Journal of Psychosomatic Research 70 (2011) 318–327

Persistence of sleep disturbances following cognitive-behavior therapyfor posttraumatic stress disorder

Geneviève Belleville⁎, Stéphane Guay, André Marchand

Centre d'Étude du Trauma, Centre de Recherche Fernand-Seguin, Hôpital Louis-H. Lafontaine, Montreal, Québec (Québec), Canada H1N 3V2

Received 1 April 2010; received in revised form 10 August 2010; accepted 30 September 2010

Abstract

Objectives: The objectives of the present study were (1) toassess the impact of cognitive-behavior therapy (CBT) forposttraumatic stress disorder (PTSD) on associated sleep dis-turbances and (2) to explore the correlates of persistent sleepdifficulties in terms of anxiety and depression symptoms andperceived health. Method: Fifty-five individuals with PTSD wereadministered a series of assessments designed to evaluate sleep,PTSD symptoms, symptoms of anxiety and depression, andperceived health before and after individual CBT for PTSD andat 6-month follow-up. Results: Significant improvements wereobserved on sleep quality, sleep onset latency, sleep efficiency, andsleep disturbances. These changes were not fully maintained after 6months, and 70% of people who reported baseline sleep difficulties

⁎ Corresponding author. École de Psychologie, Pavillon Félix-Antoine-Savard, Bureau 1334, 2325, rue des Bibliothèques, Québec (Québec),Canada G1V 0A6. Tel.: +1 418 656 2131x4226; fax: +1 418 656 3646.

E-mail address: genevieve.belleville@psy.ulaval.ca (G. Belleville).

0022-3999/10/$ – see front matter © 2011 Elsevier Inc. All rights reserved.doi:10.1016/j.jpsychores.2010.09.022

(Pittsburgh Sleep Quality Index >5) still reported significantproblems with sleep after treatment. Persistent sleep difficultieswere associated with more severe posttraumatic, anxious, anddepressive symptoms as well as poorer health. Conclusion:Although CBT for PTSD had a favorable impact on sleep, themajority of participants suffered from residual sleep difficulties.Individuals with persistent sleep difficulties posttreatment mayexperience more residual posttraumatic, depression, and anxietysymptoms and poorer mental and physical health than those whodo not report sleep problems posttreatment. Further research in thisarea will allow clinicians to treat sleep problems in theseindividuals more effectively.© 2011 Elsevier Inc. All rights reserved.

Keywords: Anxiety; Cognitive-behavior therapy; Posttraumatic stress disorder; Sleep; Insomnia; Nightmares

Introduction

Patients with posttraumatic stress disorder (PTSD) reporta wide array of complaints and symptoms of poor sleepquality that are often presumed to be symptoms of PTSD;indeed, nightmares and problems falling or staying asleepappear among the fourth edition of the Diagnostic andStatistical Manual of Mental Disorders (DSM-IV) [1] criteriafor PTSD. However, the literature on sleep difficulties,particularly insomnia, suggests that even when a catalyst(e.g., a trauma) can be identified, perpetuating factors areoften responsible for the maintenance of sleep problems over

time [2,3]. For example, a person suffering from PTSD maydevelop sleep disturbances following the trauma (precipitat-ing factor). If that person repeatedly experiences sleeplessnights and frightening nightmares, he or she could begin tofear and avoid going to bed at night. Subsequent behaviorssuch as prolonged daytime napping or chronic worry aboutloss of sleep constitute perpetuating factors and maycontribute more to the maintenance of the sleep dysfunctionthan does the original trauma.

If sleep disturbances are solely symptoms of PTSD, theywould be expected to remit after successful PTSD treatment.However, if perpetuating factors develop, sleep difficultiesmay persist posttreatment despite successful therapy forPTSD. The majority of the empirical data on this subjectsuggest that cognitive-behavior therapy (CBT) for PTSD hasa small to moderate impact on concomitant sleep distur-bances. One study assessed the impact of a single session of

1 These participants were initially excluded for reasons related to theprimary objectives of the mother study, that is, to assess the effect of socialsupport in PTSD treatment outcome. This study required that participants'spouse or significant other participate in the assessment and in a fewtreatment sessions. Since the standardized treatment under study was notdesigned to deal with marital violence, individuals with a history ofviolence in their current relationship were excluded to avoid potentialexacerbation of existing relationship problems.

319G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

exposure on each PTSD symptom, as measured by theClinician-Administered PSTD Scale (CAPS) [4,5]. Theresults revealed a significant moderate treatment effect onnightmares (d=0.6) and a small, nonsignificant effect oninsomnia (d=0.3). An online psychoeducation and exposuretherapy program for participants with mild to severeposttraumatic symptoms produced a similar significantmoderate improvement in sleep problems posttreatment(reported Cohen's d=0.6) [6]. Another study reportedsignificant improvements in sleep after CBT for PTSD,with a large effect for nightmares and a moderate effect forinsomnia [7]. In a study that assessed the persistence ofinsomnia following individualized CBT for PTSD, 13 out of27 patients (48%) reported residual insomnia symptoms aftertreatment. Further, these authors found that insomnia wasone of the most severe residual symptoms posttreatment [8].

