Don't worry, be constructive: A randomized controlled feasibility study comparing behaviour therapy singly and combined with constructive worry for insomnia

Objectives. Based on the lack of research on interventions targeting intrusive and worrisome thinking for insomnia, the aim was to examine whether a constructive worry (CW) intervention adds to the effects of behaviour therapy (BT). Design. A randomized, controlled design was used. The design included a 2‐week baseline, a 4‐week intervention phase (sleep restriction and stimulus control [BT] or sleep restriction and stimulus control plus constructive worry [BT + CW]), and a 2‐week follow‐up. Methods. Twenty‐two patients with primary insomnia participated. The primary outcome was the Anxiety and Preoccupation about Sleep Questionnaire (APSQ), and secondary endpoints were subjective sleep estimates, the Insomnia Severity Index (ISI), and the Work and Social Adjustment Scale (WSAS). Results. Although both conditions produced significant improvements in subjective sleep estimates, no significant group differences over time were shown for total wake time (TWT) and total sleep time (TST). Both interventions resulted in reductions over time in insomnia severity, worry, and dysfunction. Compared to BT, BT + CW led to a larger decrease in insomnia severity at all three time points (controlled d= 1.10–1.68). In comparison with BT, BT + CW resulted in a larger reduction in worry at two of the time points (controlled d= 0.76–1.64). No significant differences between the two conditions were demonstrated for dysfunction. While more participants responded positively to treatment in the BT + CW (80–100%) than in the BT condition (18–27%), none of the participants remitted. Conclusions. The findings suggest that, compared to BT alone, CW might result in additional improvements in insomnia severity and worry. Given the small sample size and short follow‐up, future studies are warranted.

Insomnia is a chronic difficulty that involves problems initiating sleep, maintaining sleep, or waking in the morning not feeling restored. It is one of the most prevalent health problems reported by 10% of the population ([1]). The consequences for the sufferer are severe and include functional impairment, work absenteeism, impaired concentration and memory, increased use of medical services, and a heightened risk of subsequently developing another psychiatric disorder ([7]; [15]; [42]). Insomnia is therefore regarded as a serious public health problem.

There has been a growing interest in the importance of intrusive and worrisome thinking in the maintenance of insomnia (e.g., [5]; [13]; [18]; [34]; [38]). Analysis of the content of pre‐sleep cognitive arousal suggests that it is composed of several components, for example, problem solving, reflection on quality of thoughts, environmental influences, arousal state, and the consequences of not sleeping ([50]). According to several models of insomnia, intrusive thoughts prior to sleep and during nightly awakenings are viewed as exacerbating the sleep difficulties associated with the condition ([13]; [18]; [34]; [38]). For example, in a cognitive model of insomnia, it has been proposed that worry can trigger autonomic arousal and emotional distress, culminating in an anxious state under which sleep is likely to be disturbed and daytime functioning may be impeded ([18]). The evidence for a role for worry has accrued (see [21] for review). Experiments have shown that an increase in worry result in an increase in the subjective estimation of time taken to fall asleep ([16]; [17]; [33]) and experimental manipulations that decrease worry result in a decrease in sleep onset latency ([27]; [23]; [31]).

Despite the theoretical models and existent research underscoring intrusive and worrisome thinking as a maintaining factor in insomnia, very little empirical work has yet focused upon intervention strategies pinpointing such thinking. A few exceptions should be mentioned. A cognitive intervention, consisting of cognitive restructuring, paradoxical instructions, and thought stopping, has been shown to be superior to biofeedback training in reducing pre‐sleep intrusions ([44]). In a second investigation, both mental relaxation and cognitive control procedures were superior, compared to muscle relaxation, in decreasing pre‐sleep intrusive cognitions ([36]). In a third study, a combined approach of relaxation training and cognitive control procedures was more effective in reducing pre‐sleep cognitive intrusions than both relaxation and treatment of sleep perception as well as than a passive control group ([35]). In a fourth study, imagery distraction resulted in less distressing pre‐sleep cognitive activity, relative to general distraction and no instructions ([23]). In a fifth study, articulatory suppression (repeating a set of phonemes in competition with ongoing thoughts) led to reductions in sleep maintenance problems.

