Journal of Affective Disorders 123 (2010) 249–257

Contents lists available at ScienceDirect

Journal of Affective Disorders

j ourna l homepage: www.e lsev ie r.com/ locate / j ad

Research report

Comorbidity of posttraumatic stress disorder, anxiety and depression:A 20-year longitudinal study of war veterans

Karni Ginzburg a,⁎, Tsachi Ein-Dor b, Zahava Solomon a

a Bob Shapell School of Social Work, Tel Aviv University, Tel Aviv 69978, Israelb The New School of Psychology, Interdisciplinary Center (IDC) Herzliya, Israel

a r t i c l e i n f o

⁎ Corresponding author. Tel.: +972 3 6405917; faxE-mail address: karnig@post.tau.ac.il (K. Ginzburg)

0165-0327/$ – see front matter © 2009 Elsevier B.V.doi:10.1016/j.jad.2009.08.006

a b s t r a c t

Article history:Received 20 January 2009Received in revised form 11 August 2009Accepted 11 August 2009Available online 18 September 2009

Background: This study aims to: (a) follow-up the prevalence of comorbidity of posttraumaticstress disorder (PTSD), anxiety and depression; (b) determine the chronological relationsbetween these disorder; and (c) examine whether PTSD comorbid with anxiety and depressionis implicated in more impaired functioning than PTSD by itself.Methods: 664 war veterans were followed up 1, 2, and 20 years after their participation in the1982 Lebanon War. Comorbidity was assessed by self reported PTSD, anxiety, and depressionsymptoms; impairment in psychosocial functioning was assessed by self reported problems inoccupational, social, sexual and family functioning.Results: At each point of assessment, rates of triple comorbidity (PTSD, anxiety and depression;26.7–30.1%) were higher than rates of PTSD, either by itself (9.3–11.1%), or comorbid withdepression (1.2–4.5%) or anxiety (2.9–4.5%). PTSD predicted depression, anxiety, and comorbiddisorders, but not vice versa. At time 1 and 2 assessments, triple comorbidity was associatedwith more impaired functioning than PTSD alone. In addition, triple comorbidity at Time 2 wasassociated with more impaired functioning than double comorbidity.Limitations: Since measurements did not cover the entire span of 20 years since the war, theentire spectrum of changes could not be monitored.Conclusions: Almost one half of war veterans would endorse a lifetime triple comorbidity, andthose who do, are likely to have more impaired functioning. The findings support theperspective that views PTSD as the dominant disorder following traumatic events, whichimpels the development of comorbid anxiety and depression.

© 2009 Elsevier B.V. All rights reserved.

Keywords:Posttraumatic stress disorder (PTSD)AnxietyDepressionComorbidity

1. Introduction

The exposure to war-related traumatic events has longsince been found to potentially render long term psychiatricpsychopathology (Grinker and Spiegel, 1945). Posttraumaticstress disorder (PTSD) is considered as themost characteristicpsychiatric sequelae of traumatic experiences. Yet, previousstudies found that substantial proportions of PTSD casualties,ranging between 21% and 94%, suffer from co-morbiddepression (e.g., Ginzburg, 2007; Mollica et al., 1999;Salcioglu et al., 2003; Sundquist et al., 2005; Frayne et al.,

: +972 3 6409182..

All rights reserved.

2005; Hashemian et al., 2006), 39–97% suffer from co-morbidanxiety (Sundquist et al., 2005; Hashemian et al., 2006;Zayfert et al., 2002; Mayou et al., 2001; McFarlane and Papay,1992), and 11–67% endorse a triple-comorbidity, namelyhaving anxiety and depression in addition to the PTSD(Hashemian et al., 2006; Brady and Clary, 2004).

Four exclusive mechanisms were offered to explain thephenomenon of comorbidity in PTSD (see Breslau, 2002;Solomon and Bleich, 1998). The first explanation is that pre-existing psychiatric disorders increase the risk for PTSD,either by increasing the risk for exposure to traumatic eventsor increase victims' susceptibility to the PTSD-inducingeffects of trauma. This view received support from studiesshowing that among individuals suffering from both PTSD

250 K. Ginzburg et al. / Journal of Affective Disorders 123 (2010) 249–257

and depression, in most cases depression onset precede PTSDonset (O'Toole et al., 1998). The second hypothesis suggeststhat PTSD is a causal risk for anxiety and depression, whichthus may be considered as complications of the PTSD and itsimpairment. This hypothesis was also supported, by studiesthat showed that in most cases, depression and anxiety weresecondary to PTSD (Kessler et al., 1995; Engdahl et al., 1998;Franko et al., 2005); According to the third hypothesis PTSD,anxiety and depression are independent disorders which co-occur due to shared risk factors. This view was supported bystudies showing a simultaneous onset of the comorbiddisorders (e.g., Bleich et al., 1997) as well as studies pointingto shared risk factors, such as level of exposure to thetraumatic event (e.g., Vinck et al., 2007; Hoven et al., 2005;Schumm et al., 2006). Finally, according to the fourthexplanation, the observed comorbidity is merely an artifactof symptom overlap (see Southwick et al., 1991; Franklin andZimmerman, 2001). This explanation is supported by studiesshowing that individuals with comorbid PTSD and depressiondo not differ from those with PTSD alone in PTSD severity(Solomon and Bleich, 1998; Franklin and Zimmerman, 2001),psychosocial functioning (Stein et al., 2000), and treatmentresponsiveness (Labbate et al., 2004).

As can be seen, each of these perspectives is supported byempirical evidence. Yet, for each there is also refutingindication. These contradicting findings, as well as the factthat most of the studies that assess comorbidities of PTSD areeither cross-sectional (e.g., Franklin and Zimmerman, 2001)or based on retrospective assessment of lifetime disorders(Kessler et al., 1995; Franko et al., 2005; Bleich et al., 1997),point to the need for a systematic longitudinal examination ofthe chronological relations between the development ofPTSD, depression and anxiety following exposure to trau-matic event.

