Psychological Trauma: Theory,Research, Practice, and PolicyThe Impact of Prolonged Exposure on Sleep andEnhancing Treatment Outcomes With Evidence-BasedSleep Interventions: A Pilot StudyElizabeth M. Walters, Melissa M. Jenkins, Carla M. Nappi, Jacob Clark, July Lies, Sonya B. Norman,and Sean P. A. DrummondOnline First Publication, June 27, 2019. http://dx.doi.org/10.1037/tra0000478

CITATIONWalters, E. M., Jenkins, M. M., Nappi, C. M., Clark, J., Lies, J., Norman, S. B., & Drummond, S. P. A.(2019, June 27). The Impact of Prolonged Exposure on Sleep and Enhancing Treatment OutcomesWith Evidence-Based Sleep Interventions: A Pilot Study. Psychological Trauma: Theory,Research, Practice, and Policy. Advance online publication.http://dx.doi.org/10.1037/tra0000478

The Impact of Prolonged Exposure on Sleep and Enhancing TreatmentOutcomes With Evidence-Based Sleep Interventions: A Pilot Study

Elizabeth M. WaltersMonash University

Melissa M. JenkinsCenter for Stress and Anxiety Management,

San Diego, California

Carla M. NappiGrand Junction Veterans Affairs Medical Center, Grand

Junction, Colorado

Jacob Clark and July LiesMonash University

Sonya B. NormanVeterans Affairs Center of Excellence for Stress and MentalHealth, San Diego, California, and University of California,

San Diego

Sean P. A. DrummondMonash University

Objective: Insomnia and nightmares are central features of posttraumatic stress disorder (PTSD).However, often they are inadequately assessed and ineffectively resolved following gold-standard PTSDtreatment. Here we: (a) evaluate effects of prolonged exposure (PE) on subjectively measured sleep and(b) present pilot results of an examination of whether adding sleep interventions (imagery rehearsaltherapy [IRT] and cognitive-behavioral therapy for insomnia [CBT-I]) to PE improves treatmentresponse, relative to PE alone, for night- and/or daytime PTSD symptoms among returning U.S. veteransand postdeployment personnel. Method: In a parallel-groups, randomized controlled trial, participantsreceived 12 sessions of PE followed by IRT (5 weeks) and CBT-I (7 weeks) or PE followed by 12 weekssupportive care therapy (SCT). Results: PE did not improve sleep to a clinically meaningful degree,despite significant improvements in both Clinical Administered PTSD Scale and PTSD Checklist.Enhancing treatment with IRT/CBT-I led to greater improvements in insomnia (diary-recorded sleepefficiency) symptoms with large effect size, relative to SCT (p � .068, d � 1.07). There were largeimprovements in nightmare frequency relative SCT that did not reach statistical significance (p � .11,d � 0.90). Moreover, there was small improvement in daytime symptoms (Clinical Administered PTSDScale) that did not reach statistical significance (p � .54, d � .31). Conclusion: The addition of targeted,validated sleep treatment improves effects of PE and improves nighttime symptoms. Thus, evidence-based sleep treatment should be considered in comprehensive PTSD treatment.

Elizabeth M. Walters, School of Psychological Sciences and TurnerInstitute for Brain and Mental Health, Monash University; Melissa M.Jenkins, Center for Stress and Anxiety Management, San Diego, Califor-nia; Carla M. Nappi, Grand Junction Veterans Affairs Medical Center,Grand Junction, Colorado; Jacob Clark and July Lies, School of Psycho-logical Sciences and Turner Institute for Brain and Mental Health, MonashUniversity; Sonya B. Norman, Veterans Affairs Center of Excellence forStress and Mental Health, San Diego, California, and Department ofPsychiatry, University of California, San Diego; Sean P. A. Drummond,School of Psychological Sciences and Turner Institute for Brain andMental Health, Monash University.

We express our deep thanks to Jennifer Salamat and Kathy Resovsky fortheir work on recruitment, screening, and assessment on the project as wellas the veterans for their time and their service. We thank the followingindividuals for serving as treatment fidelity rater for this study: Steven R.

Thorp (PE), Michael L. Perlis (CBT-I), Richard J. Ross (IRT), and John R.McQuaid (SCT).

This work was supported by the National Institute of Nursing Research(1-RC1-NR011728-01). Elizabeth M. Walters, Jacob Clark, and July Liesreceive financial support from the Australian Government throughResearch Training Program scholarships. Funders played no role indesign or implementation of the project or analysis or interpretation ofthe data.

The authors declare no conflicts of interest associated with this publi-cation.

Trial registry is at www.clinicaltrials.gov (identifier: NCT01009112).Correspondence concerning this article should be addressed to Sean

P. A. Drummond, Monash Institute for Cognitive and Clinical Neurosci-ences, School of Psychological Sciences, Monash University, 18 Innova-tion Walk, Clayton VIC 2300, Australia. E-mail: sean.drummond@monash.edu

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Psychological Trauma:Theory, Research, Practice, and Policy

© 2019 American Psychological Association 2019, Vol. 1, No. 999, 0001942-9681/19/$12.00 http://dx.doi.org/10.1037/tra0000478

1

Clinical Impact StatementSleep problems, namely insomnia and nightmares, are a key part of posttraumatic stress disorder(PTSD). However current gold-standard exposure treatments for PTSD do not fully resolve sleepproblems. Here we demonstrate that prolonged exposure reduced global PTSD symptoms but did notimprove sleep. Subsequent evidence-based sleep interventions, imagery rehearsal therapy andcognitive-behavioral therapy for insomnia, improved nighttime, with a modest impact on daytime,symptoms. Results show need to specifically target both daytime and nighttime PTSD symptoms foroptimal clinical outcomes.

Keywords: sleep, veterans, cognitive-behavioral therapy for insomnia, prolonged exposure, imageryrehearsal therapy

Exposure to trauma during military operations is associated withnumerous negative outcomes, including posttraumatic stress dis-order (PTSD). Lifetime prevalence of PTSD is estimated at 11–17% for recent Iraq and Afghanistan veterans (Hoge et al., 2004).Consequences of PSTD include attempted suicide, functional im-pairment, alcohol abuse and dependence, reduced health-relatedquality of life, and significant mental distress (Kessler, 2000). Forveterans with PSTD, maximizing treatment efficacy is especiallycritical after discharge when trauma survivors must reintegrate intofamily and community and return to work and/or school. Effectiveevidence-based treatments for PTSD exist; however, one set ofsymptoms often unresolved is sleep disturbance.

Sleep disturbances are part of the diagnostic criteria for PTSDand are recognized as a core feature of PTSD (Spoormaker &Montgomery, 2008). Up to 90% of veterans (Maher, Rego, &Asnis, 2006) with PTSD report sleep disturbances, including for-mal diagnoses of insomnia and nightmares (Maher et al., 2006)and reduced subjective and objective sleep duration and sleepquality (Cox & Olatunji, 2016). In treatment-seeking samples,sleep measures almost universally show impairment well withinclinical ranges, including on global sleep measures (Galovski,Monson, Bruce, & Resick, 2009; Zayfert & DeViva, 2004), spe-cific sleep symptoms (Pruiksma et al., 2016), and night-to-nightvariability (Straus, Drummond, Nappi, Jenkins, & Norman, 2015).Sleep disturbances in PTSD are a particular concern because theymay impact the overall effectiveness of interventions by increasingdistress and daytime impairment above and beyond that related todaytime symptoms (Giosan et al., 2015). Untreated sleep symp-toms can persist for years, exacerbate existing PTSD symptoms,and complicate recovery or affect efficacy of first-line PTSDtreatment (Koffel, Khawaja, & Germain, 2016). For these reasons,sleep disturbances are considered a modifiable risk factor that, ifsufficiently treated, could reduce the overall burden and risk ofrelapse in PTSD (Germain, 2013).