Based on several aspects, it is likely that strategies targeting intrusive and worrisome thinking that aim to suppress and distract such thinking may not be optimal in the long run. First, research has shown that suppression of a thought likely to dominate pre‐sleep cognitive activity, compared with non‐suppression, leads to longer sleep onset latency and poorer sleep quality in both insomniacs and good sleepers ([19]), suggesting that suppression strategies may maintain insomnia. Research on thought suppression has generally shown that individuals, who attempt suppression of a target thought, experience an increase in the frequency of occurrence of the thought (i.e., ironic effects of thought suppression). This ironic effect has, for example, been demonstrated in normal sleepers instructed to either fall asleep as quickly as possible or fall asleep whenever they desired ([2]), smokers attempting reduction or cessation of smoking ([43]), and in individuals with post‐traumatic stress disorder ([45]). Second, the literature suggests that patients with insomnia hold positive beliefs about the utility of pre‐sleep worry ([20]), indicating that positive beliefs about the benefits of worrying during the pre‐sleep period may serve to maintain insomnia. These findings are consistent with those reported in the literature on generalized anxiety disorder (e.g., [6]).

It is thus possible that interventions other than suppression and distraction may be more helpful for patients with insomnia. However, few studies on insomniacs have yet focused upon other intervention strategies to reduce intrusive and worrisome thinking. Four exceptions should be mentioned. In a case study, an early evening problem‐solving component termed worry control (pre‐sleep writing) was shown to decrease pre‐sleep worry and time awake after sleep onset ([14]). Limitations from the case study include the use of a single case only and that the worry control was implemented after another treatment component (paradoxical intention). In a second investigation, writing about worries and concerns (Pennebaker‐style writing intervention) was used in an analogue sample of poor sleepers ([22]). Compared with a group who was not assigned the writing task, the group who wrote down worries and concerns reported shorter sleep onset latency. The authors hypothesized that this intervention was effective because it facilitated processing of pre‐sleep worries. A few limitations should be underscored, for example, no pre‐treatment sleep data, no instrument assessing intrusive and worrisome thinking, an intervention phase consisting of only three nights, and the participants were an analogue sample of poor sleepers. In a third study on a Pennebaker‐style writing intervention, the writing component, relative to a control condition, was more effective in reducing pre‐sleep cognitive arousal for insomniacs ([37]). This main finding was however demonstrated for only one of two arousal measures, and there was no group difference on sleep onset latency. In a controlled trial on college students with insomnia, a constructive worry (CW) intervention (recording of worries and problem solving) was compared with a control condition, involving only recording of worries ([9]). The findings suggested that the CW intervention resulted in a significantly larger reduction than the control condition on pre‐sleep cognitive arousal and also relative to pre‐treatment. Some limitations hamper the interpretation of the study, for example, possible floor effects on sleep, arousal, and anxiety parameters, an intervention phase of only 5 days, and the use of an analogue sample.

Cognitive‐behavioural therapy is currently considered as an evidence‐based treatment for chronic insomnia ([39]). Although significant progress has been made to design effective cognitive‐behavioural treatments for patients with chronic insomnia, the point where patients can be offered a maximally effective psychological treatment have yet not been reached ([24]). An additional problem with the evidence base of cognitive‐behavioural therapy is also that very few studies investigating its effect have used measures for intrusive and worrisome thinking as an outcome, despite that such thinking is often viewed as maintaining insomnia. There is thus scope for improvements. Although little is known about the effects of CW, preliminary data indicate that it could possibly add to the effects of already evidence‐based interventions. The aim of this study was therefore to examine whether a CW intervention adds to the effects of standard behaviour therapy (BT) (sleep restriction and stimulus control) on worry, subjective sleep estimates, insomnia severity, and dysfunction.

A randomized controlled trial with assessments at pre‐treatment, post‐treatment, and at a 2‐week follow‐up compared the effects of BT (sleep restriction and stimulus control) with behaviour therapy plus constructive worry (BT + CW). The primary outcome was sleep‐related worry, and the secondary endpoints were subjective sleep estimates, insomnia severity, and dysfunction. The study followed the ethical guidelines as formulated by the American Psychological Association.

