Long-term effectiveness of cognitive–behavioural self-help intervention for nightmares


Jaap Lancee, MSc, Utrecht University, Department of Clinical and Health Psychology, PO Box 80.140, 3508TC Utrecht, the Netherlands. Tel.: +31 30 253 2387; fax: +31 30 253 4718; e-mail: J.Lancee@uu.nl


Nightmares are a prevalent disorder leading to daily impairments. Two cognitive–behavioural self-help interventions – imagery rehearsal and exposure – recently showed short-term efficacy compared to a waiting-list and a group that recorded their nightmares. This paper reports the long-term results of the imagery rehearsal (= 103) and exposure (= 95) interventions. Participants were assigned randomly to a condition after completing baseline measurements; they received a 6-week self-help intervention and completed questionnaires 4, 16 and 42 weeks after end of treatment. Initial effects on nightmare measures were almost completely sustained after 42 weeks (= 0.50–0.70); no differences were found between exposure and imagery rehearsal therapy. These results suggest that nightmares should be targeted specifically and that an internet-delivered self-help intervention seems to be a good first option in a stepped-care model.


Nightmares are a common disorder with a prevalence of around 2–5% (Bixler et al., 1979; Schredl, 2010). Nightmares cause distress and are associated with other mental complaints (Spoormaker et al., 2006). The high prevalence and impact has resulted in several treatment outcome studies, whereby cognitive behavioural treatments (CBT) such as imagery rehearsal therapy (IRT) and exposure are the treatments of choice for nightmares (Wittmann et al., 2007). Exposure instructs participants to exercise with nightmare imagery during the day; IRT is similar, but uses changed nightmare imagery instead (e.g. more positive ending of the nightmare). Both IRT and exposure gained empirical support in several trials (e.g. Burgess et al. 1998; Krakow et al., 1995).

In a recent randomized controlled trial (RCT) we compared IRT and exposure self-help interventions to a waiting-list group and a simpler form of exposure that does not directly involve exercises with nightmare imagery: recording of one’s nightmares (Lancee et al., 2010). A self-help intervention was used because it controls for possible therapist effects which could influence the efficacy of the treatments (individual therapy was found to be more effective than group therapy for nightmares –Spoormaker and van den Bout, 2006). Moreover, self-help encompasses a unique option to deliver inexpensive treatment to a large public, thereby overcoming such limitations as unavailability of cognitive behavioural therapists trained in sleep medicine (Andersson, 2009). In this trial, IRT and exposure were equally effective in ameliorating nightmares compared to recording (which was effective, but to a lesser extent) and waiting-list group (Lancee et al., 2010). However, the follow-up period of the self-help study after the intervention was relatively short (4 weeks after end of treatment). This paper discusses the long-term effectiveness of IRT and exposure.



Participants were recruited through a Dutch popular-scientific nightmare website (http://www.nachtmerries.org) and via general media (all four large national newspapers published articles about nightmares with a link to the website). Dutch internet penetration is the highest in Europe, with 83% of the Dutch households having access to the internet in 2007 (Statistics Netherlands, 2009). Inclusion criteria were: being aged 18 years or older and having self-reported nightmares based on the SLEEP-50 that has been specifically validated for nightmares (Spoormaker et al., 2005). Exclusion criteria were: extreme score on post-traumatic complaints [score > 53 on Dutch translation of the Impact of Event Scale – IES; Brom and Kleber, 1985; score > 53 is 2× standard deviations (SD) above population mean; cut-off score to indicate post-traumatic stress disorder (PTSD) is 26], currently in treatment for PTSD, suicidal ideation and schizophrenia. One hundred and ninety-eight participants with a mean age of 39.31 (SD = 15.21), including 159 women (80.3%), were randomized in IRT (= 103) or exposure (= 95; Table 1, Fig. 1). Four participants in the IRT condition were considered outliers (Z-score above 3.29 on nightmare frequency). Of the remaining 194 participants, 133 (68.6%) had a self-reported trauma, 32 (16.5%) were in psychological treatment and 40 were on medication [20.6%; mainly selective serotonin reuptake inhibitors (SSRIs)]. Seventy-seven (34.5%) participants completed all three follow-up assessments (4, 16 and 42 weeks after end of treatment).