Existing evidence-based treatments for PTSD, such as pro-longed exposure (PE), appear to be less effective in amelioratingsleep disturbance, relative to daytime symptoms, that is, all otherPTSD symptoms apart from nightmares and difficulty sleeping(Galovski et al., 2009; Gutner, Nillni, Suvak, Wiltsey-Stirman, &Resick, 2013; Larsen, Fleming, & Resick, 2019; Nappi, Drum-mond, & Hall, 2012; Schnurr & Lunney, 2018). One limitation ofthese studies, though, is they generally assessed sleep with onlylimited questions extracted from a larger, nonsleep focused mea-sure (Galovski et al., 2009; Gutner et al., 2013; Larsen et al., 2019;

Schnurr & Lunney, 2018) or global sleep measure rather thanPTSD-specific sleep issues (Galovski et al., 2009; Gutner et al.,2013). To our knowledge, no study has assessed sleep usingwell-validated multimodal tools to understand which sleep prob-lems are most likely to interfere with and least likely to remitthrough treatment. Therefore, this study’s first aim was to developa more nuanced understanding of how evidence-based PTSD psy-chotherapies may improve sleep symptoms.

Whereas it appears daytime PTSD interventions may not suffi-ciently treat sleep symptoms, evidence-based interventions existfor both insomnia and nightmares (Germain, 2013; Nappi et al.,2012; Talbot et al., 2014). Studies utilizing such interventionswithin PTSD populations have reported sleep symptoms improvedas expected when they were specifically targeted. Moreover, thesestudies reported some improvements in daytime symptoms ofPTSD (Germain, 2013; Nappi, Drummond, Thorp, & McQuaid,2010; Talbot et al., 2014). Overall, treatment studies suggestnormalization of sleep may have some benefit in reducing daytimePTSD symptoms. However, to our knowledge, no study has tes-ted whether combining traditional PTSD psychotherapy withevidence-based treatments for nightmares (imagery rehearsal ther-apy [IRT]) and insomnia (cognitive-behavioral therapy for insom-nia [CBT-I]) in patients with sleep complaints will increase effi-cacy of overall treatment. If the addition of sleep treatment tobest-practice PTSD therapies improves nighttime and/or overallseverity of PTSD, this may improve PTSD treatment outcomes. Asone of the first pilot studies of its nature, this will also provide keyinformation as to guide implementation of combined protocols inclinics.

The aims of this study were to address the two main gapshighlighted above: (a) to examine the impact of evidence-basedPTSD treatment (PE) on well-validated sleep assessments and (b)to assess the value of adding evidence-based sleep interventions tothe end of PTSD treatment to improve both sleep symptoms anddaytime PTSD symptoms. The study recruited Operation EnduringFreedom/Operation Iraqi Freedom/Operation New Dawn (OEF/OIF/OND veterans with comorbid diagnoses of PTSD and chronicinsomnia as well as nightmares. All participants received PE andthen were randomized to either receive validated sleep interven-tions (CBT-I � IRT) or control treatment (supportive care therapy[SCT]). For the first aim, we hypothesized small improvements insleep diary-based sleep efficiency (SE) and number of nightmaresduring PE but that sleep impairment would remain in the clinicalrange. We also included Clinical Administered PTSD Scale

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

2 WALTERS ET AL.

(CAPS) total score as a primary outcome to ensure PE improvedoverall PTSD symptoms as expected. The second aim focused onchanges in the same set of primary outcome variables when sleepinterventions were added onto PE in a sequential manner. Wehypothesized sleep diary-based SE and number of nightmareswould fall within the normal range and overall PTSD symptoms(CAPS total score) would show clinically meaningful improve-ments following sleep treatment but not control treatment. Becauseof the final sample size (see below), the second aim must beconsidered a pilot study. A number of other measures were exam-ined in both aims for exploratory purposes, to provide as full apicture of treatment effects as possible, and to address the researchgap identified related to limited sleep measures in prior studies.Over the course of the study, participant dropout was high; there-fore, as an exploratory aim, we also explored predictors of treat-ment completion.

Method

This was a parallel-groups, randomized controlled trial compar-ing a combined PTSD and sleep intervention to a PTSD plus SCTintervention. Participants were recruited throughout a grant periodbetween January 2010 and March, 2012 and randomized in equalnumbers to receive: (a) 12-times, 90-min twice-weekly sessions ofPE followed by sleep-specific treatments (5-times, 60-min weeklysessions of IRT for nightmares and 7-times, 60-min weekly ses-sions of CBT-I for insomnia); or (b) PE followed by 12-times,60-min weekly sessions of SCT. The trial was approved by theUniversity of California, San Diego, Veterans Affairs San DiegoHealth care System (VASDHS), and Naval Medical Center SanDiego Institutional Review Boards and registered on clinicaltrials.gov (NCT01009112) prior to first enrollment.

The study was single blinded, with the outcomes assessormasked to allocation. Randomization was conducted by primaryinvestigator Sean P. A. Drummond. in blocks of six using maskedallocation. Condition was revealed to the treating therapist onlyafter completion of PE. There were no important changes made tomethods or outcomes after trial commencement, nor were thereany unintended harms to participants throughout the protocol.

Participants

Study participants were 55 U.S. veterans and active-duty per-sonnel with PTSD and sleep disturbance from the VASDHS, andNaval Medical Center San Diego provided written informed con-sent. Inclusion criteria included (a) being deployed at least once aspart of Operation Enduring Freedom/Operation Iraqi Freedom/Operation New Dawn (OEF/OIF/OND), (b) having a current di-agnosis of PTSD from military-related trauma, (c) meeting criteriafor comorbid insomnia, and (d) reporting minimum two night-mares per week. Exclusion criteria included (a) unmanaged psy-chosis or manic episodes in the past year, (b) substance/alcoholdependence in the past 6 months, (c) untreated sleep disordersother than insomnia and nightmares, (d) use of hypnotics orprazosin for sleep, (e) history of traumatic brain injury other thanmild brain injury, (f) change in dosage and/or type of psychotropicmedication in preceding month, and (g) concurrent psychotherapyfor PTSD. All screening was conducted by a single licensed andexperienced psychiatric nurse. Any questions related to eligibility

were resolved through a consensus discussion among the investi-gators. For details of participant flow through the study, see Figure1. Briefly, of the 55 veterans consented, 10 did not meet eligibilitycriteria and so were excluded. The sample at baseline comprisedveterans (n � 37), active-duty personnel (n � 5), and reserves(n � 3).

Treatment Conditions

Prolonged exposure therapy. PE is a manualised treatment(Foa, Hembree, & Rothbaum, 2007). All participants were askedto complete 12-times, 90-min, twice-weekly sessions. The focus ofthe treatment is reducing avoidance related to trauma cues andincreasing quality of life by having patients take part in in vivoexposure to trauma reminders and imaginal exposure to the traumamemory. Weekly homework was assigned, which included listen-ing to audiotapes of sessions, practicing breathing skills, andparticipating in avoided but safe situations.

Imagery rehearsal therapy. After completing PE, partici-pants randomized to PE � sleep were invited to attend 5-times,60-min weekly sessions of IRT. IRT is a manualized therapy fornightmares (Forbes et al., 2003). Briefly, the core components ofIRT are (a) psychoeducation on sleep, nightmares, and imagery;(b) teaching and homework of 10-min daily practice of personal-ized pleasant imagery scenes; (c) rescripting of the nightmare,which involves identifying and elaborating upon an alternative,neutral, and/or pleasant ending, with a 10-min daily practice of thenew-dream imagery; (d) problem-solving difficulties; and (e) re-lapse prevention. Daily nightmare logs are kept and reviewed insessions. Notably, this version of IRT does not include exposure tothe nightmare, further differentiating it from PE.

Cognitive behavioral therapy for insomnia. CBT-I con-sisted of 7-times, 60-min weekly sessions (Perlis, Benson-Jungquist, Smith, & Posner, 2005). The core components arepsychoeducation on chronic insomnia, sleep restriction, stimuluscontrol, sleep hygiene, stress reduction and relaxation, decatastro-phization of consequences through cognitive restructuring, andrelapse prevention.

Supportive care therapy condition. Participants randomizedto PE � SCT received 12-times, 60-min weekly sessions of SCTafter PE. This controlled nonspecific treatment effects associatedwith therapist contact and homework completion. This nondirec-tive Rogerian therapy includes empathetic listening, clarificationof goals, and nondirective questioning. The focus was to helpparticipants better understand their emotional response to PTSDsymptoms and provide supportive, unconditional positive regard tomanaging their own PTSD symptoms. Components of IRT andCBT-I were strictly avoided.