Individuals with insomnia disorder were recruited through an advertisement in local newspapers and through posters placed at the university, libraries, and primary care centers in Örebro. The advertisements and posters instructed those interested to participate in a treatment trial to contact the project group via telephone or e‐mail. Over the phone, the candidates were initially informed about the project and then screened. The following inclusion criteria were required at this stage ([12]; [32]): (1) age between 18 and 75 years, (2) good language skills in Swedish, (3) presence of sleep complaints during the past 6 months, (4) clinical levels of insomnia symptoms during the past 2 weeks (sleep initiation difficulties [31 min or more], sleep maintenance difficulties [31 min or more], early morning awakening difficulties [31 min or more], non‐restorative sleep, or poor sleep quality), (5) daytime impairment during the past 2 weeks in one domain or more, and (6) insomnia symptoms not exclusively due to environmental factors, medical disorders, psychiatric disorders, or substances (i.e., medications, drugs, and beverages). The exclusion criterion at this stage was that the participant reported another sleep disorder than insomnia (i.e., nightmares, restless legs syndrome, periodic limb movement disorder, sleep apnea, hypersomnia, circadian‐rhythm disorder, and parasomnia), that the insomnia symptoms was exclusively due to the presence of these sleep disorders, and that the participant reported an unstable sleep medication regimen. The interviewer also administered the Insomnia Severity Index (ISI) to ensure a clinical level of insomnia (inclusion criterion: 15 points or more). To ensure a significant level of worry, an additional inclusion criterion was that the participant reported a score of 3 or more ('much' or more) on the ISI worry/distress item ('How worried/distressed are you about your current sleep problem?').

If the above criteria were fulfilled, the participant was invited to a face‐to‐face screening interview at the university. Again, the ISI was administered and this time completed by the participant (same inclusion criterion as above). During the interview, two interview formats were used. First, the Duke Structured Interview for Sleep Disorder (DSISD) (Edinger et al., unpublished material) was administered. The DSISD assesses Diagnostic and Statistical Manual of Mental Disorders (DSM)‐IV‐TR sleep disorders (i.e., insomnia, hypersomnia, circadian‐rhythm sleep disorder, and parasomnia) and associated psychiatric and medical conditions, medications, drugs, substances, beverages, and allergies. The DSISD has showed acceptable reliability and discriminant validity in studies ([8]). Second, the Primary Care Evaluation of Mental Disorders (PRIME‐MD) was used ([47]). The PRIME‐MD is a screening questionnaire and diagnostic interview based on the DSM‐IV system. The instrument has shown good concordance between PRIME‐MD diagnoses and those of independent mental health professionals using lengthier interview protocols. The PRIME‐MD provided the basis for identifying major psychological disorders that are known to affect sleep, including major depressive disorder, generalized anxiety disorder, panic disorder, and post‐traumatic stress disorder. At this stage, the same inclusion and exclusion criteria were applied as during the telephone screening. The participant was excluded if there was evidence of a serious or untreated psychiatric or medical disorder. It should however be underscored that a minority of the participants fulfilled criteria for either a psychiatric co‐morbidity (i.e., major depression, generalized anxiety disorder, and panic disorder) or a medical co‐morbidity (i.e., chronic pain and gastrointestinal disorders). A total of 22 individuals were included in the study. All included participants signed a written consent and were then randomized to BT or BT + CW.

A pre‐established table consisting of random numbers was used to assign each participant to the two conditions. Randomization occurred after the pre‐treatment assessment had been completed. The table of random numbers was kept in a drawer, which was only opened when a participant had met the study criteria. The assignment was made by the member of our research group who performed the interview. Eleven participants were assigned to the BT condition and 11 to the BT + CW condition.

After randomization, a 2‐week interval followed during which the participants completed three questionnaires (ISI; Anxiety and Preoccupation about Sleep Questionnaire, APSQ; and Work and Social Adjustment Scale, WSAS) on a weekly basis (i.e., once a week) and sleep diaries on a daily basis. Participants completed the material at home and returned it to each treatment session. After this interval, the participants started their assigned intervention. During the intervention phase and the 2‐week follow‐up period, the questionnaires and sleep diaries were again completed as described above. All the measures were returned at each session to the participant's assigned therapist.

Two participants dropped out during the study period. The first dropout, randomized to BT + CW, reported at the fourth treatment session an unstable sleep medication use during the past week (an exclusion criterion) and was therefore excluded from the trial. The second dropout, randomized to BT, reported difficulties completing the sleep diaries at the first treatment session and dropped out due to this reason. Only the second dropout was included in the intention‐to‐treat analyses. No other participant dropped out during the study period, neither during the intervention phase nor during the follow‐up. In sum, the total dropout rate was 9.1%. Due to the small number of dropouts, it was not possible to determine with inferential statistics whether the dropouts differed from the completers. Instead, only visual analysis was performed on clinical parameters. The visual analysis showed that the dropouts were similar to the completers on demographic parameters, sleep complaints (sleep diary), the ISI, the APSQ, and the WSAS.