The present study uses data from a longitudinal study of acohort of Israeli veterans–with and without antecedentcombat stress reaction–who participated in the 1982 LebanonWar, and were followed up for 20 years. The study assessesPTSD, anxiety and depressive symptoms, as well as impair-ment in psychosocial functioning 1, 2, and 20 years followingtheir combat experience. The study aims to examine theprevalence of comorbidity of PTSD, anxiety and depression,and to explore the four mechanisms that were presented.More specifically, it examines two major research questions:

1. What are the chronological relations between PTSD,anxiety and depression: do anxiety and depression predictthe development of PTSD, as derived by the firstmechanism for explaining comorbidity (Breslau, 2002),or does PTSD predict the development of anxiety anddepression, as suggested by the second mechanism(Breslau, 2002)?

2. What is the clinical significance of comorbidity of PTSD,anxiety and depression? Do war veterans who developtriple comorbidity show more impaired psychosocialfunctioning than those who develop PTSD alone or doublecomorbidity (PTSD with either anxiety or depression)?Lack of difference in adjustment, would support the fourthmechanism for comorbidity, presented by Breslau (2002),while a significant difference would indicate that comor-bidity is more than an artifact of symptom overlap.

2. Methods

2.1. Sample and data collection

Two groups of Israeli male veterans participated in thisstudy. The first group consisted of 363 Israeli soldiers whofought in the Lebanon War and had been identified bymilitary mental health personnel as suffering from combatstress reaction (CSR) on the battlefield. Criteria for inclusionin this group were: 1) participation in frontline battles duringthe war, 2) a referral for psychiatric intervention made by thesoldier's battalion surgeon during the war, 3) a diagnosis ofcombat stress reaction made on the battlefield by clinicianstrained and experienced in the diagnosis of combat-relatedreactions, and 4) no indication in the clinician's report ofserious physical injury and other psychiatric disorders. Theresearch staff determined eligibility by using records ofclinicians' diagnoses made on the battlefield. The targetgroup included all combatants who met these criteria, andcame for treatment in an IDF mental health facility from thebeginning of the war, up to February 1983. The samplerepresented 78% of this group.

The comparison group consisted of 301 soldiers who hadparticipated in combat in the same units as those of the CSRgroup, but were not identified as suffering from CSR (responserate=84%). The two groups were matched in age, education,military rank and assignment. While it is difficult to control forthe subjective stressfulness of any combat experience, thesampling procedure used here was chosen to ensure thatsoldiers in both groups were exposed to a similar amount andtype of objective stress (for more detailed description of thesample see Solomon and Mikulncer, 2006).

In a preliminary stage, we ran all the statistical analysesfor each group separately. The analyses indicated that withregard to the research questions examined in this study, asimilar trend was observed in the two groups. A series of chi-square tests for independence showed no statistical differ-ences between CSR and control groups in the proportions ofthe disorders and comorbidities [χ2(5)=7.12, ns at Time 1,χ2(5)=3.83, ns at Time 2, and χ2(5)=6.56, ns at Time 3].Likewise, using Cohen and Cohen's (1983) procedure wefound no statistical differences between CSR and controlgroups in all of the correlations between study variables.Furthermore, multiple group analyses to assess potentialdifferences in the testedmodels, comparing the default modelthat allowed different effects within the groups, with theconstrained model that presupposed equality of the regres-sion coefficients between the groups, were conducted. All ofthe analyses showed no significant differences between CSRand control groups (Δχ2=18.57, df=12, ns for the firstmodel, Δχ2=18.64, df=14, ns for the second model, andΔχ2=15.1, df=12, ns for the last model). On this basis, inorder to avoid redundancies, the two groups were combinedinto one group.

Participants were assessed at three points of time: one(Time 1), two (Time 2), and twenty (Time 3) years after the1982 Lebanon War. All participants (n=664) had completedall the measures at the first wave of measurement (Time 1).Of them, 67% (n=448) completed Time 2 assessment, and64% (n=424) Time 3. Data retrieved from official militaryrecords and from the questionnaires filled out at Time 1

251K. Ginzburg et al. / Journal of Affective Disorders 123 (2010) 249–257

revealed that the men who participated at all three points intime did not significantly differ from those who declined toparticipate at times 2, or 3 in sociodemographic and militarybackground, premilitary adjustment, intelligence, or mentaland somatic health 1 year after the war. In all of the reportedanalyses, the overall statistical power for detecting significantsmall effects (i.e., d=.2, η2=.01, r=.1) was .83.

Participants with missing data on either Time 2 or Time 3assessmentwere included in the sample.We used the softwareprogramWinMICE (Jacobusse, 2005) inorder tohandlemissingdata by means of Multiple Imputation (Rubin, 1987; Schafer,1997) as recommended by Bollen and Curran (2006).

Veterans' age at Time 1 ranged between 18 and 37(M=25.81, SD=4.72, Mdn=26). Sixteen percent of theparticipants had completed only eighth grade, 27% had atmost somehigh school education, 39%had completedonly highschool, and 18% had studied beyond high school. All theveterans in both groups underwent stringent physical andpsychiatric screening before commencing theirmilitary serviceand no indication of premorbid symptomatology was recordedin their medical files.