Treatment fidelity. Treatment fidelity was maintained threeways. First, therapists were all licensed or licensed eligible psy-chologist. They were trained on each intervention through: (a)didactics; (b) treating two veterans with each intervention prior toconducting therapy for the study (training conducted by authorsSonya B. Norman [PE], Sean P. A. Drummond [CBT-I and IRT],and Carla M. Nappi [SCT]); and (c) every session included atherapist checklist of key components. Second, therapists hadweekly supervision for each treatment with the same individualswho did the training. Third, every session was audio taped andindependent judges (listed in the Acknowledgments) rated treat-

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

3PE�SLEEP TREATMENT IMPACT ON SLEEP & PTSD SYMPTOMS

ment fidelity, based on a checklist, for every session of the firsttwo veterans/intervention/therapist and then 20% of sessionsfor every subsequent veteran/intervention/therapist. Unfortu-nately, these data are not available because they were lostduring an international relocation by the senior author (SeanP. A. Drummond).

Measures

Participants were assessed for daytime and nighttime symptoms,medication use, and quality of life before treatment (screening/week 0) and at weeks 6, 11, and 18. These subsequent assessmentscoincide with the end of each active treatment. Additionally,participants completed daily sleep diary and nightmare logsthroughout the 18 treatment weeks.

Screening measures. The following tools assessed eligibilitycriteria before treatment: (a) Structured Clinical Interview forDiagnostic and Statistical Manual of Mental Disorders, fourthedition, text revision; (b) Duke Structured Interview for SleepDisorders, which assesses for sleep disorders in both Diagnosticand Statistical Manual of Mental Disorders, fourth edition, textrevision, and ICSD (International Classification of Sleep Disor-ders; Edinger et al., 2004); (c) medical history interview and chartreview; (d) medication use interview; and (e) brief screening fortraumatic brain injury as per VASDHS protocol: If a potentialparticipant screened positive, further questions were asked to de-termine severity.

Primary outcome measurement tools. A daily sleep diaryand nightmare log was completed throughout the 18-week treat-ment and collected primary outcome measures of SE and number

of nightmares. This also asked for exploratory subjective measuresof bedtime, sleep latency, number and duration of awakenings,wake time, total time in bed, total sleep time, and intensity ofnightmares. Weekly averages and night-to-night variability insleep parameters were calculated (Straus et al., 2015).

The CAPS (Blake et al., 1995) is the gold-standard semistruc-tured interview that corresponds with Diagnostic and StatisticalManual of Mental Disorders, fourth edition, criteria for PTSD.This interview was used to make a current diagnosis (past month)of PTSD at baseline and symptoms in the past week at all futureassessments. Reliability was good, sample � for subscales �.74–.87, CAPS total score � � .92.

Additional exploratory outcome measurement tools. Tosupplement our primary outcome variables, a number of othermeasures were gathered at each assessment.

The Insomnia Severity Index is a seven-item self-report assess-ment of severity of insomnia in the past week (Morin, 1993). Ahigher score indicates more sleep disturbances (sample � � .80).

The Pittsburgh Sleep Quality Index (PSQI) is a 19-item self-report assessment of sleep quality over the past month (Buysse,Reynolds, Monk, Berman, & Kupfer, 1989). Scores �5 are con-sidered clinically impaired sleep quality. The PTSD Addendum forthe PSQI assessed sleep-disruptive behaviors common to PTSDpatients and those with chronic nightmares (Germain, Hall, Kra-kow, Katherine Shear, & Buysse, 2005). Scores �4 are in theclinical range. Reliability in this sample was relatively low, PSQI� � .61, PSQI-Addendum � � .68.

Participants wore an actigraph for 7 days before treatment (week0) and then 7 days each preceding all following assessments.

254 referred to study

236 phone screened

18 never reached

121 ineligible 19 did not return calls 41 declined 55 consented and

completed baseline

4 declined 41 started PE

18 discontinued PE

*of the 10 ineligible:

5 no PTSD diagnosis

3 < 2 nightmares weekly

2 no insomnia

1 current substance/alcohol dependence

1 sleep apnea diagnosis

*of the 4 declined:

1 lack of time

1 relocating

1 not interested in treatment

1 not interested in research

23 completed PE

32 completed post

PE assessment

*may have multiple reasons

PTSD = Posttraumatic Stress Disorder;

PE = Prolonged Exposure

12 randomized to

IRT/CBTI

11 randomized to

SCT

8

completed

SCT

8

completed

IRT/CBTI

10 ineligible

*of the 18 discontinued:

5 life stressors (homelessness, financial

problems, marital distress, work stress)

4 relocation (out of state or > 50 miles from

study site)

3 no longer interested in PE specifically

2 preferred alternative treatment (marital

therapy, sleep medication and treatment for

depression)

2 did not return calls

2 schedule conflicts (work schedule, family

obligations)

1 safety concerns

1 did not cite reason

*of the 3 discontinued:

1 safety issues

1 schedule conflict

1 did not cite reason

*of the 4 discontinued:

1 did not return calls

1 schedule conflict

2 did not cite reason

3

discontinued

SCT

4

discontinued

IRT/CBTI

Figure 1. Consort chart.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

4 WALTERS ET AL.

Respironics Actiwatch 2 and Actiware software (Respironics;Murrysville, Pennsylvania) algorithms generated values for objec-tive time in bed, total sleep time, wake after sleep onset, and SE.Rest intervals for analysis were manually set, corresponding totime in bed according to diary entries, although rest intervals couldbe extended �60 min on either side to account for obvious errors(Straus et al., 2015). Sleep detection settings were set to medium.

The PTSD Checklist-Specific version (Wilkins, Lang, & Nor-man, 2011) is a self-report measure of PTSD severity. A higherscore indicates greater symptom severity. The reliability and va-lidity of the measure have been established (Wilkins et al., 2011)and in this sample � � .91.

The Patient Health Questionnaire-9 (PHQ-9; Kroenke &Spitzer, 2002) assessed severity and frequency of depressionsymptoms because of the high frequency of comorbid depressionin PTSD. It comprises nine items corresponding to Diagnostic andStatistical Manual of Mental Disorders, fourth edition, criteria fordepression, with higher scores indicating greater symptoms (sam-ple � � .87).

The Medical Outcome Study Short Form-36-V (Kazis et al.,2006) is a 36-item psychometrically based index designed tomeasure general health functioning and quality of life among

veterans. Two variables are reported here: physical and mentalcomponents. Reliability was low in this sample, � � .33.

Data Analysis

Analysis was conducted using SPSS version 24 (IBM Corp;Armonk, New York). Of the 55 veterans consented for studytreatment, 23 (42%) completed PE. Following PE, 12 were ran-domized to IRT/CBT-I and 11 randomized to SCT. Of these, eightcompleted IRT/CBT-I and eight completed SCT (total study com-pleters: n � 16 [29% of those consented, 70% of those whofinished PE]; see Figure 1). All participants were invited to attendall assessment sessions, regardless of whether they completedtreatment. This enabled an intent-to-treat analytic framework, inwhich all available data for each time point were analyzed (Hollis& Campbell, 1999).