There are, however, several inconsistencies in theliterature in this area. Although two studies with PTSDsamples reported that CBT had very large and positivetreatment effects for nightmares and other sleep disturbances[9,10], a study with a similar sample reported no significantimprovement in nightmares following the application of oneof two specific behavioral strategies (imaginal exposure orapplied muscle relaxation) [11]. Further, three studies thatused the Pittsburgh Sleep Quality Index to measure sleep inPTSD patients reported contradictory results. The first studywas conducted with a sample of veteran inpatients with along history of severe PTSD symptoms. The majority ofpatients reported very small changes in sleep after CBT, andonly 19% of the sample demonstrated reliable improvementin sleep [12]. The average posttreatment score (12.7)suggested that persistent sleep disturbances were the rulerather than the exception. The second study [13] found alarge effect of CBT on sleep in the few female outpatientrape victims who responded well to treatment for PTSD. Thethird study reported that CBT was effective in reducing sleepdisturbances in a large sample of women who reported asexual assault, but that most of the participants did not returnto normal sleep functioning [14].

Type of trauma, PTSD chronicity, and the use ofempirically supported therapies are factors that couldmoderate the impact of treatment on sleep. However,inconsistencies regarding the impact on sleep still emergedwithin studies that recruited victims of rape [13,14], or thosewho recruited victims of earthquake [4,5], or those whorecruited people with combat-related PTSD [7,9,12,15].Likewise, no pattern could be drawn regarding PTSDchronicity; however, in all the reviewed studies, traumaticevents leading to PTSD had occurred at least 3 years beforeparticipation to the study on average. Whether impact onsleep is different in victims who are referred to treatmentmore rapidly is unknown. Finally, studies that have usedempirically supported psychotherapies, such as CBT,cognitive processing, or prolonged exposure [8,13,14], didnot appear to report greater improvements on sleep comparedto those who studied less well-established treatments

[5,6,10–12]. Together, these samples show a large amountof variability that may explain the inconsistencies in theliterature, highlighting the importance of investigating theserelationships with a heterogeneous sample.

The findings discussed above require replication with alarger, heterogeneous sample of patients and a validatedmeasure of sleep quality. This study aims to replicate andextend these findings in a sample of individuals sufferingfrom PTSD from various traumatic events. The objectives ofthe present study are (1) to assess the impact of CBT forPTSD on associated sleep disturbances and (2) to explore thecorrelates of persistent sleep difficulties in terms of anxietyand depression symptoms and perceived health. The primaryhypothesis is that CBT for PTSD will have a significantfavorable impact on concomitant sleep disturbances, but thatmost participants will report significant residual sleepdifficulties after treatment. The second hypothesis is thatpersistent sleep difficulties will be associated with moresevere anxiety and depression symptoms and lower levels ofperceived health.

Method

Participants and procedure

Participants with PTSD were recruited through mediaadvertisements and through referrals from hospitals in theMontreal metropolitan area in Quebec, Canada. Further,since this study was embedded in a larger study designed toassess the impact of social support on treatment for PTSD,each participant's spouse or significant other was required toparticipate. Exclusion criteria included (a) under 18 years ofage; (b) history of aggression by the spouse or significantother1; (c) alcohol or substance abuse/dependence; and (d)past or present psychotic episode, bipolar disorder, ororganic mental disorder.

A total of 583 individuals were screened using atelephone interview. Of these, 215 appeared eligible andwere invited for a thorough initial clinical assessment.Ninety-four met the criteria for PTSD and inclusion in thestudy. Nine individuals failed to complete the pretreatmentevaluation and were excluded or dropped out after their firsttreatment session. Thirteen further participants were exclud-ed during treatment: reasons for exclusion were repeatedabsenteeism to therapy sessions (n=8), marital separation(n=4), and the emergence of suicidal ideation (n=1). Tenwithdrew their participation: reasons were lack of

320 G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

availability (n=8) and moves (n=2). Finally, data from sevenparticipants who did not fully complete the sleep question-naire were excluded from the analyses. The final sample wascomposed of 55 individuals.

Participants signed a consent form and were administereda structured clinical interview that assesses PTSD and otherpsychiatric disorders (Structured Clinical Interview forDSM-IV, or SCID-IV [16]. Participants completed self-reportquestionnaires at home and returned them at the first treat-ment session. Participants were readministered the SCID-IVand completed the same series of questionnaires immedi-ately following treatment and 6 months posttreatment.

Measures

The Pittsburgh Sleep Quality IndexThe Pittsburgh Sleep Quality Index (PSQI) [17] is a 19-

item self-report questionnaire that assesses seven compo-nents of sleep quality: (a) subjective sleep quality; (b) sleeplatency; (c) sleep duration; (d) sleep efficiency; (e) sleepdisturbances; (f) use of sleep medication; and (g) impairmentof daytime functioning. Each component or subscale is ratedon a scale from 0 to 3, with higher scores indicating higherproblem severity. A global sleep quality score ranging from0 to 21 is calculated by summing the seven componentscores. The PSQI has demonstrated good internal consisten-cy, test–retest reliability, and overall validity in validationstudies with individuals suffering from insomnia or otherpsychiatric and medical disorders [17]. Validation of theFrench version has yielded satisfactory results [18].