The measures consisted of both daily sleep diaries and weekly questionnaires. All the measures were filled out during pre‐treatment, mid‐treatment, post‐treatment, and follow‐up. While the pre‐treatment and follow‐up scores consisted of an average of 2 weeks of either daily or weekly instruments, the mid‐ and post‐treatment scores were averaged based on 1 week of diaries and questionnaires. The sleep diaries contained one item assessing total sleep time (TST in minutes) and two items (sleep onset latency and wake after sleep onset) tapping total wake time (TWT in minutes). Three psychometrically validated self‐report scales were also used. The instrument to determine sleep disturbance was the ISI, containing seven items (ISI; [3]). The ISI has standardized cut‐offs: 0–7, no clinically significant insomnia; 8–14, sub‐threshold insomnia; 15–21, moderate clinical insomnia; and 22–28, severe clinical insomnia. The measure to assess worry was the APSQ ([30]; [48]). The APSQ, consisting of 10 items, determines worries about the consequences of poor sleep and worries about the uncontrollability of sleep. Though there are several other self‐report instruments tapping sleep‐related worry (e.g., Pre‐Sleep Arousal Scale), we chose the APSQ for two reasons: (1) its psychometric properties are of good quality ([30]) and (2) its items are insomnia specific and capture the distinctive experience of worry in insomnia patients (both worries concerning the consequences of poor sleep and the uncontrollability of sleep). The instrument to determine function was the five‐item WSAS ([41]). The WSAS assesses functioning across work, home management, social leisure activities, private leisure activities, and relationships with others.

Both the interventions followed a written manual and consisted of four sessions each lasting 30 min (except for the first session that lasted 2 h). The interventions were administered by one of the authors (two psychology students in their fifth year of training and a clinical psychologist). To increase the likelihood that the intervention was delivered as intended, the therapists received supervision on the progress of treatment. Also, before starting the project, the therapists used role‐play to incorporate the interventions. The therapists were instructed to note any deviations from the manual during the intervention phase. No deviations occurred. They were also instructed to note if they used any non‐intended interventions during the intervention phase. No such interventions were reported. Before starting each new session, the therapists discussed the previous session in detail and discussed difficulties in adhering to the manual. Written material was distributed to the patients to enhance compliance. In terms of treatment adherence, all but one patient attended all sessions.

The patients received only sleep restriction and stimulus control ([4]; [46]). With sleep restriction, participants were instructed to reduce their time in bed to the TST (never less than 5 h) based on pre‐treatment sleep diaries and during subsequent weeks to increase their time in bed (30 min per week) until sleep efficiency reached or exceeded approximately 85%. The stimulus control guidelines included unwinding before going to bed, going to bed only when sleepy, getting out of bed whenever unable to fall asleep within 15 min, having a fixed time for getting out of bed in the morning, using the bedroom only for sleep and sex, avoiding napping at daytime, and avoiding clock watching at nighttime. The first session consisted of basic psycho‐education on physiological aspects of sleep. The patients also received a rationale and detailed instructions for using sleep restriction and stimulus control. Sessions 2–3 were devoted to individualizing the two components based on information from the sleep diary. The fourth session consisted of a summary of the intervention, an individualized plan for the future, and relapse prevention. The follow‐up session, which took place 2 weeks after the fourth session, was used as a booster session, again focusing upon individualizing the two components.

The patients received sleep restriction, stimulus control, and a CW intervention. The two first components were administered as described above. The CW intervention was based upon previous work as described by [9]. With the CW intervention, participants were instructed to record worries on a worksheet during the early evening (at least 2 h before bedtime) concerning all areas of life (e.g., sleep, work, and relationships). The instruction was further to record worries that the participants thought had the greatest likelihood of keeping them awake at nighttime. On the same worksheet, they were asked to brainstorm ideas on how they could solve the worries. This procedure was repeated for all the listed worries. When they had completed the intervention each evening, they were instructed to fold the worksheet in half and place it on the nightstand. To cope with worries off the 15‐min CW period, participants were asked to tell themselves that they had dealt with their worries at an appropriate time for problem solving, and not when fatigued or in bed, and that the intervention would be repeated the next day. The participants were instructed to bring the worksheet to each session for further problem solving with the therapist. At the first session, the patients were given a rationale and detailed instructions for using CW. The second and third sessions consisted of individualizing the component based on homework assignments. The fourth session was devoted to a summary of the intervention, an individualized plan for the future, and relapse prevention. The follow‐up session was used as a booster session, focusing upon individualizing the component. Though the BT + CW intervention contained one component more than the BT condition, the allotted time for the first and subsequent sessions allowed adequate room for discussing homework and formulating new assignments for both the BT and the CW parts. The sequencing of the two parts during all the sessions was accomplished by first exclusively focusing on sleep restriction and stimulus control (approximately 20 min during subsequent sessions), which was then followed by a sole focus on the CW intervention (approximately 10 min).