Table 1.   Overview of the interventions
  1. IRT, imagery rehearsal therapy. *Participants kept a diary during the whole treatment; participants were advised to keep practising the relaxation exercises.

Week 1Psychoeducation about nightmaresPsychoeducation about nightmares
 Explanation of IRTExplanation exposure
 Writing down the nightmaresWriting down the nightmares
Week 2Thinking about origin of nightmareProgressive muscle relaxation
 Mental relaxation with a ‘safe place’Reading about and practising exercises
Week 3Progressive muscle relaxationExercising with nightmare imagery (10–15 min a day) in original format
 Create a new ending of the nightmare
 Exercising with nightmare imagery (10–15 min a day) in changed format
Week 4Exercising with nightmare imagery (10–15 min a day) in changed formatExercising with nightmare imagery (10–15 min a day) in original format
Week 5Exercising with nightmare imagery (10–15 min a day) in changed formatExercising with nightmare imagery (10–15 min a day) in original format
Week 6Only diaryOnly diary
Figure 1.

 Flowchart. *Z-score on nightmare frequency above 3.29 (25 or more nightmares per week); for flowchart of recording and waiting-list group see supporting Fig. S1.


The original study was approved by the Medical Ethics Review Committee of the Medical Centre of Utrecht University and was registered at http://www.clinicaltrials.gov (ID: NCT00513045). Participants received access to the digital baseline questionnaire after informed consent was provided. Participants were subsequently randomized into one of four conditions: IRT (= 103); exposure (= 95); recording group (= 105); and waiting-list group (= 95). For a description of the short-term results see Lancee et al. (2010). Eleven weeks after completion of the baseline measurements (4 weeks after intervention) participants in all conditions completed the online post-test. After completion of the post-test, participants in the two control conditions (recording group; waiting-list group) were sent an IRT intervention to fulfil ethical obligations. No data are thus available on the long-term effects of recording or waiting-list groups. Participants completed follow-up measurements 16 and 42 weeks after the end of treatment. Participants were considered dropouts after three unanswered reminders (two e-mails, one postal).


The SLEEP-50 (Spoormaker et al., 2005) was used to assess nightmare frequency for the past week, the number of nights with nightmares per month and the sleep rating (1, ‘very bad’ to 10, ‘very good’). This questionnaire has good reliability (0.85, test–retest reliability 0.78). For nightmares, the sensitivity was found to be 0.84 and the specificity 0.77. Moreover, six items (range 6–24) of the sleep impact subscale were used for a nightmare distress scale targeting the last 7 days. In our study the sleep impact subscale was preceded by: ‘Because of my nightmares…’ (e.g. I am told that I am easily irritated’). Anxiety was measured by the Dutch version of the 20-item Spielberger Trait Anxiety Inventory (Van der Ploeg et al., 1980). Depression was measured by a Dutch translation of the 20-item Centre of Epidemiological Studies–Depression scale (CES-D; Bouma et al., 1995). Post-traumatic stress complaints were measured by a Dutch translation of the 15-item IES (Brom and Kleber, 1985). Only participants who experienced a trauma completed this questionnaire.


For this study the IRT treatment (Krakow and Zadra, 2006) was transformed to a self-help treatment booklet (50 pages). The exposure treatment booklet (48 pages) was based on the self-help book used in the study by Burgess et al. (1998). The key components of the 6-week step-by-step programme are depicted in Table 1. Both booklets offered information on the fact that treatments had proved effective in earlier trials. Moreover, participants were warned about possible complications with post-traumatic nightmares. Information was provided on common difficulties with the imagery exercises in the troubleshooting section (e.g. too much anxiety; no changes in the nightmare).