The PSQI Addendum for PTSDThe PSQI Addendum for PTSD (PSQI-A) [19] is an

extension of the PSQI designed to assess the frequency ofseven disruptive nocturnal behaviors: (a) hot flashes; (b)general nervousness; (c) memories or nightmares of atraumatic experience; (d) severe anxiety or panic, not relatedto traumatic memories; (e) bad dreams, not related totraumatic memories; (f) episodes of terror or screamingduring sleep without fully awakening; and (g) episodes of“acting out” dreams, such as kicking, punching, running, orscreaming. Each item is rated on a 0–3 scale referring tofrequency of each disturbance, where 0=not in the pastmonth, 1=less than once a week, 2=once or twice a week,and 3=three or more times a week. A global score can beobtained from the sum of all seven items. To document thefrequency of each specific disturbances, however, we choseto analyze each item separately.

The Modified PTSD Symptom Scale—Self-ReportThe Modified PTSD Symptom Scale—Self-Report

(MPSS-SR) [20] is a 17-item self-report questionnaire thatassesses the frequency and severity of PTSD symptoms. TheMPSS-SR items correspond to the PTSD symptoms listed inDSM-IV-TR [1], and total scores range from 0 to 119. TheFrench version of the MPSS-SR has demonstrated good

psychometric properties in clinical samples [21]. To avoidoverlap between the MPSS-SR and measures of sleep, thecalculation of the total score excluded Items 2 (recurrentdistressing dreams of the event) and 13 (difficulty falling orstaying asleep).

The Beck Depression Inventory—Second EditionThe Beck Depression Inventory—Second Edition (BDI-

II) [22] is a 21-item measure of depressive symptoms.Participants assign a score from 0 to 3 to each item,indicating the severity of the symptom over the past 7 days.The BDI has been extensively validated and the psychomet-ric properties reported for the French version are satisfactory[23]. To avoid overlap with other measures, BDI-II Item 16(changes in sleeping patterns) was excluded from thecalculation of total score.

The Beck Anxiety InventoryThe Beck Anxiety Inventory (BAI) [24] is a 21-item self-

report questionnaire that targets cognitive and somaticsymptoms of anxiety. Participants assign a score from 0 to3 to each item, indicating the severity of the symptom overthe past 7 days. Total score ranges from 0 to 63; higherscores indicate greater anxiety. The French version of theBAI has good psychometric properties [25].

The Medical Outcome Survey—Short Form-12The Medical Outcome Survey—Short Form-12 (SF-12)

[26] is a self-report questionnaire that assesses eight areas ofhealth-related quality of life. The results can be summed toproduce two health-related component scores: perceivedmental health and perceived physical health. Norms for bothvariables are based on a 1998 normative population samplein the United States [26]. The SF-12 is an abbreviated formof the SF-36. A French version of the SF-36 has beenvalidated [27], but no psychometric data on the Frenchversion of the SF-12 are currently available.

Treatment

Trained clinical psychologists delivered individual treat-ment for PTSD. Treatment delivery was supervised by apsychologist specialized in CBT for PTSD (S. Guay). Astandardized treatment manual outlined the material to becovered in each session and guided therapists in theapplication of various therapeutic strategies. Therapistswere instructed to give a 20-session CBT, but could reducethe number of sessions down to 16, if they judged thatparticipants were in remission and no longer benefited fromtherapy, or increase it up to 24, if more time was needed tocomplete the exposure exercises. Thus, CBT consisted of amean number of 19.06 sessions (S.D.=3.03) lasting between60 and 90 min. The CBT protocol included strategies thatfollowed expert consensus guidelines [28]: (1) psychoeduca-tion about PTSD, (2) anxiety management techniques, (3)cognitive restructuring, (4) prolonged exposure, and (5)

Table 1Sociodemographic and clinical characteristics of sample

Mean S.D. Range

Age 40.73 12.81 18–68No. of traumatic events over the lifetime 1.50 1.04 1–5

n %Work statusEmployed 22 40Employed—currently on sick leave 8 15Unemployed 15 27Students 5 9Retired 4 7

Civil statusMarried or common-law relationship 25 45Single 23 42Divorced/separated 6 11

College education (or higher) 30 55Incomeb$20,000 CAD 21 38$20,000 CAD–$40,000 CAD 19 35N$40,000 CAD 14 26

Type of traumaPhysical aggression or threat 19 35Motor vehicle accident 14 25Witnessing a traumatic event 10 18Sexual aggression 3 5Marital violence 3 5Work accident 2 4Natural disaster 1 2Other 3 5

No. of years since the occurrence of traumab1 12 221–2 13 243–4 9 165+ 21 38

Comorbid major depressive disorder 24 44Comorbid anxiety disorder 21 38Use of psychotropic medicationAntidepressants 35 64Benzodiazepines 18 33Atypical antipsychotics 5 9Nonbenzodiapine hypnotics 4 7

321G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

relapse prevention. Participants received a manual thatcovered psychoeducation and CBT for PTSD. The wholecontent of this manual was reviewed in the psychoeducationsession at the beginning of therapy. Each therapy sessionfurther included the assignment of various exercises for theparticipant to practice at home. Participants were encouragedto observe and record their mood and symptoms on a dailybasis.

