate that the prevalence of depressive symptoms was higher than that of probable PTSD among both study groups, at all three assessments. However, while the prevalence of depressive symptoms increased over time in both study groups, PTSD symptoms increased only among ex-POWs, but not among controls. Likewise, captivity contributed to the bidirectional association between symptoms of PTSD and depression over time. Importantly, in accord with the notion of a unified traumatic stress construct, our findings show that symptoms of depression and PTSD were inter-related across times. The finding that ex-POWs endorsed higher depressive symptom severity and probable PTSD is consistent with previous research, as well as earlier investigations of this sample conducted by our group (e.g., Dekel et al., 2013a, 2013b, 2013c). Several explanations may exist. First, captivity is a particularly severe traumatic experience, which entails hardships such as torture and humiliation that are inter-personal, deliberately inflicted and repeatedly experienced. Second, in this sample, the ex-POWs were exposed to a ‘double trauma’ pertaining to both captivity and combat exposure. Our finding that ex-POWs also reported more co-occurrence of depressive and PTSD symptoms suggests that their pathology entails a more severe type of posttraumatic disorder (Dekel et al., 2013a, 2013b, 2013c) or complex PTSD (Herman, 1992). While depression has been conceived as a derivate of PTSD (Kessler et al., 1995), our findings suggest that in relation to sever trauma, such as the experience of captivity, symptoms of depression may subsequently worsen symptoms of PTSD. This is consistent with some prospective short-term assessments showing that depression fuels PTSD (Erikson et al., 2001; King et al., 2009). Dysphoria symptoms, entailing a sense of helplessness and loss of mastery, may prevent seeking social support, self-disclosure, and other self-efficacious behaviors that buffer against the development of symptoms of PTSD (King et al., 2009). Moreover, sleep and concentration impairments seen in depression may trigger a hypervigilance response (King et al., 2009). The contribution of our findings is in underscoring that the severity of event affects the nature of the bi-directional relationships between PTSD and depressive symptoms. The findings suggest that severe traumatic exposure not only yields higher levels of psychopathology at a given time point, but also entails enduring long-term associations between different psychopathological conditions. Interestingly, our findings show that the prevalence of depressive symptoms increased among both study groups, while that of probable PTSD increased only among ex-POWs. While PTSD is the most common psychiatric sequela of trauma in the initial period following exposure (e.g., Shalev et al., 1998), a different pattern appears evident in the long-term, with depressive symptoms emerging as the dominant clinical response. This may be explained by Seyle’s (1956) conceptualization of a general adaptation syndrome in response to an external stressor. First, the individual exhibits an alarmed and the fight-or flight response. As the stressor persists, the individual enters a resistance state and attempts to deploy means of coping. In the final state, exhaustion, resources are depleted and normal functioning is impaired. When this phase is
prolonged, it may manifest in long-term pathogenic outcomes and the development of depressive symptoms. The findings that depressive symptoms may worsen over time among individuals exposed to relatively moderate traumatic events, similarly to survivors of extreme and prolonged trauma, suggests that unlike PTSD (e.g., Solomon and Dekel, 2005), the enduring nature of depressive symptoms is possibly attributed primarily to post-event factors such as level of social support rather than to the level of exposure and other peritraumatic factors. Finally, our findings revealed a single underlying dimension for depressive and PTSD symptoms in all three waves of measurement. While we cannot fully negate an illusionary unified post-trauma construct merely due to symptom overlap (Franklin and Zimmerman, 2001; Southwick et al., 1991); in a recent study, a single factor emerged even after removing shared symptoms and symptoms of PTSD that are depression-related (Elhai et al., 2011). This may suggest that PTSD and depression share a common diathesis, possibly indicative of a single general traumatic response (Breslau et al., 2000; O’Donnell et al., 2004). This study has several methodological limitations. First, it does not allow inferences regarding causal relations between depressive and PTSD symptoms following trauma. Second, in spite of the strength of longitudinal designs, there are inherent limitations. Not all participants took part in all assessments, although the samples did not differ with respect to socio-demographic variables. Since