Statistical analysis

Multi-level regression was used to evaluate the within-group (time) and between-group (time × condition) effects (Hox, 2002). Multi-level regression is an intention-to-treat procedure, which can be considered as a sophisticated form of regression that allows participants with only one measurement to be included in the analyses (Hox, 2002). A significance level of < 0.05 (two-sided) was used throughout the study. Dropout was predicted by trauma and age for IRT. No variable predicted dropout in the exposure group. IRT scored higher on depression on baseline measurement than exposure (< 0.05). Attrition could have influenced post-test scores and therefore Cohen’s d estimations. To correct for this problem, multiple imputation based on the ‘missing at random assumption’ was employed (Sterne et al., 2009). For the missing scores, 10 separate data sets were generated with ‘predictive mean matching’. Cohen’s d was calculated with (Mpre1Mpre2)/σpretest (Morris, 2008) on the pooled mean of these data sets.


Pre–post measurements and Cohen’s d are depicted in Table 2 and Fig. 2. For both IRT and exposure the effect on nightmares, nightmare distress, depression and sleep complaints was sustained after 42 weeks (< 0.001). No differences were found between IRT and exposure after 42 weeks. We observed a trend for exposure being superior to IRT in ameliorating depression after 42 weeks (= −3.28; standard error (SE) = 1.78; = 0.06). However, IRT ameliorated nightmares more effectively compared to exposure 16 weeks after the end of the treatment (= −0.39; SE = 0.18; < 0.05); this effect disappeared at the final measurement (see supporting Table S1 for regression coefficients). The recording and waiting-list groups (with IRT intervention after the first follow-up) showed no significant differences in effect compared to IRT and exposure at the 42-week follow-up. Pre–post measurements of the recording and waiting-list group can be found in supporting Table S2 (with IRT after the first follow-up).

Table 2.   Mean, standard deviation (SD) and corresponding Cohen’s d at baseline, 4-week, 16-week and 42-week follow-up
 Pre4 weeksd16 weeksd42 weeksd
Mean (SD)Mean (SD)Mean (SD)Mean (SD)
  1. EXP, exposure; IRT, imagery rehearsal therapy; PTSD, post-traumatic stress disorder.*< 0.05; **< 0.01; ***< 0.001; NS, not significant; significance levels are based on multi-level regression analyses and Cohen’s d is based on the data set imputed with multiple imputation; see supporting Table S1 for pre–post-test means of recording and waiting-list group.

Nightmare frequency/week
 IRT5.15 (3.56)2.76 (2.74)−0.66***2.03 (1.75)−0.84***2.52 (1.75)−0.68***
 EXP4.66 (3.28)2.86 (2.67)−0.58***2.98 (2.47)−0.58***2.82 (2.48)−0.62***
Nights with nightmares/month
 IRT14.67 (7.97)9.71 (8.49)−0.69***7.08 (6.31)−0.84***8.09 (6.59)−0.64***
 EXP13.97 (7.85)9.34 (7.61)−0.65***9.67 (9.31)−0.65***10.55 (9.78)−0.50***
Nightmare distress
 IRT15.31 (4.42)12.76 (4.67)−0.49***12.03 (4.94)−0.64***11.05 (4.26)−0.64***
 EXP15.34 (4.10)12.27 (3.93)−0.71***12.79 (4.14)−0.61***12.79 (4.60)−0.70***
 IRT46.31 (9.75)42.88 (9.37)−0.25**41.61 (10.27)−0.36**42.55 (10.94)0.18NS
 EXP43.43 (11.67)41.07 (11.82)−0.17**41.53 (10.56)−0.21NS43.36 (11.57)0.07NS
PTSD complaints
 IRT30.12 (17.35)26.36 (21.34)−0.10NS23.75 (18.28)−0.20NS22.91 (18.38)−0.36NS
 EXP26.50 (16.85)23.53 (18.03)−0.16NS19.33 (16.40)−0.35*20.12 (16.07)−0.35NS
 IRT21.33 (6.39)15.18 (10.19)−0.64***13.92 (11.08)−1.06***15.06 (11.15)−0.76***= 0.06
 EXP21.55 (5.59)13.73 (10.18)−1.34***14.82 (8.48)−1.13***12.85 (10.10)−1.60***
Subjective sleep quality
 IRT5.64 (1.34)6.19 (1.35)0.24***6.57 (1.37)0.49***6.70 (1.19) 0.71***
 EXP5.61 (1.036.20 (1.03)0.57***6.10 (1.30)0.55***6.24 (1.35) 0.77***
Figure 2.