The manuscript detailing the results of the mother study iscurrently in preparation. Treatment effects on posttraumaticsymptoms were large, with effect sizes ranging from 0.93 to2.29. Treatment integrity was assessed by independentjudges on a sample of 45 sessions. Overall, 98% of thestrategies employed by therapists corresponded to those inthe treatment manual.

Data analyses

Repeated-measures analyses of variance (ANOVAs)assessed improvements in overall sleep quality and variousaspects of sleep over time. Significant time effects werefollowed by post hoc analyses of linear and quadratictendencies. To keep the probability of making a type I errorabove 5% familywise, only linear and quadratic tendenciesassociated with a P value of .025 were interpreted. A secondseries of repeated-measures ANOVAs explored the effects oftime, persistence of sleep difficulties, and their interaction onposttraumatic, depression, and anxiety symptoms, as well asperceived health. Significant group effects or interactionswere followed by three separate post hoc between-groupcomparisons, one for each assessment time. To keep theprobability of making a type I error above 5% familywise,only between-group comparisons associated with a P valueof .017 were interpreted.

Follow-up data were missing for 11 participants. To avoidreducing statistical power for the above-mentioned analyses,a “last observation carried forward” strategy was used—missing data on follow-up questionnaires were replaced byscores obtained at posttreatment.

Results

The final sample constituted 55 individuals withPTSD, 38 women (69%) and 17 men (31%). Socio-demographic and clinical characteristics of sample arepresented in Table 1.

Table 2 presents the means and standard deviations of thePSQI total and subscales scores and the items from the PSQI-A. Significant time effects were observed for the PSQI totalscore and the Sleep Quality, Sleep Onset Latency, SleepEfficiency, and Sleep Disturbances subscales. Linear andquadratic tendencies were analyzed post hoc for thesevariables. Significant linear effects, indicating improvementof sleep over time, were found for the PSQI total score [F(1,54)=6.71, P=.012], and the Sleep Quality [F(1,53)=6.70,

P=.012], Sleep Onset Latency [F(1,49)=10.47, P=.002],and Sleep Disturbances [F(1,54)=5.94, P=.018] subscales.Significant quadratic effects, indicating an initial improve-ment from pre- to posttreatment and a further worsening ofsymptoms from posttreatment to the 6-month follow-up,were observed for the PSQI total score [F(1,54)=10.22,P=.002] and the Sleep Onset Latency [F(1,49)=6.33,P=.015], Sleep Efficiency [F(1,51)=8.02, P=.007], andSleep Disturbances [F(1,54)=11.88, P=.001] subscales. Thequadratic effect associated to the Sleep Quality subscale didnot reach the significance criterion [F(1,53)=5.12, P=.028].

Regarding items of the PSQI-A, significant time effectswere observed for difficulty sleeping due to GeneralNervousness, Memories or Nightmares of Traumatic Expe-rience, Anxiety or Panic Not Related to TraumaticMemories, and Terror or Screaming. Linear and quadratictendencies were analyzed post hoc for these variables.

Table 2Means (and standard deviations) of PSQI total and subscale scores and items from the PSQI-A

Pretreatment Posttreatment 6-month follow-upMain effect oftime F (df) P

PSQI (total) 10.74 (4.77) 8.41 (4.85) 8.92 (5.04) 7.93 (2,108) .001PSQI (subscales)Sleep Quality 1.82 (0.90) 1.36 (0.89) 1.43 (0.92) 6.03 (2,106) .003Sleep Onset Latency 1.98 (1.06) 1.47 (1.14) 1.52 (1.20) 8.71 (2,98) .000Sleep Duration 1.07 (1.03) 0.91 (0.96) 0.94 (1.05) 1.11 (2,104) .371Sleep Efficiency 1.19 (1.23) 0.80 (1.07) 0.87 (1.12) 4.10 (2,102) .019Sleep Disturbances 1.93 (0.66) 1.58 (0.63) 1.69 (0.69) 8.21 (2,108) .000Impairment of Daytime Functioning 1.44 (0.71) 1.25 (0.93) 1.27 (0.76) 1.16 (2,108) .319Use of Sleep Medication 1.30 (1.37) 1.04 (1.33) 1.24 (1.37) 1.86 (2,106) .161

PSQI-A (“Difficulty sleeping because of…”)Hot flashes 1.20 (1.11) 0.92 (0.98) 0.83 (1.06) 2.68 (2 ,102) .073General nervousness 2.07 (0.91) 1.28 (1.01) 1.17 (1.11) 19.89 (2,102) .000Memories or nightmares of traumatic experience 1.83 (1.04) 0.75 (1.04) 0.83 (1.06) 28.66 (2,102) .000Anxiety or panic, not related to traumatic memories 1.02 (1.11) 0.62 (0.86) 0.87 (1.05) 3.40 (2,102) .037Bad dreams, not related to traumatic memories 1.17 (1.01) 0.83 (0.80) 0.98 (0.86) 2.26 (2,102) .110Terror or screaming 0.56 (0.84) 0.29 (0.67) 0.32 (0.73) 3.61 (2,100) .031“Acting out” dreams 0.65 (1.01) 0.34 (0.63) 0.44 (0.80) 1.75 (2,96) .180