3. Measures

Comorbidity of PTSD, anxiety and depression was exam-ined by the three following scales:

PTSD inventory

The PTSD inventory (Solomon et al., 1993) is based onDSM-III (APA, 1980), which was the standard used when thestudy commenced (1983). It was employed at all 3 points inorder to enable comparison over time. The inventory wascomprised of 13 items, corresponding to the 13 PTSDsymptoms listed in the DSM-III, each of them was anchoredat the war experiences (for example: “Did you experiencerecurrent scenes or thoughts about the war?”, “Did you try toavoid thought or feelings about the war?” Respondents wereasked to indicate for each statement, whether or not theyexperienced each symptomduring the lastmonth. Time 1 and2 measures used a dichotic response format (yes/no). In Time3, respondents were asked to indicate on a 4-point scale, theextent to which they endorse each symptom. As used in othersimilar measures (see for example the PCL, Forbes et al.,2001), responses of 3–4 were considered as indications forsymptom endorsement.

Respondents are identified as having PTSD according tothe DSM criteria: endorsing at least one each from thereexperiencing and numbing categories, and two hyperarousal symptoms. In addition, PTSD symptomatology wascalculated as the number of symptoms endorsed.

The PTSD inventory was used in various veteran popula-tions (Solomon and Dekel, 2005; Defrin et al., 2008). It hasconcurrent validity compared with structured clinical diag-nosis (SCID). Previous study demonstrated considerable ratesof agreement between the two measures (84–85%), revealinghigh level of specificity (94%) and moderate sensitivity (48–61%) of the PTSD Inventory (Solomon et al., 1993). Cronbachαs for the current sample were .90 at Time 1, .88 at Time 2,and .87 at Time 3, indicating high reliability.

Depression and anxiety subscales of the Symptoms Checklist-90(SCL-90)

Based on norms for psychiatric outpatients (Derogatis,1977) scores above .73 at each subscalewere considered as anindication for depression or anxiety.

High Cronbach αs indicated considerable reliability for thetwo subscales (for depressive symptoms, Cronbach αs were.92 at Time 1, .90 at Time 2, and .90 at Time 3; For anxietysymptoms, Cronbach αs were .89 at Time 1, .89 at Time 2, and.93 at Time 3).

Computation of co-morbidity scores

Based on the above assessments, comorbidity scores werecalculated for eachwave ofmeasurement. Each participantwasclassified into oneof the following eight categories: nodisorder,PTSD, depression, anxiety, PTSD+depression, PTSD+anxiety,depression + anxiety; and triple comorbidity (PTSD +depression + anxiety).

Psychosocial functioning

Impairment in psychosocial functioning was measured bya 29-item self-report questionnaire, that assesses problems infive areas of functioning: occupational performance, familyfunctioning, sexual functioning, social functioning and inter-personal relations, and social independence (see Solomonand Mikulincer, 2007).

In this study, Cronbach α coefficients were high, rangingbetween .88 and .91 across the three waves of measurement.

4. Statistical analyses

First, descriptive statistics (means and standard devia-tions) for PTSD, anxiety and depression, at Time 1, 2 and 3were presented. To investigate the trend of the changes inthese variables over time, we used LGM (Latent GrowthModels; see Bollen and Curran, 2006). The LGM estimatedwhether the trajectory of change was constant overtime(i.e. linear), or took another shape. This model estimatedtwo latent factors: the intercept of the developmentaltrajectory, and the slope of the trajectory (with factor load-ings set to 0, 1 and 19 to define the passage of time).A significant slope of LGM relates to a linear change in eachvariable (e.g. PTSD) following a change of one unit of time(i.e., 1 year).

The relations between PTSD, depression and anxietysymptoms were examined by Pearson correlations. Toexamine whether there are differential patterns of correla-tions, Cohen and Cohen's (1983) procedure, assessingsignificance of differences between correlations (Fisher Z),was employed. This was followed by calculations of distribu-tion of co-morbidity at each wave of assessment.

To examine the first research question, i.e., the longitudi-nal relations between PTSD, depression, and anxiety, Auto-regressive Crosslagged (ARCL) modeling strategy wasemployed, based on the classic simplex model (Guttman,1954; Anderson, 1960; Hemphreys, 1960; Heise, 1969;Jöreskog, 1979). This modeling strategy incorporates twomain components. First, later measures of a construct are

252 K. Ginzburg et al. / Journal of Affective Disorders 123 (2010) 249–257

predicted by earlier measures of the same construct, thusgiving rise to the “autoregressive” term. For example,

PTSDit = μ t + ρtPTSDi;t−1 + εit ð1:1Þ

indicating that the measure of PTSD for individual i at timepoint t is an additive combination of a time specific intercept(μt), a weighted contribution of the prior measure of PTSD(ρt−1), and an individual and time specific random error (εit).Larger positive values of the regression parameter are usuallyinterpreted as indicating greater stability of the constructover time, that is, on average, scores above the mean at time ttend also to be above the mean at time t+1.

This model can be extended to a multivariate model inwhich two or more construct are examined simultaneously.Here, not only measures of one construct regressed uponearlier measures of the same construct but the later measuresof one constructs are also regressed on earlier measures ofother constructs as well. For example, the measures ofdepression could be incorporated such that

PTSDit = μ t + ρ1PTSDi;t−1 + ρ2Depressioni;t−1 + εit ð1:2Þ

indicating that later measures of PTSD are a function of anintercept, the weighted contribution of prior measure ofPTSD, the weighted contribution of prior measure ofdepression, and a random error term. The model in Eq. 1.2thus has both the autoregressive component, but also has acrosslagged component such that earlier measures of depres-sion predict later measures of PTSD. This model can beextended to examine bi-directional relations such that earliermeasures of PTSD predict later measures of depression aswell. In order to assess the appropriateness of the ARCLmodels, we used the EQS 6.1 Structural Equation Models(SEM) software (Bentler and Wu, 1995). We estimated themodels' fit by using the comparative fit index (CFI) and theroot-mean-square error of approximation (RMSEA). A modelis judged as reasonably fitting the data when CFI and 1-RMSEA are larger then .90 (Schafer, 1997; Bollen & Curran,2006). In all of the reported SEM analyses we used family-wise Bonferroni correction to account for a possible inflationof Type I error when estimating multiple paths.