 Nightmare frequency per week, nights with nightmares per month, and nightmare distress at baseline, 4 weeks (post-1), 16 weeks (post-2) and 42 weeks (post-3) after treatment.

Clinical changes

In the IRT condition, nightmare frequency per week was reduced by 46.4% (4 weeks), 60.6% (16 weeks) and 51.0% (42 weeks); nights with nightmares: 33.8%/51.7%/44.9%; nightmare distress: 27.4%/35.2%/45.8%. Nightmare frequency per week < 1 was reported by: 0.0% (baseline), 16.9% (4 weeks), 24.3% (16 weeks) and 15.2% (42 weeks).

In the exposure condition, nightmare frequency per week was reduced by 38.6% (4 weeks), 36.1% (16 weeks) and 39.5% (42 weeks); nights with nightmares: 33.1%/30.8%/24.5%; nightmare distress: 32.9%/27.0%/16.6%. Nightmare frequency per week < 1 was reported by: 0.0% (baseline) 16.1% (4 weeks), 12.5% (16 weeks) and 21.2% (42 weeks).


This study showed that the reduction in nightmare frequency and short-term improvements in sleep quality after self-help IRT as reported in Lancee et al. (2010) were sustained after 42 weeks; no significant differences were found between IRT and exposure. In the IRT condition, nightmare frequency was reduced by 61% at 16-week follow-up and by 51% at 42-week follow-up. The results after 16 weeks are similar to outcomes obtained in face-to-face intervention studies [e.g. Krakow et al. (1995) found a 67% decrease at 3 months]. However, the effectiveness at 42-week follow-up was smaller [e.g. Krakow et al. (1996) found an 83% decrease after 18 months].

Moreover, this study replicated the long-term effectiveness of self-help exposure for nightmares. In the exposure condition, nightmare frequency dropped by 39% 4 weeks after the intervention and by 40% at 42 weeks. Burgess et al. (1998) found a similar nightmare frequency decrease after the intervention (43%) and a larger decrease at 6-month follow-up (58%). In the Burgess et al. (1998) study, however, participants performed self-exposure up to an hour a day (in contrast to 10–15 min in the current study) and completed a diary as part of the follow-up, which also ameliorates nightmares (Lancee et al., 2010). Similar effects were also observed in the control conditions that received IRT after the initial waiting-list or nightmare recording period; at the last follow-up we observed a decrease of 39.7% on nightmare frequency for the waiting-list and 53.6% for the recording condition. To exclude any time effects, future studies should include longer-term control conditions.

A limitation is that for a large group of participants nightmares did not disappear completely in the IRT and exposure conditions. Some form of interaction with a therapist might enhance effectiveness (e.g. e-mail support; Andersson, 2009). Moreover, for this study we did not include sleep hygiene, as we aimed to strictly compare the different types of interventions. However, including this technique may result in higher effect sizes (especially for the sleep complaints). In addition, some people have different nightmares every night. Lucid dreaming therapy (Spoormaker and van den Bout, 2006) might be more effective for these people, as in this treatment participants are taught to alter their nightmare within the dream itself.

Our results suggest that specific treatment of a sleep disorder has long-term effectiveness and it may therefore be useful to address specific sleep complaints in mental health care. This finding is relevant for nightmares, which are viewed commonly as a symptom of an underlying affective disorder by mental health professionals. Diagnosis of nightmares according to the DSM-IV-TR (American Psychiatric Association, 2000) is allowed only if nightmares do not occur during the course of another disorder, which may have the consequence that readily available and effective treatments are not delivered to patients. Self-help interventions (especially in a stepped-care context) seem to be a good option to deliver this treatment because of its low cost and easy-to-deliver format.


This study was funded by the Dutch Foundation for Mental Health, located in Amersfoort, the Netherlands (FPG20066126).

Conflict of interest

This was not an industry-supported study. The authors have indicated no financial conflicts of interest.