322 G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

Significant linear effects, indicating a decrease in sleepdifficulties due to the following reasons, were found only forGeneral Nervousness [F(1,51)=25.63, Pb.001] and Memo-ries or Nightmares of Traumatic Experience [F(1,51)=29.33,Pb.001]. Significant quadratic effects, indicating an initialdecrease from pre- to posttreatment and a further reappear-ance of sleep difficulties due to the following reasons fromposttreatment to the 6-month follow-up, were observed forGeneral Nervousness [F(1,51)=9.55, P=.003], Memories orNightmares of Traumatic Experience [F(1,51)=27.35,Pb.001), and Anxiety or Panic Not Related to TraumaticMemories [F(1,51)=8.75, P=.005]. The quadratic effectassociated to Terror or Screaming did not reach thesignificance criterion [F(1,51)=5.12, P=.028].

The impacts of type of trauma, PTSD chronicity, andtherapy duration on the treatment effect were explored. Forthese analyses, only the PSQI total score was used as adependent variable. There was no significant interactionbetween the effect of treatment on sleep and the type oftrauma [F(4,50)=1.35, P=.264]. Visual exploration of datashowed similar moderate sleep improvement for victims ofphysical aggression, victims of motor vehicle accidents,victims of other traumatic events and individuals whowitnessed a traumatic event. However, mean PSQI totalscore of victims of sexual aggression (n=3) increased aftertreatment [from 7.67 (S.D.=1.53) to 9.33 (S.D.=5.77)].Number of months since the occurrence of the traumaticevent did not correlate with PSQI total score at anyassessment time. Number of therapy sessions did notcorrelate with pretreatment PSQI total score; however,positive correlations, indicating that people who received agreater number of treatment sessions had more sleepdifficulties, were observed during posttreatment (r=.35,P=.01) and 6-month follow-up (r=.30, P=.03) assessments.

Impact of persistent sleep difficulties

Prior to CBT, 85% of the sample (47/55) reportedsignificant sleep difficulties, as indicated by a PSQI scoregreater than 5. Following CBT, 30% (14/47) no longer metthis criterion, and 70% (33/47) still reported significantsleep difficulties (PSQI N5). Of the 15% (8/55) who didnot report significant sleep difficulties at baseline (PSQIscore of 5 or lower), 63% (5/8) were still good sleepersfollowing CBT and 37% (3/8) developed significantproblems with sleep.

To explore the impact of persistent sleep difficulties,the sample was separated in two groups, that is,individuals with persistent posttreatment sleep problems(n=33) and participants whose sleep difficulties did notpersist (n=14). [Individuals without sleep difficulties atbaseline (n=8) were not included.] Prior to CBT, bothgroups were equivalent: no significant differences wereobserved in terms of gender, age, employment, civilstatus, education, income, type of trauma, number oftraumatic events reported in the lifetime, time since theoccurrence of the traumatic event that led to currentPTSD, presence of comorbid anxiety or mood disorders,and use of psychotropic medication.

To assess differences between the two groups over thethree assessment times, repeated-measures ANOVAs werecomputed on MPSS-SR (Fig. 1), BDI-II (Fig. 2), BAI (Fig.3), SF-12 Mental Health (Fig. 4), and SF-12 Physical Health(Fig. 5) scores. All variables revealed significant timeeffects, with scores decreasing over time. All but one (theSF-12 Physical Health score) revealed significant groupeffects, with greater difficulties in participants withpersistent sleep difficulties. Significant Group × Timeinteractions were observed for MPSS-SR and both SF-12

70.38

39.9944.31

65.67

12.8615.43

0

10

20

30

40

50

60

70

80

90

100

Pre Post 6-month FU

MPS

S-SR

Mea

n Sc

ores

With Persistent Sleep Difficulties (n=33)

Without Persistent Sleep Difficulties (n=14)

Main Effects and Interaction

Time: F(2,90) = 77.25, P<.001

Group: F(1,45) = 10.86, P=.002

Time X Group: F(2,90) = 6.58, P=.002

Post Hoc Comparisons Between Groups

Pre: t(45) = -.89, P=.376

Post: t(44.34) = -4.52, P<.001

6-month FU: t(42.93) = -4.36, P<.001

Fig. 1. Posttraumatic symptoms (MPSS-SR scores) among individuals with and without persistent sleep difficulties.

323G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

scores, with less improvement over time in participants withpersistent sleep difficulties. Post hoc group comparisonswere computed at each assessment times. Results did notreveal any significant baseline differences. Differencesfavoring participants without persistent sleep difficultiesemerged at posttreatment in all five variables and weremaintained at the 6-month follow-up in all but one (the SF-12 Physical Health score).

27.88

22.79

0

5

10

15

20

25

30

35

40

Pre P

BD

I M

ean

Scor

es

Main Effects and Interaction

Time: F(2,90) = 22.98, P<.001

Group: F(1,45) = 16.88, P<.001

Time X Group: F(2,90) = 2.85, P=.063

Post Hoc Comparis

Pre: t(45) = -1.60, P

Post: t(44.18) = -4.