The second research question, concerning the clinicalimplications of triple comorbidity was examined in a series of

Table 1Correlations, means, and standard deviations for PTSD, anxiety and depression sym

PTSDT1 DEPT1 ANXT1 PTSDT2

M 4.76 0.78 0.94 4.70SD 3.93 0.83 0.91 3.62PTSDT1 –

DEPT1 0.49 –

ANXT1 0.50 0.87 –

PTSDT2 0.80 0.40 0.41 –

DEPT2 0.22 0.76 0.76 0.37ANXT2 0.21 0.87 0.79 0.40PTSDT3 0.48 0.22 0.22 0.50DEPT3 0.34 0.19 0.17 0.38ANXT3 0.33 0.17 0.15 0.37

Note. N=664 participants. All effect sizes are significant at p<.01. PTSDT1=ANXT1=anxiety symptoms at Time 1; PTSDT2=PTSD symptoms at Time 2; DEPTPTSDT13 = PTSD symptoms at Time 3; DEPT3=depressive symptoms at Time 3; A

MANOVAs. In each MANOVA, the relation of comorbidity(PTSD only vs. triple comorbidity, at either Time 1, Time 2, orTime 3) with psychosocial functioning difficulties (at eitherTime 1, Time 2, or Time 3) was examined. To ascertain thesource of the significant relation, eachMANOVAwas followedby a series of ANOVAs.

5. Results

5.1. Descriptive statistics

Means and standard deviations for PTSD, depression andanxiety symptoms are presented in Table 1. As can be seen,for each of these variables there is somewhat differentpattern of change over time. LGMs evaluating linearityindicated that whereas the model for PTSD symptomsshowed excellent fit for the data χ2(1)=.78, p=.37,CFI=1, 1-RMSEA=1, the models for depression and anxietysymptoms showed poor fit for the data χ2(1)=22.23,p<.001, CFI=.99, 1-RMSEA=.82 for anxiety, and χ2(1)=12.36, p<.001, CFI=1, 1-RMSEA=.87 for depression. Theanalyses showed that PTSD symptoms tended to slightlyincrease over time (a .02 yearly increase in symptoms level).Although depression and anxiety symptoms did not tend tochange linearly, the results imply that they tend to decreaseover time.

A series of Pearson correlations yielded significant relationsbetween PTSD, depression, and anxiety, both cross-sectionaland across time (see Table 1). Though the associations betweenall the three measures were high, the associations betweendepression and anxiety were significantly higher than theassociations between PTSD and each of the two othermeasures(Fisher's Zs>10.5, ps<.001).

Finally, the frequencies of comorbidity in each wave ofmeasurement are presented in Table 2. As can be seen, amongthose who endorsed at least one disorder, the most commoncategory was triple comorbidity. 48.8% of the participantsreported a lifetime triple comorbidity. Unlike other catego-ries, the rates of triple co-morbidity seem to increase overtime: Compared to triple co-morbidity in Time 1, 53 newparticipants showed triple co-morbidity in Time 2. Likewise,compared to Time 2, 112 new participants showed triple co-morbidity in Time 3.

ptoms.

DEPT2 ANXT2 PTSDT3 DEPT3 ANXT3

0.93 0.84 5.16 0.74 0.740.91 0.80 3.55 0.70 0.70

–

0.90 –

0.28 0.30 –

0.23 0.21 0.76 –

0.22 0.21 0.76 0.92 –

PTSD symptoms at Time 1; DEPT1=depressive symptoms at Time 1;2=depressive symptoms at Time 2; ANXT2=anxiety symptoms at Time 2;NXT3=anxiety symptoms at Time 3.

Table 2Rates of comorbidity according to wave of assessment.

Time 1 Time 2 Time 3

No disorder 273 (41.11%) 236 (35.54%) 248 (37.35%)PTSD only 67 (10.09%) 74 (11.14%) 62 (9.34%)Depression only 12 (1.81%) 10 (1.51%) 18 (2.71%)Anxiety only 31 (4.67%) 50 (7.53%) 20 (3.01%)PTSD + depression 16 (2.41%) 8 (1.20%) 30 (4.52%)PTSD + anxiety 30 (4.50%) 24 (3.61%) 19 (2.86%)Depression + anxiety 58 (8.73%) 80 (12.05%) 62 (9.30%)Triple co-morbidity 177 (26.66%) 182 (27.41%) 205 (30.09%)

Note. Triple co-morbidity=PTSD+depression+anxiety.

253K. Ginzburg et al. / Journal of Affective Disorders 123 (2010) 249–257

5.2. The longitudinal relations between PTSD, depression, andanxiety

In this section, we examined (a) the stability and cross-lagged influence of PTSD, depression, and anxiety, (b) thestability and crosslagged influence of PTSD and co-morbiditybetween depression and anxiety, and (c) the stability andcrosslagged influence of PTSD, co-morbidity between depres-sion and anxiety, and triple co-morbidity. In order to assessthe likelihood of suffering from a certain disorder we haveinsert the model variables as dummy variables (0=do not

Fig. 1. ARCLmodel assessing longitudinal stability and crosslagged effects of PTSD, deconstructs and/or error terms. Dashed lines represent non-significant predictions.p<.01.

have the disorder, and 1=have the disorder) and assessedthe SEM model using the ‘category variables as continuousvariables’ function in EQS (for additional information seeBentler & Wu, 1995). A positive coefficient denotes a relationin which suffering from a certain disorder (i.e., coding of 1) isassociated with increased likelihood of suffering from theother disorder as well (i.e., coding of 1). A negative coefficientdenotes a relation in which suffering from a certain disorder(i.e., coding of 1) is associated with decreased likelihood ofsuffering from the other disorder (i.e., coding of 0).