Where there were missing data, a complete case analysis wasconducted. Some or all of the pretreatment assessment was com-pleted by n � 55 participants. At week 6, 32 participants com-pleted some or all post-PE assessment. At weeks 11 and 18, 15participants completed some or all assessment. Reasons for miss-ing data included participants declining to complete measures,

Table 1Demographic Information and Trauma Characteristics by Treatment Condition

Measure

All consented Completed PE IRT/CBTI SCT p value

(n � 55) (n � 23) (n � 12) (n � 11) (IRT/CBTI vs. SCT)

SexFemale 9 4 4 0 .035

Age (years)M � SD 35.0 � 1.4 36.54 � 9.8 36.8 � 10.4 36.0 � 9.5 .86

RaceBlack 10 7 3 4 .70Caucasian 28 12 1 6Pacific islander 4 3 2 1Asian 1 1 1 0Biracial 2 0 0 0

EthnicityNon-Hispanic 19 18 10 8 .54Hispanic 26 5 2 3

MinorityMinority 19 7 4 8 .75Nonminority 26 16 8 3

Marital statusSingle 11 5 2 3 .26Married 19 8 3 5Divorced 10 6 4 2Separated 4 3 3 0Widowed 1 1 0 1

Number of childrenM � SD 1.40 � 1.71 1.61 � 1.97 1.58 � 2.31 1.64 � 1.63 .95

Number of deployments1 22 11 7 4 .532 17 8 3 53 6 4 2 2

Months since deploymentM � SD 59.4 � 4.5 60.1 � 35.12 52.9 � 39.1 67.9 � 30.0 .32

Months since index traumaM � SD 78.0 � 8.4 84.0 � 64.8 78.4 � 68.1 90.2 � 63.7 .67

Months since symptomsM � SD 63.6 � 8.7 67.0 � 65.5 34.5 � 36.6 92.7 � 81.1 .07

Note. IRT � imagery rehearsal therapy; CBTI � cognitive behavioral therapy for insomnia; SCT � supportive care therapy; PE � prolonged exposure.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

5PE�SLEEP TREATMENT IMPACT ON SLEEP & PTSD SYMPTOMS

nonadherence to sleep diary and actigraphy, or technical issueswith actigraphy. Exact sample size for each set of measures at eachassessment is noted within Tables 1–4.

For completeness of presentation of data, results presented in-clude primary outcome measures of diary-derived SE, frequencyof nightmares, and CAPS score as well as other exploratory dataincluding retrospective sleep measures, other diary-derived vari-ables, actigraphy, and self-reported PTSD, depression, and quality-of-life measures. We will present these measures for completenesswithout discussion or interpretation.

To address aim 1 and test whether sleep improved over PE,repeated-measures t tests compared results at week 0 (pretreat-ment) and week 6 (immediately after treatment). Because of sig-nificant skew in variable distributions, nonparametric Wilcoxonsigned-ranks test compared groups for all diary-derived variablesexcept total sleep time.

To address aim 2, which is the pilot study component, analysisused t tests for change scores from the prior assessment for eachvariable. Our a priori hypothesis was that there would be a Time �Treatment Interaction and change score t tests provide a direct test

of that specific effect. Whereas a repeated-measures ANOVAwould also be appropriate, small sample size lacks sufficientpower for an ANOVA. Where significant skewness in distributionsof sleep diary variables indicated nonparametric analysis, Mann-Whitney U tests were conducted.

Because attrition was high, we decided to explore predictors oftreatment completion as an exploratory analysis to inform futureresearch and treatment design. To assess differences between studycompleters and noncompleters, we considered a PE completer anyparticipant who received all components of the protocol (Foa et al.,2007) in a minimum of 10 sessions. We used a broad range ofvariables to explore possible individual-level predictors. We con-sidered general and military-related predisposing determinants (in-dividual characteristics existing before onset of illness), enablingfactors (factors allowing access to health care services), and psy-chiatric and sleep-related factors (perception of illness and itsseverity). In addition, we examined variables related to durationsince symptom onset and trauma exposure. For each set of vari-ables, we conducted a standard logistic regression to determinewhich predictors uniquely contributed to treatment completion.

Table 2Baseline and After Prolonged Exposure Treatment Means for PTSD, Depression, and Quality of Life

Measures

Before treatment After treatment Paired-sample t test

n M � SD M � SD t or z value p ES (d)

CAPS total score 30 75.27 � 14.91 62.30 � 21.56 3.97 �.001 .70Reexperiencing 30 20.77 � 5.99 19.47 � 6.52 1.21 .24 .21Avoidance/numbing 30 29.27 � 7.43 21.47 � 11.13 4.12 �.001 .82Hyperarousal 30 25.20 � 4.64 21.37 � 6.98 3.60 .001 .65

PCL total score 32 60.47 � 8.78 46.84 � 16.74 5.33 �.001 1.02PHQ-9 32 16.28 � 5.23 11.84 � 6.07 3.57 .001 .78VR-36 physical component 28 44.93 � 11.00 43.68 � 10.75 .76 .45 .12VR-36 mental component 28 29.52 � 10.29 37.66 � 11.81 –4.79 �.001 .74

Note. CAPS � Clinical Administered PTSD Scale; PCL � PTSD Checklist; PHQ-9 � The Patient Health Questionnaire-9; VR-36 � Veterans Rand-36.

Table 3Baseline and After Prolonged Exposure Treatment Means for Prospective and Retrospective Sleep Measures

Measures

Beforetreatment

After PEtreatment

Paired-samplet tests

M � SD M � SD t or Z p ES (d)

Sleep diary (n � 25)Sleep efficiencya 71.97 � 15.05 77.34 � 12.64 –1.66 .098 .39Total number of nightmaresa 6.72 � 4.30 4.36 � 3.57 –2.76 .006 .60Total sleep time (min) 336.03 � 94.32 339.94 � 88.05 –.22 .83 .04Sleep latency (min)a 57.35 � 55.07 42.54 � 35.59 –1.06 .29 .32Wake after sleep onset (min)a 75.50 � 51.80 56.01 � 45.20 –2.46 .014 .40Highest nightmare ratinga 7.56 � 2.02 6.04 � 2.94 –1.73 .083 .60Night-to-night variability: SLa 35.54 � 30.81 26.90 � 28.42 –1.84 .065 .29Night-to-night variability: WASOa 55.02 � 38.27 45.93 � 30.51 –.44 .66 .26Night-to-night variability: TSTa 123.56 � 58.95 107.52 � 68.75 –1.19 .23 .25Night-to-night variability: SEa 15.16 � 7.72 13.20 � 8.87 –1.09 .28 .24

Retrospective sleep measuresInsomnia Severity Index (n � 32) 18.97 � 4.01 16.22 � 6.29 2.31 .028 .52Pittsburgh Sleep Quality Index (n � 29) 14.76 � 3.30 13.07 � 3.22 2.07 .048 .52Pittsburgh Sleep Quality Index-Addendum (n � 30) 9.67 � 3.62 8.47 � 3.88 1.58 .13 .32

Note. Primary outcome variables are presented in boldface. SL � sleep latency; WASO � wake after sleep onset; TST � total sleep time; SE � sleepefficiency; PE � prolonged exposure.a Used nonparametric, Wilcoxon test.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

6 WALTERS ET AL.

Tab

le4

Mea

nsan

dSt

anda

rdD

evia

tion

sat

Eac

hA

sses

smen

tP

oint

,Se

para

ted

Into

the

Evi

denc

e-B

ased

Slee

pT

reat

men

tan

dSu

ppor

tive

Car

eT

reat

men

tG

roup

s

Var

iabl

es

Supp

ortiv

eca

reth

erap

ySl

eep

trea

tmen

ts

Bef

ore

trea

tmen

tPE

com

plet

edPE

�5

wee

ksSC

TPE

�12

wee

ksSC

TB

efor

etr

eatm

ent

PEco

mpl

eted

PE�

IRT

PE�

IRT

�C

BT

-I

Wee

k0

Wee

k6

Wee

k11

Wee

k18

Wee

k0

Wee

k6

Wee

k11

Wee

k18

M�

SDM

�SD

M�

SDM

�SD

M�

SDM

�SD

M�

SDM

�SD

PTSD

,de

pres

sion

and

qual

ityof

life

n�

11n

�11

n�

8n

�8

n�

12n

�12

n�

6n

�7

CA

PSto

tal

scor

e70

.09

�15

.56

61.3

6�

19.5

0a55

.13

�23

.44

52.3

8�

18.5

2a74

.08

�13

.70

53.3

3�

19.3

8a44

.33

�19

.05a

37.5

7�

20.9

7a,b

Ree

xper

ienc

ing

19.4

5�

4.78

20.2

7�

5.62

17.3

8�

6.85

16.7

5�

6.78

19.4

2�

5.90

16.9

2�

6.13

12.8

3�

9.89

13.1

7�

8.13

Avo

idan

ce/n

umbi

ng25

.82

�9.

0918

.54

�11

.30a

14.8

8�

12.6

3a14

.00

�8.

38a

30.4

2�

6.14

17.8

3�

9.45

a14

.67

�7.

12a

12.0

0�

8.32

a

Hyp

erar

ousa

l24

.82

�4.