6-month FU: t(45)

Fig. 2. Depressive symptoms (BDI-II scores) among indiv

Discussion

The impact of CBT on associated sleep disturbances wasassessed in a heterogeneous sample of people with PTSD.Significant improvements were observed on sleep quality,sleep onset latency, sleep efficiency, and sleep disturbances.These changes were not fully maintained after 6 months, and70% of the subsample with baseline sleep difficulties still

18.56 20.82

5.936.02

ost 6-month FU

With Persistent Sleep Difficulties (n=33)

Without Persistent Sleep Difficulties (n=14)

ons Between Groups

=.116

43, P<.001

= -3.79, P<.001

iduals with and without persistent sleep difficulties.

26.42

17.72 19.4621.31

4.215.24

0

5

10

15

20

25

30

35

40

45

Pre Post 6-month FU

BA

I M

ean

Scor

es

With Persistent Sleep Difficulties (n=33)

Without Persistent Sleep Difficulties (n=14)

Main Effects and Interaction

Time: F(2,90) = 22.19, P<.001

Group: F(1,45) = 14.87, P<.001

Time X Group: F(2,90) = 2.85, P=.063

Post Hoc Comparisons Between Groups

Pre: t(45) = -1.36, P=.181

Post: t(42.63) = -4.77, P<.001

6-month FU: t(40.99) = -4.52, P<.001

Fig. 3. Anxiety symptoms (BAI scores) among individuals with and without persistent sleep difficulties.

324 G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

reported problems with sleep after treatment—confirmingour first hypothesis. Similar patterns were observed for sleepdifficulties due to general nervousness, nightmares, andanxiety or panic not related to traumatic memories.

Persistent sleep difficulties—a PSQI total score above 5at both baseline and posttreatment—were associated withmore severe posttreatment posttraumatic, anxious, anddepressive symptoms as well as poorer health. Paired with

31.3630.64

0

10

20

30

40

50

60

70

Pre P

SF-1

2 M

enta

l Hea

lth M

ean

Scor

es

Main Effects and Interaction

Time: F(2,90) = 25.14, P<.001

Group: F(1,45) = 8.45, P=.006

Time X Group: F(2,90) = 6.04, P=.003

Post Hoc Comparis

Pre: t(45) = -.251, P

Post: t(45) = -3.14,

6-month FU: t(45)

Fig. 4. Self-perceived mental health (SF-12 mental health component scor

the observation that such differences were not detected atbaseline, these findings confirmed our second hypothesis.

The present findings are congruent with trends observedin the literature that suggest that CBT for PTSD has afavorable, albeit moderate, impact on concomitant sleepproblems [4–7,14,15] and that residual sleep difficulties canbe expected [8,14]. Three prior studies used the PSQI toassess the evolution of sleep problems. One reported much

38.21

36.97

49.2948.54

ost 6-month FU

With Persistent Sleep Difficulties (n=33)

Without Persistent Sleep Difficulties (n=14)

ons Between Groups

=.803

P=.003

= 3.05, P=.004

es) among individuals with and without persistent sleep difficulties.

41.0841.13 43.2142.78

51.4050.34

0

10

20

30

40

50

60

70

Pre Post 6-month FU

SF-1

2 Ph

ysic

al H

ealth

Mea

n Sc

ores

With Persistent Sleep Difficulties (n=33)

Without Persistent Sleep Difficulties (n=14)

Main Effects and Interaction

Time: F(2,90) = 7.93, P=.001

Group: F(1,45) = 3.61, P=.064

Time X Group: F(2,90) = 5.24, P=.007

Post Hoc Comparisons Between Groups

Pre: t(45) = .434, P=.667

Post: t(41.17) = 3.68, P=.001

6-month FU: t(45) = 1.95, P=.057

Fig. 5. Self-perceived physical health (SF-12 mental physical component scores) among individuals with and without persistent sleep difficulties.

325G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

smaller changes than those found in the present study [12];however, the discrepancy may be a function of theconsiderable difference in the populations studied. Fordand colleagues' participants were inpatients in a Departmentof Veterans Affairs hospital. Every patient suffered from wartrauma (90% had served in the Vietnam War) and had ahistory of severe psychosocial impairment. The lack ofimprovement in sleep difficulties after treatment in thissample may reflect the nature, severity, and chronicity of theposttraumatic symptomatology in this sample rather than thetrue effect of CBT for PTSD on sleep. In our sample,duration of therapy—but not time since the occurrence of thetraumatic event—was positively correlated with problemswith sleep at posttreatment and follow-up. People whoneeded more therapy had more residual sleep difficulties.This may suggest the need to integrate sleep managementstrategies in order to optimize the benefits of CBT for PTSD.

The second study [11] found an improvement in PSQIscores among treatment responders that was greater than thatfound in the present study. This inconsistency may alsoreflect sample differences between the two studies. Nishithand colleagues' sample only reported sleep data for treatmentresponders, whereas our analyses were performed with thewhole sample of recruited participants, regardless oftreatment outcomes. Finally, the conclusions of the presentstudy are similar to the ones of a study published recentlythat examined the question among women reporting sexualassault [14]. Our findings suggested that sleep difficulties ofvictims of sexual aggression may be even more resistant tochange than that of other victims. Given the very smallnumber of victims of sexual assault in our sample, however,this findings need replication. The conclusions that CBT for

PTSD helps to improve sleep but that residual sleepproblems are to be expected appear generalizable tooutpatient women and men who suffer from PTSDsecondary to a variety of traumatic events.