Fig. 1 presents the stability and crosslagged influence ofPTSD, depression and anxiety symptoms. The model fit thedata fairly well χ2(13)=112.56, p<.05, CFI=.96, 1-RMSEA=.90. The analyses revealed that the stability ofPTSD disorder was noticeably high: Participants who sufferedfrom PTSD in Time 1 tended to suffer from PTSD both in Time2 and Time 3. In contrast, the stability of both depression andanxiety was low: There were no significant autoregressiveeffects, since Time 1 depression was associated with Time 2depression, but this pattern was not extended to time 3.Similarly, anxiety at Time 1 was associated with anxiety atTime 2, but this pattern was not extended to Time 3.

More importantly, the analyses revealed that PTSD pre-dicted depression and anxiety above and beyond the auto-regressive effects, but not vice versa. In other words, suffering

pression (DEP) and anxiety (ANX). Dotted lines represent covariates betweenSolid lines represent significant predictions. All effect sizes are significant at

254 K. Ginzburg et al. / Journal of Affective Disorders 123 (2010) 249–257

from PTSD increased the likelihood of depression or anxiety insubsequent waves of measurement. However, suffering eitherfrom depression or from anxiety did not change the likelihoodof a PTSD in subsequent waves of measurement.

Since the two most frequent phenomena that precededtriple co-morbidity were PTSD and co-morbidity betweendepression and anxiety, another ARCL was conducted, inorder to assess the stability and crosslagged influence of PTSDand co-morbidity between depression and anxiety. Themodel fit the data fairly well χ2(4)=28.15, p< .05,CFI=.98, 1-RMSEA=.91. As can be seen in Fig. 2, the stabilityof suffering from both depression and anxiety at the sametime was low: participants suffering from depression andanxiety in Time 1 did not seem to continue to suffer from bothdepression and anxiety over time. Time 1 comorbid anxietyand depression was associated with that of Time 2, but thispattern was not extended to Time 3.

More importantly, the analyses revealed that PTSD pre-dicted co-morbidity between depression and anxiety aboveand beyond the autoregressive effects, but not vice versa. Inother words, suffering from PTSD increased the likelihood ofco-morbidity of depression and anxiety in subsequent wavesof measurement; however, suffering from co-morbidity ofdepression and anxiety symptoms did not change thelikelihood of PTSD symptoms in subsequent waves ofmeasurement.

To examine whether PTSD predict triple co-morbidity, athird ARCL model was conducted, in which we tested (a) thestability of full co-morbidity across time, and (b) the cross-lagged influence of PTSD and co-morbidity between depres-sion and anxiety on triple co-morbidity. The model fit thedata fairly well χ2(13)=59.29, p< .05, CFI=.98, 1-RMSEA=.93. As can be seen in Fig. 3, the stability of sufferingfrom triple co-morbidity was high: participants suffering

Fig. 2. ARCL model assessing longitudinal stability and crosslagged effects of PTSD arepresent covariates between constructs and/or error terms. Dashed lines represeneffect sizes are significant at p<.05.

from triple co-morbidity in Time 1 tended to suffer from fullco-morbidity over time.

The analysis also revealed that both PTSD and co-morbiditybetween depression and anxiety in Time 1 predicted triple co-morbidity in Time2.However, using Cohen andCohen's (1983)procedure, we have found that the association between PTSDand triple co-morbidity was significantly larger than theassociation between co-morbidity of depression and anxietyand triple co-morbidity, Fisher's Z=4.25, p<.001. Moreover,triple co-morbidity in Time 3, was predicted by PTSD in Time 2but not by co-morbidity between depression and anxiety inTime2.All of thepredictionswere significant above andbeyondautoregressive effects.

In sum, we have found support for the notion that PTSDimpelled the phenomenon of triple co-morbidity. It predicteddepression and anxiety; it predicted co-morbidity betweendepression and anxiety; and lastly, it solely predicted tripleco-morbidity over time.

5.3. The clinical significance of triple comorbidity

To examine whether participants with triple comorbiditydiffer in their psychosocial functioning than those with PTSDonly and thosewith double comorbidity (PTSD and depression,PTSD and anxiety) a series of ANOVAswas conducted. The firstANOVA,with Time1 psychosocial functioning difficulties as thedependant variables, and comorbidity at Time 1 (PTSD only,double comorbidity, triple comorbidity) as the independentvariable, yielded a significant effect (F(2,283)=4.18; p<.05;see Fig. 4a). Planned contrasts revealed that participants withtriple comorbidity at Time 1, reported on more impairedfunctioning than participants with only PTSD (t(283)=2.83,p<.01). The difference in the degree of psychosocial function-ing difficulties between participants with triple comorbidity

nd co-morbidity between depression and anxiety (ANX + DEP). Dotted linest non-significant predictions. Solid lines represent significant predictions. Al

l

Fig. 3. ARCL model assessing longitudinal stability and crosslagged effects of PTSD (PT), co-morbidity between depression and anxiety (ANX + DEP), and co-morbidity between PTSD, depression and anxiety (CM). Dotted lines represent covariates between constructs and/or error terms. Dashed lines represent non-significant predictions. Solid lines represent significant predictions. All effect sizes are significant at p<.05.

255K. Ginzburg et al. / Journal of Affective Disorders 123 (2010) 249–257

and double comorbidity was in the predicted direction but notstatistically significant (t(283)=1.10, p=.27).