4022

.54

�4.

6822

.88

�5.

8421

.63

�5.

43c

24.2

5�

4.81

18.5

8�

6.13

a16

.83

�7.

22a

12.0

0�

9.14

a,b,c

.d

CA

PSw

ithou

tsl

eep

item

s56

.91

�15

.59

49.4

5�

19.0

042

.63

�22

.50

40.6

3�

17.3

9a61

.67

�12

.99

42.1

7�

18.6

6a35

.17

�16

.03

31.1

7�

19.2

6a

PCL

tota

lsc

ore

58.2

7�

9.73

45.8

0�

16.7

3e40

.75

�14

.54a

40.3

8�

14.1

7a58

.92

�9.

2739

.75

�14

.58a

33.7

1�

12.4

3a34

.43

�16

.15a

PHQ

14.9

1�

5.56

11.6

4�

6.67

10.2

5�

5.97

11.7

5�

6.98

17.0

8�

5.57

9.42

�5.

37a

6.33

�4.

59a

7.29

�5.

77a

VR

-36

phys

ical

com

pone

nt43

.71

�11

.21

43.5

8�

10.8

844

.32

�7.

9643

.20

�9.

9346

.07

�11

.24f

45.3

5�

10.2

847

.23

�7.

79g

49.2

4�

7.54

VR

-36

men

tal

com

pone

nt31

.70

�13

.26

38.2

8�

12.7

240

.33

�11

.61

42.2

7�

11.6

630

.52

�8.

20f

40.0

4�

11.2

8a45

.83

�10

.75a,

g42

.63

�11

.26a

Slee

pdi

ary

n�

11n

�10

n�

8n

�8

n�

12n

�11

n�

6n

�7

Slee

pef

fici

ency

h76

.51

�10

.07

81.7

4�

10.5

574

.70

�17

.75

74.6

1�

11.5

9c73

.33

�12

.11

78.7

7�

9.79

82.8

2�

8.29

91.2

9�

4.89

c,d

,i

Tot

alnu

mbe

rof

nigh

tmar

esh

6.64

�4.

204.

10�

3.84

a5.

88�

6.88

4.50

�6.

786.

50�

3.94

4.00

�3.

77a

3.67

�3.

272.

14�

2.73

Tot

alsl

eep

time

(min

)35

7.36

�99

.82

361.

57�

106.

8233

2.45

�13

5.32

304.

12�

103.

1032

1.42

�57

.85

324.

45�

70.4

036

5.26

�65

.68a

326.

29�

91.1

5Sl

eep

late

ncy

(min

)h55

.93

�37

.80

32.2

4�

24.1

238

.02

�17

.01

39.5

4�

18.6

635

.76

�29

.83

30.3

3�

17.5

731

.79

�18

.38

21.3

7�

19.5

2d

Wak

eaf

ter

slee

pon

set

(min

)h52

.24

�16

.60

42.7

0�

24.0

050

.58

�29

.01

51.3

1�

23.8

685

.54

�57

.10

59.2

7�

44.2

242

.15

�32

.09a,

b31

.12

�39

.01b

,d,i

,j

Hig

hest

nigh

tmar

era

tingh

8.00

�1.

345.

30�

2.98

a6.

63�

2.39

5.50

�3.

427.

17�

2.48

6.27

�3.

295.

33�

3.08

j4.

57�

3.36

Nig

ht-t

o-ni

ght

vari

abili

ty:

SLh

40.1

5�

25.0

218

.95

�14

.02a

20.8

9�

13.7

5a29

.84

�20

.18

30.1

9�

38.5

415

.90

�13

.36

15.8

8�

13.1

114

.61

�27

.23

Nig

ht-t

o-ni

ght

vari

abili

ty:

WA

SOh

41.2

1�

24.1

634

.92

�31

.64

40.1

5�

25.4

134

.29

�19

.15

72.2

1�

47.4

346

.26

�27

.06

25.7

1�

16.9

120

.67

�12

.35a,

b

Nig

ht-t

o-ni

ght

vari

abili

ty:

TST

h91

.72

�36

.32

102.

49�

51.0

311

1.45

�69

.03

115.

91�

84.9

713

3.55

�64

.37

82.0

6�

37.5

5a65

.42

�29

.66

57.5

4�

43.1

4a

Nig

ht-t

o-ni

ght

vari

abili

ty:

SEh

12.2

6�

7.67

10.2

8�

10.1

114

.43

�14

.52

14.8

8�

4.98

16.9

5�

6.75

12.4

5�

5.23

7.86

�5.

20a

8.02

�8.

40

Act

igra

phy

n�

9n

�8

n�

6n

�6

n�

9n

�10

n�

7n

�7

Tot

alsl

eep

time

(min

)33

0.65

�39

.70

344.

56�

49.7

433

9.89

�25

.63

350.

19�

37.9

036

5.21

�62

.08

334.

06�

44.6

635

4.94

�65

.24

283.

75�

65.2

0W

ake

afte

rsl

eep

onse

t55

.18

�29

.12

55.0

0�

25.7

847

.31

�21

.92

57.9

4�

30.4

644

.22

�9.

0739

.58

�16

.44

43.3

2�

18.6

153

.76

�30

.02

Slee

pef

fici

ency

68.0

3�

8.69

85.8

9�

6.67

a75

.68

�5.

99b

75.4

5�

7.11

b75

.48

�11

.63

89.5

4�

3.90

a76

.35

�9.

49b

69.5

7�

15.9

6b

Tim

ein

bed

385.

83�

58.4

340

2.77

�60

.43

387.

20�

36.1

840

8.13

�61

.47c

410.

30�

60.5

937

3.64

�51

.55

398.

26�

62.0

333

7.56

�48

.37c

Ret

rosp

ectiv

esl

eep

mea

sure

sn

�11

n�

11n

�8

n�

8n

�12

n�

12n

�5

n�

7In

som

nia

Seve

rity

Inde

x18

.73

�4.

1016

.36

�5.

5418

.50

�5.

35d

16.7

5�

6.02

18.7

5�

3.98

14.4

2�

5.84

a9.

20�

6.26

c,j

9.71

�7.

13a,

b,d

Pitts

burg

hSl

eep

Qua

lity

Inde

x15

.09

�3.

7313

.36

�2.

0614

.75

�1.

67c

13.1

3�

3.31

c14

.75

�3.

0511

.50

�2.

58a

9.60

�4.

04c,

j7.

71�

4.50

a,b,c

,d

Pitts

burg

hSl

eep

Qua

lity

Inde

x–A

dden

dum

10.8

2�

2.86

8.00

�3.

66a

8.75

�4.

4014

.38

�5.

76a,

b,c

8.42

�3.

927.

92�

4.19

4.60

�4.

39b

8.29

�5.

59

Not

e.Pr

imar

you

tcom

eva

riab

les

are

pres

ente

din

bold

face

.V

R-3

6�

Vet

eran

sR

and-

36;

SL�

slee

pla

tenc

y;W

ASO

�w

ake

afte

rsl

eep

onse

t;T

ST�

tota

lsl

eep

time;

IRT

�im

ager

yre

hear

sal

ther

apy;

CB

T-I

�co

gniti

vebe

havi

oral

ther

apy

for

inso

mni

a;SC

T�

supp

ortiv

eca

reth

erap

y;PE

�pr

olon

ged

expo

sure

;PC

L�

PTSD

chec

klis

t;PH

Q�

The

Patie

ntH

ealth

Que

stio

nnai

re.

aD

iffe

rent

from

wee

k0

ofsa

me

cond

ition

.b

Dif

fere

ntfr

omw

eek

6m

easu

reof

sam

eco

nditi

on.

cD

iffe

rent

from

sam

ew

eek

ofdi

ffer

ent

cond

ition

.d

Dif

fere

ntfr

omw

eek

11of

sam

eco

nditi

on.

en

�10

.fn

�11

.g

n�

6;al

lsi

gnif

ican

cede

sign

atio

nba

sed

onp

�.0

5.h

Use

dno

npar

amet

ric,

Man

n-W

hitn

eyU

,an

dW

ilcox

onsi

gned

-ran

kte

st.

iC

hang

efr

omw

eek

11is

sign

ific

antly

grea

ter

than

SCT

.jC

hang

efr

omw

eek

6is

sign

ific

antly

grea

ter

than

SCT

.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

7PE�SLEEP TREATMENT IMPACT ON SLEEP & PTSD SYMPTOMS

We then tested an overall model containing all significant predic-tors from the initial regressions, again using standard logisticregression. Standard regression was selected because of lack oftheory to generate hypotheses about relative importance needed forhierarchical regression.