Several factors may limit the generalizability of thepresent findings. First, our sample of outpatient volunteerswith PTSD may not be representative of individuals withPTSD who present in community medical or psychiatricsettings. Participants who had been victims of physicalaggression, threat, or sexual violence within the marital orspousal relationship were excluded from the study (as afunction of the exclusion criteria in the mother study); thisexclusion may prevent the results from generalizing to thosePTSD patients. It is noteworthy, however, that participantsreported a wide range of traumatic events, comorbidconditions, and sleep problems, and had all been diagnosedwith severe PTSD by trained clinical evaluators. Second,with the exception of psychiatric diagnoses, the resultsrelied heavily on data from self-report questionnaires,measures that are subject to memory and motivationalbiases. Nevertheless, despite its limitations, subjectiveevaluation is an appropriate method for assessing sleepdifficulties [29], and provides more readily accessible datathan results from sleep laboratories. Further, many clin-icians are accustomed to working with subjective data andmay therefore find this type of information more meaning-ful. Third, this study had an elevated attrition rate and manyparticipants did not provide complete data: only 55 out of94 participants provided complete pre- and posttreatmentdata, and an additional 11 individuals were lost to follow-up. Attrition is very prevalent in PTSD treatment studies[30] and has been associated with elevated levels of

326 G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

catastrophic cognitions and avoidance [31]. Integratingsleep management strategies focusing on insomnia ornightmares would allow therapists to avoid addressing thetrauma directly and to help the more avoidant patients tocommit to therapy before deciding to include, or not, moretrauma-focused strategies. Finally, for the 18 participantswho used benzodiazepines, no information was collectedabout whether they used it as a sleep aid, and changes in thepatterns of use of sedative psychotropic medication—eitherbenzodiazepines or others—after treatment were nottracked. Changes in use of sedative medication couldpossibly have accounted for some of the findings.

The present findings highlight the importance ofevaluating and addressing sleep disturbances in patientssuffering from PTSD. Fortunately, several nonpharmaco-logical strategies for managing sleep have been demon-strated to be consistently efficient in cases of chronicinsomnia [32] and nightmares [33]. Moreover, recent workhas demonstrated that it is feasible, even preferable, toincorporate sleep management strategies into treatment forPTSD [34]. Empirical replication and extension of thesefindings is necessary.

Acknowledgments

This research was supported by a postdoctoral grantfrom the Fonds de Recherche en Santé du Québec awardedto the first author. The authors have no conflict of interestto disclose.

References

[1] American Psychiatric Association. Diagnostic and Statistical Manualof Mental Disorders. 4th ed. Text Revised. Washington, DC: AmericanPsychiatric Association, 2000.

[2] Espie CA. Understanding insomnia through cognitive modelling.Sleep Med 2007;8:S3–8.

[3] Spielman AJ, Glovinsky P. The varied nature of insomnia. In: HauriPJ, editor. Case studies in insomnia. New York: Plenum Press, 1991.

[4] Salcioglu E, Basoglu M, Livanou M. Effects of live exposure onsymptoms of posttraumatic stress disorder: the role of reducedbehavioral avoidance in improvement. Behav Res Ther 2007;45:2268–79.

[5] Basoglu M, Salcioglu E, LivanouM. A randomized controlled study ofsingle-session behavioural treatment of earthquake-related post-traumatic stress disorder using an earthquake simulator. Psychol Med2007;37:203–13.

[6] Lange A, Rietdijk D, Hudcovicova M, van de Ven J-P, Schrieken B,Emmelkamp PMG. Interapy: a controlled randomized trial of thestandardized treatment of posttraumatic stress through the internet. JConsult Clin Psychol 2003;71:901–9.

[7] Frueh BC, Turner SM, Beidel DC, Mirabella RF, et al. Traumamanagement therapy: a preliminary evaluation of a multicomponentbehavioral treatment for chronic combat-related PTSD. Behav ResTher 1996;34:533–43.

[8] Zayfert C, DeViva JC. Residual insomnia following cognitivebehavioral therapy for PTSD. J Trauma Stress 2004;17:69–73.

[9] Cooper NA, Clum GA. Imaginal flooding as a supplementarytreatment for PTSD in combat veterans: a controlled study. BehavTher 1989;20:381–91.

[10] Gersons BPR, Carlier IVE, Lamberts RD, van der Kolk BA.Randomized clinical trial of brief eclectic psychotherapy for policeofficers with posttraumatic stress disorder. J Trauma Stress 2000;13:333–47.

[11] Vaughan K, Armstrong MS, Gold R, O'Connor N, Jenneke W, TarrierN. A trial of eye movement desensitization compared to imagehabituation training and applied muscle relaxation in post-traumaticstress disorder. J Behav Ther Exp Psychiatry 1994;25:283–91.