we focused on the long-term phase, we may have overlooked fine fluctuations in depressive and PTSD symptoms given short interval assessments in the immediate phase and earlier exposure to trauma. Finally, our use of self-report measures, although well validated, rather than having obtained clinical diagnosis, leaves open the possibility of overestimating symptom prevalence. Future studies may benefit from the inclusion of clinical assessments, as well as from the examination of additional factors associated with co-occurring PTSD-depression, such as psychiatric history, personality traits, and cognitive variables. Nevertheless, the present study makes an important contribution to the literature regarding the co-occurrence of PTSDdepression. The repeated evidence in support of a tight interplay between symptoms of PTSD and depression assessed in three time points following the war in a veterans sample may indicate that PTSD symptoms are not the only direct outcome of trauma exposure. Instead, survivors of trauma may experience a general PTSDdepression response to include symptoms of re-experiencing, avoidance, and arousal along with negative alteration in mood, according, in part, with recent criteria revisions (DSM-V) for PTSD diagnosis. In practical terms, our findings suggest that mental health professionals may benefit from a close monitoring of depressive symptoms, as those may be a clinical marker of ensuing symptoms of PTSD. Role of the funding source Part of this work has been funded by the NARSAD Young Investigator Grant sponsored by Dr. Dylan Tauber which was awarded to Sharon Dekel. Contributors Dr. Sharon Dekel initiated and generated the study paradigm with respect to the sample, aims, measurements, and design, and prepared the manuscript. Prof. Zahava Solomon supervised the research project. Dr. Tsachi Ein-Dor conducted the statistical analysis. Dr. Danny Horesh contributed to the writing of the manuscript.
S. Dekel et al. / Journal of Psychiatric Research 54 (2014) 64e69
Conflict of interest The authors of the study have no conflict of interest. Acknowledgment We thank Dr. Talya Greene who assisted in the grammatical editing of the manuscript. References American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 3rd ed., revised. Washington, DC: American Psychiatric Press; 1987. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Washington, DC: American Psychiatric Press; 2013. Bentler PM, Wu EJC. EQS for Windows user’s guide. Encino, CA: Multivariate Software, Inc; 1995. Bollen KA, Curran PJ. Latent curve models. A structural equation perspective. Hoboken, NJ: Wiley; 2006. Breslau N, Davis CG, Peterson EL, Schultz LR. Psychiatric sequelae of post-traumatic stress disorder in women. Arch Gen Psychiatry 1997;54:81e7. Breslau N, Davis GC, Peterson EL, Schultz LR. A second look at comorbidity in victims of trauma: the post-traumatic stress disorder-major depression connection. Biol Psychiatry 2000;48:902e9. Campbell DG, Felker BL, Liu CF, Yano E, Kirchner JE, Chan D, et al. Prevalence of depressionePTSD comorbidity: Implications for clinical practice guidelines and primary care-based interventions. J Gen Intern Med 2007;22:711e8. Cheong J, MacKinnon DP, Khoo ST. Investigation of mediational processes using parallel process latent growth curve modeling. Struct Equ Model 2003;10:238e62. Davidson JR, Tupler LA, Wilson WH, Connor KM. A family study of chronic posttraumatic stress disorder following rape trauma. J Psychiatric Res 1998;32: 301e9. Dekel S, Bonanno G. Changes in trauma memory and patterns of posttraumatic stress. Psychol Trauma Theory Res Pract Policy 2013;5:26e34. Dekel S, Ein-Dor T, Solomon Z. Posttraumatic growth and posttraumatic distress: a longitudinal study. Psychol Trauma Theory Res Pract Policy 2012;4:94e101. Dekel S, Mandl C, Solomon Z. Is the holocaust implicated in PTG in second generation holocaust survivors? A prospective study. J Trauma Stress 2013a;26:1e4. Dekel S, Peleg T, Solomon Z. The relationship of PTSD to negative cognitions: a 17year longitudinal study. Psychiatry Interpers Biol Process 2013b;76:241e55. Dekel S, Solomon Z, Rozenstreich E. Secondary salutogenic effects in veterans whose parents were holocaust survivor? J Psychiatric Res 2013c;47:266e71. Derogatis LR. SCL-90: Administration, scoring, and procedure manual e I. Baltimore, MD: Johns Hopkins; 1977. Derogatis LR, Rickles K, Rock AF. SCL-90 and the MMPI: a step in the validation of a new self-report scale. Br J Psychiatry 1976;128:280e9. Elhai JD, Carvalho LF, Miguel FK, Palmieri PA, Primi R, Frueh B. Testing whether posttraumatic stress disorder and major depressive disorder are similar or unique constructs. J Anxiety Disord 2011;25:404e41. Erickson DJ, Wolfe J, King DW, King LA, Sharkansky EJ. Post-traumatic stress disorder and depression symptomatology in a sample of Gulf War veterans: a prospective analysis. J Consult Clin Psychol 2001;69:41e9.