A second ANOVA, with Time 2 psychosocial functioningdifficulties as the dependant variables, and comorbidity at Time2 (PTSD only, double comorbidity, triple comorbidity) as theindependent variable, yielded a significant effect (F(2,191)=9.34; p<.001; see Fig. 4b). Planned contrasts revealed thatparticipants with triple comorbidity at Time 2, reported onmore impaired functioning than participants with doublecomorbidity (t(191)=3.62, p<.001) or with only PTSD(marginally significant; t(191)=1.72, p=.08).

Finally, a third ANOVA, with Time 3 psychosocial func-tioning difficulties as the dependant variables, and comor-bidity at Time 3 (PTSD only, double comorbidity, triplecomorbidity) did not reveal significant results (F(2,311)=.22; p=.80; see Fig. 4c).

6. Discussion

Almost half of the veterans who participated in this studyendorsed a lifetime triple comorbidity. Rates of triple comor-bidity (26.7–30.1%) were higher than rates of PTSD, either byitself (9.3–11.1%), or with comorbid depression (1.2–4.5%) oranxiety (2.9–4.5%). 74–80% of the veterans with PTSD alsoendorsed comorbid anxiety or depression, or both. That is, inthis veteran population, PTSD is more often accompanied by

another disorder than manifested alone. This complexity ledsome authors to suggest that the diagnosis of PTSD may bereorganized, as a category of posttraumatic spectrum disorder,rather than a distinctive diagnostic entity (Jung, 2004; Moreauand Zisook, 2002). According to this suggestion, this categorywould include several subtypes of PTSD, PTSD with depressivefeatures and PTSD with anxiety features, are two of them.Similar view was presented by O'Donnell et al. (2004), whoconceptualized the posttrauma psychopathology as a generaltraumatic stress factor which is characterized by mixed PTSDand depressive symptoms.

While this possibility deserves further investigation, thefindings of our study also indicate that PTSD is the core reactionto traumatic stress. Our findings suggest that although manyPTSD sufferers endorse either depression or anxiety or both,PTSD is more stable over time than the other two disorders.Moreover measuring patterns of relations between PTSD,anxiety and depression over time revealed that in cases ofcomorbidity, depression and anxiety are secondary to PTSD.That is, PTSD predicts subsequent depression and anxiety,while suffering either from depression or from anxiety did notchange the likelihood of a PTSD in subsequent waves ofmeasurement. Moreover, the findings support the notion thatPTSD impels the phenomenon of triple co-morbidity. Itpredicted depression and anxiety; it predicted co-morbiditybetween depression and anxiety; and lastly, it solely predicted

Fig. 4. a. Time 1 psychosocial functioning difficulties according to comorbidityat Time 1. b. Time 2 psychosocial functioning difficulties according tocomorbidity at Time 2. c. Time 3 psychosocial functioning difficulties accordingto comorbidity at Time 3.

256 K. Ginzburg et al. / Journal of Affective Disorders 123 (2010) 249–257

triple co-morbidity over time. Consistent with previous studies(Kessler et al., 1995; Engdahl et al., 1998; Franko et al., 2005),these findings suggest that PTSD is a causal risk for anxiety anddepression, which, in turn,may be considered as complicationsof the PTSD and its impairment.

The second study question, regarding the clinical implica-tions of comorbidity, is important in light of the contrastingviews in the literature. While some scholars view thiscomorbidity as genuine, others claim that it is not an actualclinical phenomenon, but an artifact of overlap in symptoms(see Southwick et al., 1991; Franklin and Zimmerman, 2001).Our findings suggest that comorbidity has certain clinicalsignificance. At the Time 1 and 2 assessments, participantswith triple comorbidity endorsed more severe adjustmentdifficulties than those with PTSD alone. In addition, triplecomorbidity at Time 2 was implicated in more psychosocialfunctioning difficulties than double comorbidity. Thesefindings are consistent with several studies that suggestthat comorbidity complicates the trauma survivors' adjust-ment, showing that PTSD sufferers with comorbid depressionexhibit more severe PTSD (Maes et al., 2000), depression

(Kozaric-Kovacic et al., 2001), difficulties in psychosocialadjustment (Ginzburg, 2007; Momartin et al., 2004), and lesslow wave sleep (Woodward et al., 1996), than PTSD suffererswithout comorbid depression. Similar findings were found ina recent study of a large population-based sample, showingthat comorbidity of mental health disorders is associatedwithincrease in disability days (Merikangas et al., 2007).

Previous report indicated that veterans who sustained CSRhad higher rates of PTSD, compared to veterans without CSR(for comprehensive comparisons in of the course of PTSDamong the two groups see Solomon and Mikulincer, 2006).Yet, similar trends were observed in the two groups: In bothgroups, triple comorbidity was the most prevalent distresscategory. In addition, the nature of relations among the threecomorbid disorders, and that with psychosocial adjustmentdifficulties, was similar in the two groups. This findingsuggests that the course and nature of PTSD is robust,regardless to the timing of the rise of the stress disorder.

This study suffers from several limitations. The first isattrition, a well recognized problem in longitudinal studies.Second, the fact that PTSD, anxiety and depression wereassessed via self-report questionnaires, rather than by clinicalinterviews must be taken into account. Turner et al. (2003),who assessed PTSD, and depression among refugees, dem-onstrated that self-report measures overestimate the preva-lence of these disorders, compared to diagnoses made byclinical interviews. Thus, although previous study indicatedthat the self report measure of PTSD has a considerableconvergent validity when compared to clinical interview, thevalidity of self-report assessment should be reconsidered.