To assess whether there were differences in individuals whocompleted the sleep portion of the protocol, the same predictors asthose assessing PE completion were assessed as well as PTSDnighttime and daytime symptoms after PE completion. Post-PEdiary-based SE was dropped due to collinearity (r � .7) with othersleep variables.

To assist interpretation of effects for the pilot component of thisstudy, effect sizes were calculated and reported.

Results

Table 1 provides a detailed characterization of the sample. Table2 presents data demonstrating the efficacy of PE on global PTSDsymptoms (clinical and self-rated), depression, and quality of life,with medium to large effect sizes.

Aim 1: Effect of PE on Sleep (See Table 3)

There was no significant effect of PE on SE (mean changescore � 5.37%, z � 1.66, p � .098, d � .39). Participants didreport significantly fewer weekly nightmares after PE (mean de-crease � 2.36 nightmares, z � 2.76, p � .006, d � .60).However, each of these measures remained in the clinical range forthe majority of participants. After PE, 80% (n � 20) continued toexperience minimum two nightmares per week and 76% (n � 19)had SE �85%. Figure 2 displays results of paired-samples t testsbefore and after PE.

Aim 2: Effect of Evidence-Based Sleep Treatments onPrimary Outcomes

Table 4 displays primary and exploratory outcome variables foreach group at week 6 (all participants had completed PE), week 11(after SCT or IRT), and week 18 (after SCT or CBT-I). FollowingPE treatment, participants in the SCT and IRT/CBT-I groups didnot significantly differ on any primary outcome variable, all p �.05. Figure 3 displays results of independent samples t tests be-tween sleep and control conditions.

Relative to the end of PE (week 6), IRT increased diary-derivedSE with nonsignificant but medium-large effect size (U � 10.00,p � .11, d � .64). Nightmare frequency decreased with largeeffect size but also did not meet statistical significance, t(11) �1.76, p � .10, d � .96. CAPS scores did not change over IRTdifferently according to condition, t(12) � 0.15, p � .88, d � .08.

Relative to week 11 (end of IRT or 5 weeks of SCT), CBT-Isignificantly increased SE compared with SCT, t(12) � 2.21,p � .04, d � 1.25. There was no significant change in nightmarefrequency according to treatment condition (U � 21.00, p � .68,d � .01). Compared with SCT, CBT-I showed a small treatmenteffect on CAPS total scores, t(12) � 0.66, p � .51, d � .36.

In examining the combined effects of both sleep treatments,relative to SCT, (i.e., week 6 vs. week 18), IRT/CBT-I trendedtoward improved SE with large effect size, t(12) � 2.00, p �.068, d � 1.07. Nightmare frequency showed large treatmenteffects but did not reach statistical significance, t(12) � 1.68, p �.11, d � .90. CAPS total scores showed small treatment effects,t(13) � 0.61, p � .54, d � .31.

Exploratory Analysis: Predictors of TreatmentCompletion

There were no demographic differences between those whostarted and those who did not start treatment. Reasons for reporteddiscontinuation and are presented in Figure 1. A majority ofparticipants who discontinued PE (n � 12; 66%) did so in the firsthalf of treatment (before Session 7). Logistic regression analysesrevealed baseline PHQ-9 total score and night-to-night variabilityof total sleep time uniquely predicted PE treatment completion;2(5) � 22.61, p � .001. For each 1-point increase in PHQ-9, theodds of completing PE treatment increased by 31.8%. The moreconsistent one’s total sleep time was night to night, the more likelyone was to complete PE treatment. All participants who completedPE went on to attend a minimum of one session of the sleep/supportive care treatment. Analyses assessing post-PE predictorsof full treatment completion revealed that in combination, a suiteof sleep predictors (diary-based total sleep time, night-to-nightvariability of total sleep time, night-to-night variability of SE, andInsomnia Severity Index) predicted full treatment completion;2(5) � 19.31, p � .001. Night-to-night variability in total sleeptime uniquely predicted completion such that the more consistent

Figure 2. M � SD scores at weeks 0 (baseline) and 6 (after PE). Arrows indicate direction of improvement.Dotted lines indicate clinical thresholds used: for CAPS, moderately severe (60) and probably PTSD (40); sleepefficiency, 85%; number of nightmares, two per week. � p � .05. ��� p � .001. See the online article for the colorversion of this figure.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

8 WALTERS ET AL.

one’s total sleep time, the more likely one was to complete allphases of treatment.

Discussion

The present study examined the impact of PE on well-validatedsubjective measures of sleep and then piloted the potential benefitof adding sleep-specific treatments (i.e., IRT and CBT-I) followingPE to explore whether sleep interventions would enhance night-time and/or daytime treatment outcomes. As expected, PE wasefficacious for improving day- but not nighttime PTSD symptoms,whereas a combination of PE and sleep-specific treatment showedpreliminary evidence of being efficacious at improving both out-comes. Thus, this study adds to the evidence that a comprehensiveapproach to PTSD treatment should address both nighttime anddaytime symptoms of the disorder.

Prolonged Exposure Improves Day but Not NighttimeSymptoms of PTSD

PE was, as expected, effective for treating daytime symptoms,specifically global PTSD severity, quality of life, and depression(Table 2, Figure 2). However, PE did not significantly improve SE,and participants still experienced clinically significant sleep im-pairment in both SE and nightmare frequency. This replicatesprevious findings, which showed sleep symptoms were residualafter PE (Gutner et al., 2013; Pruiksma et al., 2016; Zayfert et al.,2004). Despite previous literature consistently demonstrating im-proved global PTSD symptoms with similar large effect sizes asthe present data, sleep difficulties (self-reported duration or qual-ity) were consistently maintained in the clinical range after behav-ioral treatment. Our findings of 75–80% of individuals fallingwithin clinical insomnia/nightmare ranges echoed a recent reportby Schnurr & Lunney, 2018 showing over half their sample stillreported sleep difficulties after PE. Overall, the conclusion drawnis although PE is clearly effective for treating daytime symptomsof PTSD, it is insufficient for treating sleep-specific nighttimesymptoms.

Importantly, this finding also supports the notion of impairedsleep being a core feature of PTSD (Germain, 2013; Spoormakeret al., 2008). Historically, sleep disruption was considered a sec-ondary symptom of the underlying disorder. If this were the case,sleep-related symptoms should be alleviated once the underlyingdisorder is treated. The fact this does not occur is concerning fortwo reasons: (a) sleep impairment is highly prevalent in patientsdiagnosed with PTSD (Germain, 2013); and (b) it suggests thegold-standard PTSD treatments need to be augmented to providemaximal relief of the 24-hr symptom profile in PTSD.