69
Franklin CL, Zimmerman M. Post-traumatic stress disorder and major depressive disorder: Investigating the role of overlapping symptoms in diagnostic comorbidity. J Nerv Ment Dis 2001;189:548e51. Gaffney M. Factor Analysis of treatment response in post-traumatic stress disorder. J Trauma Stress 2003;16:77e80. Ginzburg K, Ein-Dor T, Solomon Z. Comorbidity of post-traumatic stress disorder, anxiety and depression: a 20-year longitudinal study of war veterans. J Affect Disord 2010;23:249e57. Grieger TA, Cozza SJ, Ursano RJ, Hoge C, Martinez PE, Engel CC, et al. Posttraumatic stress disorder and depression in battle-injured soldiers. Am J Psychiatry 2006;163:1777e83. Herman JL. Complex PTSD: a syndrome in survivors of prolonged and repeated trauma. J Trauma Stress 1992;5:377e91. Keane TM, Kaloupek DG. Comorbidity psychiatric disorders in PTSD: implications for research. Ann New York Acad Sci 1997;821:24e34. Kessler RC, Sonnega A, Bromet E, Hughes M, Nelson C. Post-traumatic stress disorder in the National Comorbidity Survey. Arch Gen Psychiatry 1995;52:1048e 60. King DW, King LA, McArdle JJ, Shalev AY, Doron-Lamarca S. Sequential temporal dependencies in associations between symptoms of depression and posttraumatic stress disorder: an application of bivariate latent difference score structural equation modeling. Multivar Behav Res 2009;44:437e64. McNally RJ. Progress and controversy in the study of post-traumatic stress disorder. Annu Rev Psychol 2003;54:229e52. Muthén LK, Muthén BO. Mplus user’s guide. 6th ed. Los Angeles, CA: Muthén & Muthén; 1998e2010. O’Toole BI, Marshall RP, Schureck RJ, Dobson M. Post-traumatic stress disorder and comorbidity in Australian Vietnam veterans: risk factors, chronicity and combat. Aust New Zeal J Psychiatry 1998;32:32e42. Oquendo MA, Echavarria G, Galfalvy HC, Grunebaum MF, Burke A, Barrera A. Lower cortisol levels in depressed patients with comorbid post-traumatic stress disorder. Neuropsychopharmacology 2003;28:591e8. O’Donnell M, Creamer MC, Pattison P. Post-traumatic stress disorder and depression following trauma: understanding comorbidity. Am J Psychiatry 2004;161: 1390e6. Rubin DB. Multiple imputation for nonresponse in surveys. New York: John Wiley & Sons, Inc; 1987. Schafer JL. Analysis of incomplete multivariate data. New York: Chapman and Hall; 1997. Selye H. Annual report of stress. New York: McGraw-Hill; 1956. Shalev AY, Freedman SA, Peri T, Brandes D, Sahar T, Orr SP, et al. Prospective study of post-traumatic stress disorder and depression following trauma. Am J Psychiatry 1998;155:630e7. Sher L. The concept of post-traumatic mood disorder. Med Hypotheses 2005;65: 205e10. Solomon Z, Dekel R. Posttraumatic stress disorder among Israeli ex-prisoners of war18 and 30 years after release. J Clin Psychiatry 2005;66:1031e7. Southwick SM, Yehuda R, Giller EL. Characterization of depression in war-related post-traumatic stress disorder. Am J Psychiatry 1991;148:179e83. Spitzer RL, First MB, Wakefield JC. Saving PTSD from itself in DSM-IV. J Anxiety Disord 2007;21:233e41. Yehuda R. Current status of cortisol findings in post-traumatic stress disorder. Psychiatric Clin North Am 2002;25:341e68. Zlotnick C, Warshaw M, Shea MT, Keller MB. Trauma and chronic depression among patients with anxiety disorders. J Consult Clin Psychol 1997;65:333e6.