Another limitation is the absence of pre-war level ofdistress. This fact is of particular importance with respect todepression and anxiety, since PTSD is reactive construct bydefinition. Post war anxiety and depression, on the other hand,may reflect prior vulnerability rather than a reaction to thetraumatic event. All the soldiers in both groups underwentstringent physical and psychiatric screening before commenc-ing their military service and no indication of diagnosablepremorbid symptomatologywas recorded in theirmedicalfiles.Yet, one cannot rule out the possibility of pre-war vulnerability,or anxiety and depression that precede the development oflater PTSD, at any point of time since the war until Time 1assessment. Finally, ourmeasurements did not cover the entirespan of 20 years since the war. Therefore, we were unable tomonitor the entire spectrum of changes in the course of PTSDand its comorbid disorders, between 1984 and 2002.

These limitations notwithstanding, the study findings showthat almost one half of war veterans would suffer from triplecomorbidity, at least atonepointof timeduring the twodecadesafter the war, and that those who do, are likely to endorse highlevels of impairment in functioning. Moreover, examining thechronological relations among the three comorbid disordersupports the perspective that views PTSD as the dominantdisorder following exposure to traumatic stress, which impelsthe development of the comorbid disorders.

Role of funding sourceNothing declared.

Conflict of interestNo conflict declared.

257K. Ginzburg et al. / Journal of Affective Disorders 123 (2010) 249–257

References

American Psychiatric Associationc, 1980. Diagnostic and Statistical Manual ofMental Disorders, (3rd ed.). APA, Washington, DC.

Anderson, T.W., 1960. Some stochastic process models for intelligence testscores. In: Arrow, K.J., Karlin, K., Suppes, P. (Eds.), Mathematical Methodsin the Social Sciences. Stanford University Press, Stanford, CA.

Bentler, P.M., Wu, E.J.C., 1995. EQS for Windows User's Guide. MultivariateSoftware, Encino, CA.

Bleich, A., Koslowsky, M., Dolev, A., Lerer, 1997. Post-traumatic stressdisorder and depression. An analysis of comorbidity. Br. J. Psychiatry 170,479–482.

Bollen, K.A., Curran, P.J., 2006. Latent Curve Models. A Structural EquationPerspective. Wiley-Interscience, Hoboken, New Jersey.

Brady, K.T., Clary, C.M., 2004. Affective and anxiety comorbidity in post-traumatic stress disorder treatment trials of sertraline. Com. Psychiatry44, 360–369.

Breslau, N., 2002. Epidemiologic studies of trauma, posttraumatic stressdisorder, and other psychiatric disorders. Can. J. Psychiatry 47, 923–929.

Cohen, J., Cohen, P., 1983. Applied Multiple Regression/Correlation Analysisfor the Behavioral Sciences. Erlbaum, Hillsdale, NJ.

Defrin, R., Ginzburg, K., Solomon, Z., Polad, E., Bloch, M., Govezensky, M.,Schreiver, S., 2008. Quantitative testing of pain perception in subjectswith PTSD— implications for the mechanism of the coexistence betweenPTSD and chronic pain. Pain 38, 450–459.

Derogatis, L.R., 1977. The SCL-90-R Manual I: Scoring, Administration, andProcedure for the SCL-90. Johns Hopkins University, School of Medicine,Baltimore.

Engdahl, Dikel, T.N., Eberly, R., Blank, A., 1998. Comorbidity and course ofpsychiatric disorders in a community sample of former prisoners of war.Am. J. Psychiatry 155, 1740–1745.

Forbes, D., Creamer, M., Biddle, D., 2001. The validity of the PTSD checklist asa measure of symptomatic change in combat-related PTSD. BehaviourRes Therapy 39, 977–986.

Franklin, C.L., Zimmerman, M., 2001. Posttraumatic stress disorder andmajordepressive disorder: investigating the role of overlapping symptoms indiagnostic comorbidity. J. Nervous Mental Dis. 189, 548–551.

Franko, D.L., Thompson, D., Barton, B.A., 2005. Prevalence and comorbidity ofmajor depressive disorder in young black and white women. J. PsychiatricRes. 39, 275–283.

Frayne, S.M., Seaver, M.R., Loveland, S., Christansen, C., Spiro, A., Parker, V.A.,Skinner, K.M., 2005. Burden of medical illness in women with depressionand posttraumatic stress disorder. Archives Internal Med. 164, 1306–1312.

Ginzburg, K., 2007. Comorbidity of PTSD and depression following myocardialinfarction. J. Affect. Disord. 94, 135–143.

Grinker, R.R., Spiegel, J.P., 1945. Men Under Stress. Philadelphia, Blakiston.Guttman, L.A., 1954. A new approach to factor analysis. In: Lazarsfeld, P.F. (Ed.),

Mathematical Thinking in the Social Sciences. InColumbia University Press,New York, pp. 258–348.

Hashemian, F., Khoshnood, K., Desai, M.M., Falahati, F., Kasl, S., Southwick, S.,2006. Anxiety, depression, and posttraumatic stress in Iranian survivorsof chemical warfare. JAMA 296, 560–566.

Heise, D.R., 1969. Separating reliability and stability in test–retest correla-tion. Am. Sociological Rev. 34, 93–101.

Hemphreys, L.G., 1960. Investigations of the simplex. Psychometrika 25,313–323.

Hoven, C.W., Duarte, C.S., Lucas, et al., 2005. Psychopathology among NewYork City public school children 6 months after September 11. Arch. Gen.Psychiatry 62, 545–552.

Jacobusse, G., 2005. WinMICE User's Manual. TNO, The Netherlands.Jöreskog, K.G., 1979. Statistical estimation of structural models in longitu-

dinal developmental investigations. In: Nesselroade, J.R., Baltes, P. (Eds.),Longitudinal Research in the Study of Behavior and Development.Academic Press, New York.

Jung, K.E., 2004. Posttraumatic Spectrum Disorder: A Radical Revision:Psychiatric Times. XVIII.

Kessler, R.C., Broment, E., Hughes, M., 1995. Posttraumatic stress disorder inthe national comorbidity survey. Arch. Gen. Psychiatry 52, 1048–1060.