[12] Ford JD, Fisher P, Larson L. Object relations as a predictor of treatmentoutcome with chronic posttraumatic stress disorder. J Consult ClinPsychol 1997;65:547–59.

[13] Nishith P, Duntley SP, Domitrovich PP, Uhles ML, Cook BJ, SteinPK. Effect of cognitive behavioral therapy on heart rate variabilityduring REM sleep in female rape victims with PTSD. J Trauma Stress2003;16:247–50.

[14] Galovski TE, Monson C, Bruce SE, Resick PA. Does cognitive-behavioral therapy for PTSD improve perceived health and sleepimpairment? J Trauma Stress 2009;22:197–204.

[15] Keane TM, Fairbank JA, Caddell JM, Zimering RT. Implosive(flooding) therapy reduces symptoms of PTSD in Vietnam combatveterans. Behav Ther 1989;20:245–60.

[16] First MB, Spitzer RL, Gibbon M, Williams JBW. Structured ClinicalInterview for DSM-IV Axis I Disorders, Clinician Version (SCID-CV).Washington, D.C.: American Psychiatric Press, Inc., 1996.

[17] Buysse DJ, Reynolds 3rd CF, Monk TH, Berman SR, Kupfer DJ. ThePittsburgh Sleep Quality Index: a new instrument for psychiatricpractice and research. Psychiatry Res 1989;28:193–213.

[18] Blais FC, Gendron L, Mimeault V, Morin CM. Evaluation del'insomnie: validation de trois questionnaires. [Evaluation of insomnia:validity of 3 questionnaires]. Encephale 1997;23:447–53.

[19] Germain A, Hall M, Krakow B, Shear MK, Buysse DJ. A brief sleepscale for Posttraumatic Stress Disorder: Pittsburgh Sleep Quality IndexAddendum for PTSD. Anxiety Disord 2005;19:233–44.

[20] Falsetti SA, Resnick H, Resick PA, Kilpatrick DG. The ModifiedPTSD Symptom Scale: a brief self-report measure for assessingposttraumatic stress disorder. Behav Ther 1993;16:161–2.

[21] Guay S, Marchand A, Iucci S, Martin A. Validation de la versionquébécoise de l'Échelle modifiée des symptômes du trouble de stresspost-traumatique auprès d'un échantillon clinique. [Validation of theFrench-Canadian version of the Modified Post-Traumatic SymptomScale in a clinical setting]. Revue québécoise de psychologie 2002;23:257–67.

[22] Beck AT, Steer RA, Ball R, Ranieri W. Comparison of BeckDepression Inventories-IA and -II in psychiatric outpatients. J PersAssess 1996;67:588–97.

[23] Gauthier J, Morin C, Thériault F, Lawson JS. Adaptation françaised'une mesure d'auto-évaluation de l'intensité de la dépression. [Frenchadaptation of a self-administered measure of depression severity]. RevQué Psychol 1982;3:13–27.

[24] Beck AT, Steer RA. Beck Anxiety Inventory Manual. San Antonio,TX: The Psychological Corporation, 1990.

[25] Freeston MH, Ladouceur R, Thibodeau N, Gagnon F, Rhéaume J.L'inventaire d'anxiété de Beck: propriétés psychométriques d'unetraduction française. [Beck Anxiety Inventory: psychometric proper-ties of a French translation]. Encephale 1994;20:47–55.

[26] Ware JE, Kosinski M, Turner-Bowker DM, Gandeck B. User's manualfor the SF-12v2 Health Survey. Lincoln, Rhode Island: QualityMetricIncorporated, 2007.

[27] Dauphinee SW, Gauthier L, Gandek B, Magnan L, Pierre U. Readyinga US measure of health status, the SF-36, for use in Canada. Clin InvestMed 1997;20:224–38.

[28] Anxiety Disorders Association of America US. The expert consensusguideline series: treatment of posttraumatic stress disorder. J ClinPsychiatry 1999;60(Suppl 16):4–76.

[29] Buysse DJ, Ancoli-Israel S, Edinger JD, Lichstein KL, Morin CM.Recommendations for a standard research assessment of insomnia.Sleep 2006;29:1155–73.

327G. Belleville et al. / Journal of Psychosomatic Research 70 (2011) 318–327

[30] Matthieu M, Ivanoff A. Treatment of human-caused trauma: attritionin the adult outcomes research. J Interpers Violence 2006;21:1654–64.

[31] Bryant RA, Moulds ML, Mastrodomenico J, Hopwood S, FelminghamK, Nixon RDV. Who drops out of treatment for post-traumatic stressdisorder? Clin Psychol 2007;11:13–5.

[32] Morin CM, Bootzin RR, Buysse DJ, Edinger JD, Espie CA,Lichstein KL. Psychological and behavioral treatment of insomnia:

update of the recent evidence (1998–2004). Sleep 2006;29:1398–414.

[33] Lancee J, Spoormaker VI, Krakow B, van den Bout J. A systematicreview of cognitive-behavioral treatment for nightmares: toward awell-established treatment. J Clin Sleep Med 2008;4:475–80.

[34] DeViva JC, Zayfert C, Pigeon WR, Mellman TA. Treatment ofresidual insomnia after CBT for PTSD: case studies. J Trauma Stress2005;18:155–9.