PE With Sleep-Specific Treatment Improves Nighttimeand Daytime Symptoms of PTSD

A critical finding of the second part of this study, representinga pilot study into efficacy of adding evidence-based sleep treat-ments onto PE, is that nighttime and, to a more modest extent,daytime symptoms of PTSD continued to improve among partic-ipants who received a sleep-specific therapy following PE (i.e.,IRT � CBT-I; Figure 3). Specifically, the first sleep treatment,IRT, led to large improvements in nightmare frequency as well asdaily sleep diary sleep efficiency. This is consistent with otherliterature showing the efficacy of IRT (Forbes et al., 2003; Nappiet al., 2010). Following completion of CBT-I, participants dem-onstrated further improvement in nighttime PTSD symptoms. In-deed, at the conclusion of all treatments, participants who receivedCBT-I did not, on average, meet clinical insomnia thresholds (i.e.,SE �85%). Participants who received SCT, on the other hand, hadfinal scores well into the clinical range (SE � 74.6%). Thisdemonstrates sleep disturbances do not resolve naturally with time.Additionally, effect sizes observed for SE (d � 1.07) and night-mare frequency (d � .90) are substantially larger than moderatelysized sleep improvements previously recorded both in our studyand previous research over the course of PE or other gold-standardPTSD treatment (Gutner et al., 2013; Schnurr & Lunney, 2018).Moreover, CBT-I did not lead to further improvements in night-mare symptoms. Thus, these preliminary findings support includ-

Figure 3. M � SD scores at each assessment: weeks 0 (pretreatment), 6 (after PE), 11 (after IRT or 5 weeksSCT), and 18 (after IRT/CBT-I or 12 weeks SCT) for the evidence-based sleep treatment (blue/light gray) andSCT (red/dark gray) groups. For SCT, n � 11 (weeks 0 and 6) and n � 8 (weeks 11 and 18). For evidence-basedsleep treatments, n � 12 (weeks 0 and 6), n � 6 (week 11), and n � 7 (week 18). Arrows indicate direction ofimprovement. Dotted lines indicate clinical thresholds used: for CAPS, moderately severe (60) and probablyPTSD (40); sleep efficiency, 85%; number of nightmares, two per week. � p � .05. See the online article for thecolor version of this figure.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

9PE�SLEEP TREATMENT IMPACT ON SLEEP & PTSD SYMPTOMS

ing a combination of nightmare specific treatment with CBT-I intreatment of patients with PTSD (Germain, 2013).

Interestingly, the present data also reveal improvement in day-time symptoms following the combined sleep interventions. Spe-cifically, we saw improvement in CAPS scores following sleeptreatment, relative to SCT, when compared with scores after PE(Table 3, Figure 3). The final assessment for the sleep treatmentgroup was the only time point participants scored below the CAPSclinical cutoff. This finding illustrates the critical role sleep likelyplays in maintenance of and recovery from PTSD. Sleep has arestorative function, is vital for emotional processing (Walker &van Der Helm, 2009), and relates to fear processes known to beabnormal in PTSD (Marshall, Acheson, Risbrough, Straus, &Drummond, 2014; Straus, Acheson, Risbrough, & Drummond,2017). Given this, individuals with nightmares, insomnia, andPTSD could be more vulnerable to trauma cues and negative affectand thus more likely to engage in avoidance behaviors, comparedwith an equivalent well-rested individual with PTSD. Therefore, itis possible improvements in daytime symptoms of PTSD could bedue to enhanced coping or improved extinction processes resultingfrom improved sleep. The relationship between disrupted sleep andimpaired emotion regulation and fear processes also argues forstudies examining whether treating sleep prior to daytime PTSDsymptoms (i.e., reversing the order of treatment presentation fromthis study design) might also be valuable. Whereas we elected totrial a protocol that most closely resembles a typical PTSD treat-ment pathway, it is certainly important to examine benefits ofreceiving sleep treatments first, followed by PE.

This study exclusively enrolled veterans of OEF/OIF/OND op-erations and generalization beyond that cohort should be done withcaution, especially given study limitation described below. Spec-ulatively, though, one would expect the results related to residualsleep symptoms after PE to apply to veterans of other eras andnonveterans, given the consistency of this finding in the literature.Results from the sleep interventions could be more mixed, though.CBT-I has proven efficacious in all age groups as well as specif-ically in PTSD samples (Talbot et al., 2014), suggesting thosefindings would likely generalize widely. On the other hand,whereas IRT appears to be effective compared with no treatmentor very nonspecific controls in mixed eras of veterans (Nappi et al.,2010) and in civilians (Ellis, Rufino, & Nadorff, 2017), IRT wasnot at all effective in a large sample of Vietnam veterans (n �124), when compared with an active control condition (Cook et al.,2010). Thus, it is unclear to what extent those specific findingswould generalize.

It is important to acknowledge limitations within this study. Asis common in PTSD treatment studies, retention was challengingand attrition high: Our dropout from PE was 43% (n � 18).Attrition continued within this protocol after PE, throughout thesleep treatments and SCT. Our dropout during the second phase ofthis protocol was 30% (four discontinued sleep treatments, threeSCT). A primary reason for attrition of this sample was timecommitment: This was a lengthy 18-week protocol. As life eventsarose throughout, participation waned. This resulted in the secondaim of this study presenting a small sample size that must beconsidered only pilot data for this type of protocol. The study wasdesigned to present each intervention in one of its most common,complete forms. In clinics, many may not choose to engage in an18-week treatment, and of those who do, many more may not

make it through. The potential benefits we demonstrated however,highlight need to test ways to shorten the combined intervention.For example, Colvonen, Drummond, Angkaw, and Norman (2018)recently piloted a protocol in which CBT-I and PE were deliveredmostly simultaneously over 15 weeks. Whereas this does notreduce the length much, it is a step in the right direction. Addi-tionally, the final sample size led to limited statistical power forour analysis in Aim 2. A post hoc power analysis based on effectsizes observed in Aim 2 (change from week 6 to week 18) revealedwe would have needed the following sample sizes to obtain sta-tistical power at the .80 level: SE, n � 18, frequency of night-mares, n � 48. Current sample size affects generalisability of thesefindings, as does a lack of follow-up data. Finally, a key limitationof this study was data loss. A large amount of actigraphy data waslost because of technical problems and participant adherence. As aresult, actigraphy sample size was lower than ideal and data notanalyzed as a primary outcome. Similarly, treatment fidelity rat-ings were lost, and thus, we cannot confirm the extent to whichtreatments protocols were strictly followed. Although we tookseveral steps to maintain good fidelity (see Method), poor fidelitywould have reduced treatment effects and/or reduced differencesbetween sleep interventions and SCT.

Overall, this study showed sleep did not improve in a clinicallymeaningful manner throughout PE. Treating nighttime symptomsof PTSD with validated, targeted treatments improved symptomsmore than SCT. Current findings coupled with other evidencesuggest IRT and CBT-I lead to significant sleep improvement inpatients with insomnia and PTSD and argue that insomnia andnightmare symptoms should be specifically assessed followingcompletion of PE. For those who experience residual sleep symp-toms after PE, sleep treatments may be effective for improvingboth nighttime and daytime symptoms.

References

Blake, D. D., Weathers, F. W., Nagy, L. M., Kaloupek, D. G., Gusman,F. D., Charney, D. S., & Keane, T. M. (1995). The development of aClinician-Administered PTSD Scale. Journal of Traumatic Stress, 8,75–90. http://dx.doi.org/10.1002/jts.2490080106

Buysse, D. J., Reynolds, C. F., III, Monk, T. H., Berman, S. R., & Kupfer,D. J. (1989). The Pittsburgh Sleep Quality Index: A new instrument forpsychiatric practice and research. Psychiatry Research, 28, 193–213.http://dx.doi.org/10.1016/0165-1781(89)90047-4

Colvonen, P. J., Drummond, S. P. A., Angkaw, A. C., & Norman, S. B.(2018). Piloting cognitive-behavioral therapy for insomnia integratedwith prolonged exposure. Psychological Trauma: Theory, Research,Practice, and Policy, 11, 107–113.

Cook, J. M., Harb, G. C., Gehrman, P. R., Cary, M. S., Gamble, G. M.,Forbes, D., & Ross, R. J. (2010). Imagery rehearsal for posttraumaticnightmares: A randomized controlled trial. Journal of Traumatic Stress,23, 553–563. http://dx.doi.org/10.1002/jts.20569

Cox, R. C., & Olatunji, B. O. (2016). A systematic review of sleepdisturbance in anxiety and related disorders. Journal of Anxiety Disor-ders, 37, 104–129. http://dx.doi.org/10.1016/j.janxdis.2015.12.001

Edinger, J., Kirby, A., Lineberger, M., Loiselle, M., Wohlgemuth, W., &Means, M. (2004). The Duke Structured Interview for Sleep Disorders.Durham, NC: University Medical Center.

Ellis, T. E., Rufino, K. A., & Nadorff, M. R. (2017). Treatment ofnightmares in psychiatric inpatients with imagery rehearsal therapy: Anopen trial and case series. Behavioral Sleep Medicine, 17, 1–14.