Kozaric-Kovacic, D., Hercigonja, D.K., Grubisic-Ilic, M., 2001. Posttraumaticstress disorder and depression in soldiers with combat experiences.Croatian Med. J. 42, 165–170.

Labbate, L., Sonne, S., Randal, C., Anton, R., Brady, K., 2004. Does comorbidanxiety or depression affect clinical outcomes in patients with post-

traumatic stress disorder and alcohol use disorders? ComprehensivePsychiatry 45, 304–310.

Maes, M., Mylle, J., Delmeire, L., Altamura, C., 2000. Psychiatric morbidity andcomorbidity following accidental man-made traumatic events: inci-dence and risk factors. Eur. Arch. Psychiatry Clin. Neurosci. 250, 156–162.

Mayou, R., Bryant, B., Ehlers, A., 2001. Prediction of psychological outcomes oneyear after a motor vehicle accident. Am. J. Psychiatry 158, 1231–1238.

McFarlane, A.C., Papay, P., 1992. Multiple diagnoses in posttraumatic stressdisorder in the victims of a natural disaster. J. Nervous Mental Dis. 180,498–504.

Merikangas, K.R., Ames, M., Cui, L., Stang, P.E., Ustun, T.B., Von Korff, M.,Kessler, R.C., 2007. The impact of comorbidity of mental and physicalconditions on role disability in the US adult household population. Arch.Gen. Psychiatry 64, 1180–1188.

Mollica, R.F., McInnes, K., Sarajlic, N., Lavelle, J., Sarajlic, I., Massagli, M.P.,1999. Disability associated with psychiatric comorbidity and healthstatus in Bosnian refugees living in Croatia. JAMA 282, 433–439.

Momartin, S., Silove, D., Manicavasagar, V., Steel, Z., 2004. Comorbidity of PTSDand depression: associationswith trauma exposure, symptom severity andfunctional impairment in Bosnian refugees resettled in Australia. J. Affect.Disord. 80, 231–238.

Moreau, C., Zisook, S., 2002. Rationale for posttraumatic stress spectrumdisorder. Psychiatr. Clin. North Am. 25, 775–790.

O'Donnell, M.L., Creamer, M., Pattison, P., 2004. Posttraumatic stress disorderand depression following trauma: understanding comorbidity. Am. J.Psychiatry 161, 1390–1396.

O'Toole, B.I.,Marshall, R.P., Schureck, R.J., Dobson,M., 1998. Posttraumatic stressdisorder and comorbidity in Australian Vietnam veterans: risk factors,chronicity and combat. Austr. New Zealand J. Psychiatry 32, 32–42.

Rubin, D.B., 1987. Multiple Imputation for Nonresponse in Surveys. Wiley,New York.

Salcioglu, E., Basoglu, M., Livanou, M., 2003. Long-term psychologicaloutcome for non-treatment-seeking earthquake survivors in Turkey. J.Nervous Mental Dis. 191, 154–160.

Schafer, J.L., 1997. Analysis of Incomplete Multivariate Data. Chapman & Hall,New York.

Schumm, J.A., Briggs, P.M., Hobfoll, S.S., 2006. Cumulative interpersonal traumasand social support as risk and resiliency factors in predicting PTSD anddepression among inner-city women. J. Traumatic Stress 19, 825–836.

Solomon, Z., Benbenishty, R., Neria, Y., Abramovitz, M., Ginzburg, K., Ohry, A.,1993. Assessment of PTSD: validation of the revised PTSD Inventory.Israeli J. Psychiatry Related Sciences 30, 110–115.

Solomon, Z., Bleich, A., 1998. Comorbidity of posttraumatic stress disorderand depression in Israeli veterans. CNS Spectrum 3, 16–21.

Solomon, Z., Dekel, R., 2005. PTSD among Israeli ex-prisoners of war: 18 and30 years after release. J. Clinical Psychiatry 66, 1031–1037.

Solomon, Z., Mikulincer, M., 2006. Trajectories of PTSD: a 20-year longitudinalstudy. Am. J. Psychiatry 163, 659–666.

Solomon, Z., Mikulincer, M., 2007. Posttraumatic intrusion, avoidance, andsocial functioning: a 20-year longitudinal study. J. Consulting Clin.Psychology 75, 316–324.

Southwick, S.M., Yehuda, R., Giller, E.L., 1991. Characterization of depression inwar-related posttraumatic stress disorder. Am. J. Psychiatry 148, 179–183.

Stein, M.B., McQuaid, R.J., Pedrelli, P., Lenoz, R., McCahill, M.E., 2000.Posttraumatic stress disorder in the primary care medical setting. Gen.Hosp. Psychiatry 22, 261–269.

Sundquist, K., Johansson, L., DeMarinis, V., Johansson Sundquist, J., 2005.Posttraumatic stress disorder and psychiatric co-morbidity: symptomsin a random sample of female Bosnian refugees. European Psychiatry 20,158–164.

Turner, S.W., Bowie, C., Dunn, G., Shapo, L., Yule, W., 2003. Mental health ofKosovan Albanian refugees in the UK. Br. J. Psychiatry 182, 444–448.

Vinck, P., Pham, P.N., Stover, E., Weinstein, H.M., 2007. Exposure to warcrimes and implications for peace building in Northern Uganda. JAMA298, 543–554.

Woodward, S.H., Friedman, M.J., Bliwise, D.L., 1996. Sleep and depression incombat related PTSD inpatients. Bio. Psychiatry 39, 182–192.

Zayfert, C., Becker, C.B., Unger, D.L., Shearer, D.K., 2002. Comorbid anxietydisorders in civilians seeking treatment for posttraumatic stressdisorder. J. Traumatic Stress 15, 31–38.