Foa, E., Hembree, E., & Rothbaum, B. (2007). Prolonged exposure therapyfor PTSD: Emotional processing of traumatic experiences therapist

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

10 WALTERS ET AL.

guide (treatments that work). New York, NY: Oxford University Press.http://dx.doi.org/10.1093/med:psych/9780195308501.001.0001

Forbes, D., Phelps, A. J., McHugh, A. F., Debenham, P., Hopwood, M., &Creamer, M. (2003). Imagery rehearsal in the treatment of posttraumaticnightmares in Australian veterans with chronic combat-related PTSD:12-month follow-up data. Journal of Traumatic Stress, 16, 509–513.http://dx.doi.org/10.1023/A:1025718830026

Galovski, T. E., Monson, C., Bruce, S. E., & Resick, P. A. (2009). Doescognitive-behavioral therapy for PTSD improve perceived health andsleep impairment? Journal of Traumatic Stress, 22, 197–204. http://dx.doi.org/10.1002/jts.20418

Germain, A. (2013). Sleep disturbances as the hallmark of PTSD: Whereare we now? American Journal of Psychiatry, 170, 372–382. http://dx.doi.org/10.1176/appi.ajp.2012.12040432

Germain, A., Hall, M., Krakow, B., Katherine Shear, M., & Buysse, D. J.(2005). A brief sleep scale for posttraumatic stress disorder: PittsburghSleep Quality Index Addendum for PTSD. Journal of Anxiety Disorders,19, 233–244. http://dx.doi.org/10.1016/j.janxdis.2004.02.001

Giosan, C., Malta, L. S., Wyka, K., Jayasinghe, N., Evans, S., Difede, J.,& Avram, E. (2015). Sleep disturbance, disability, and posttraumaticstress disorder in utility workers. Journal of Clinical Psychology, 71,72–84. http://dx.doi.org/10.1002/jclp.22116

Gutner, C. A., Nillni, Y. I., Suvak, M., Wiltsey-Stirman, S., & Resick, P. A.(2013). Longitudinal course of anxiety sensitivity and PTSD symptomsin cognitive-behavioral therapies for PTSD. Journal of Anxiety Disor-ders, 27, 728–734. http://dx.doi.org/10.1016/j.janxdis.2013.09.010

Hoge, C. W., Castro, C. A., Messer, S. C., McGurk, D., Cotting, D. I., &Koffman, R. L. (2004). Combat duty in Iraq and Afghanistan, mentalhealth problems, and barriers to care. New England Journal of Medicine,351, 13–22. http://dx.doi.org/10.1056/NEJMoa040603

Hollis, S., & Campbell, F. (1999). What is meant by intention to treatanalysis? Survey of published randomised controlled trials. British Med-ical Journal, 319, 670–674. http://dx.doi.org/10.1136/bmj.319.7211.670

Kazis, L. E., Selim, A., Rogers, W., Ren, X. S., Lee, A., & Miller, D. R.(2006). Dissemination of methods and results from the veterans healthstudy: Final comments and implications for future monitoring strategieswithin and outside the veterans healthcare system. Journal of Ambula-tory Care Management, 29, 310 –319. http://dx.doi.org/10.1097/00004479-200610000-00007

Kessler, R. C. (2000). Posttraumatic stress disorder: The burden to theindividual and to society. Journal of Clinical Psychiatry, 61(Suppl. 5),4–12�13–14.

Koffel, E., Khawaja, I. S., & Germain, A. (2016). Sleep disturbances inposttraumatic stress disorder: Updated review and implications for treat-ment. Psychiatric Annals, 46, 173–176. http://dx.doi.org/10.3928/00485713-20160125-01

Kroenke, K., & Spitzer, R. L. (2002). The PHQ-9: A new depressiondiagnostic and severity measure. Psychiatric Annals, 32, 509–515.http://dx.doi.org/10.3928/0048-5713-20020901-06

Larsen, S. E., Fleming, C. J. E., & Resick, P. A. (2019). Residual symp-toms following empirically supported treatment for PTSD. Psychologi-cal Trauma: Theory, Research, Practice, and Policy, 11, 207–215.http://dx.doi.org/10.1037/tra0000384

Maher, M. J., Rego, S. A., & Asnis, G. M. (2006). Sleep disturbances inpatients with post-traumatic stress disorder: Epidemiology, impact andapproaches to management. CNS Drugs, 20, 567–590. http://dx.doi.org/10.2165/00023210-200620070-00003

Marshall, A. J., Acheson, D. T., Risbrough, V. B., Straus, L. D., &Drummond, S. P. (2014). Fear conditioning, safety learning, andsleep in humans. Journal of Neuroscience: The Official Journal ofthe Society for Neuroscience, 34, 11754 –11760. http://dx.doi.org/10.1523/JNEUROSCI.0478-14.2014

Morin, C. M. (1993). Insomnia: Psychological assessment and manage-ment. New York, NY: Guilford Press.

Nappi, C. M., Drummond, S. P., & Hall, J. M. (2012). Treating nightmaresand insomnia in posttraumatic stress disorder: A review of currentevidence. Neuropharmacology, 62, 576–585. http://dx.doi.org/10.1016/j.neuropharm.2011.02.029

Nappi, C. M., Drummond, S. P., Thorp, S. R., & McQuaid, J. R. (2010).Effectiveness of imagery rehearsal therapy for the treatment of combat-related nightmares in veterans. Behavior Therapy, 41, 237–244. http://dx.doi.org/10.1016/j.beth.2009.03.003

Perlis, M. L., Benson-Jungquist, C., Smith, M. T., & Posner, D. A. (2005).Cognitive behavioral treatment of insomnia: A session-by-session guide.New York, NY: Springer.

Pruiksma, K. E., Taylor, D. J., Wachen, J. S., Mintz, J., Young-McCaughan, S., Peterson, A. L., . . . Resick, P. A. (2016). Residual sleepdisturbances following PTSD treatment in active duty military person-nel. Psychological Trauma: Theory, Research, Practice, and Policy, 8,697–701. http://dx.doi.org/10.1037/tra0000150

Schnurr, P. P., & Lunney, C. A. (2018). Residual symptoms followingprolonged exposure and present-centered therapy for PTSD in femaleveterans and soldiers. Depression and Anxiety, 36, 162–169.

Spoormaker, V. I., & Montgomery, P. (2008). Disturbed sleep in post-traumatic stress disorder: Secondary symptom or core feature? SleepMedicine Reviews, 12, 169–184. http://dx.doi.org/10.1016/j.smrv.2007.08.008

Straus, L. D., Acheson, D. T., Risbrough, V. B., & Drummond, S. P. A.(2017). Sleep deprivation disrupts recall of conditioned fear extinction.Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, 2,123–129. http://dx.doi.org/10.1016/j.bpsc.2016.05.004

Straus, L. D., Drummond, S. P. A., Nappi, C. M., Jenkins, M. M., &Norman, S. B. (2015). Sleep variability in military-related PTSD: Acomparison to primary insomnia and healthy controls. Journal of Trau-matic Stress, 28, 8–16. http://dx.doi.org/10.1002/jts.21982

Talbot, L. S., Maguen, S., Metzler, T. J., Schmitz, M., McCaslin, S. E.,Richards, A., . . . Neylan, T. C. (2014). Cognitive behavioral therapy forinsomnia in posttraumatic stress disorder: A randomized controlled trial.Sleep (Basel), 37, 327–341. http://dx.doi.org/10.5665/sleep.3408

Walker, M. P., & van der Helm, E. (2009). Overnight therapy? The role ofsleep in emotional brain processing. Psychological Bulletin, 135, 731–748. http://dx.doi.org/10.1037/a0016570

Wilkins, K. C., Lang, A. J., & Norman, S. B. (2011). Synthesis of thepsychometric properties of the PTSD checklist (PCL) military, civilian,and specific versions. Depression and Anxiety, 28, 596–606. http://dx.doi.org/10.1002/da.20837

Zayfert, C., & DeViva, J. C. (2004). Residual insomnia following cognitivebehavioral therapy for PTSD. Journal of Traumatic Stress, 17, 69–73.http://dx.doi.org/10.1023/B:JOTS.0000014679.31799.e7

Received February 18, 2019Revision received May 8, 2019

Accepted May 20, 2019 �

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

11PE�SLEEP TREATMENT IMPACT ON SLEEP & PTSD SYMPTOMS