t‐Tests and chi‐square analyses were used to examine whether there were differences between the two groups on the outcome measures at pre‐treatment. On the outcome measures, two‐way repeated measures Analysis of Variance (ANOVA) were used. All the analyses were based on an intention‐to‐treat principle (estimations by carrying forward the last valid observation). In these analyses, group assignment (BT and BT + CW) was treated as the between‐subjects factor and time (pre‐treatment, mid‐treatment, post‐treatment, and follow‐up assessment) was viewed as a repeated measure. Although the main effects for group or time are also reported when relevant, the Group × Time interaction was the main statistical contrast of interest. A significant Group × Time interaction effect was followed by an examination of the time effect within each group individually. A significant time effect for either group was then followed by the appropriate within‐subject contrasts. Finally, between‐groups comparisons were conducted on the mid‐treatment, post‐treatment, and follow‐up scores.

Within‐group effect sizes were calculated [(pre‐treatment minus post‐treatment or follow‐up)/pooled standard deviation] to gain an impression of the magnitude of improvement associated with the two interventions. Between‐group effect sizes were also calculated (post‐treatment or follow‐up for BT + CW minus post‐treatment or follow‐up for BT/pooled standard deviation). [10] proposed a three‐fold classification of effect sizes: small (0.20–0.49), medium (0.50–0.79), and large (0.80 and above). The outcome measure that was employed to examine treatment response and remission resulting from the two conditions was the ISI. The remaining five outcomes were not used due to the lack of standardized criteria for response and remission. Chi‐square statistic was employed to explore whether there were significant differences on response and remission between the two conditions.

The pre‐treatment characteristics for the two groups are reported in Table 1. The mean age in the sample was 56.4 years and 52.4% were women. Approximately, 24% of the participants reported sleep maintenance as their primary sleep complaint and 76% both sleep onset and sleep maintenance as complaints. The mean duration of insomnia in the sample was 10.3 years. Five of the 21 participants reported a stable use of sleep medication (i.e., zolpidem and zopiclone). No significant differences between the groups emerged regarding demographic and clinical parameters at pre‐treatment. t‐Tests showed that there were no significant differences between the two groups at pre‐treatment on the five outcome measures: TWT (t=−0.99, p=.33), TST (t=−0.08, p=.94), ISI (t=−1.33, p=.20), APSQ (t=−1.11, p=.28), and WSAS (t=−0.97, p=.34).

1 Overview of the study participants

1 Note. The numbers in parenthesis represent either standard deviations or percent.

The descriptive statistics across the two conditions for TWT and TST are displayed in Table 2.

2 The sleep diary outcome measures at four assessment points across the two groups

2 Note. TWT, total wake time; TST, total sleep time. aSignificant improvement from previous assessment point. bSignificant deterioration from previous assessment point. =No significant change from previous assessment point.

The 2 × 4 ANOVA of TWT revealed a significant Time effect (F= 58.50, p <.001) and a significant Group × Time interaction (F= 4.89, p=.004). For BT, TWT was significantly reduced from pre‐ to mid‐treatment (t= 6.61, p <.001), but no significant change occurred from mid‐ to post‐treatment (t=−2.00, p=.073) and from post‐treatment to follow‐up (t= 2.09, p=.064) (percentage change: 33–46% decrease; within‐group effect sizes: d= 1.16–2.47). For BT + CW, TWT was significantly decreased from pre‐ to mid‐treatment (t= 8.49, p <.001), no significant change occurred from mid‐ to post‐treatment (t= 1.65, p=.133), and a significant increase then followed from post‐treatment to follow‐up (t=−2.28, p=.048) (percentage change: 53–63% reduction; within‐group effect sizes: d= 2.96–4.75). Between‐groups contrasts showed that there were no significant differences on TWT between the two groups at mid‐treatment (t= 0.36, p=.724), at post‐treatment (t= 2.04, p=.066), and at follow‐up (t= 1.06, p=.309). The between‐group effect sizes were small to large (d: 0.16–1.03), in favour for the BT + CW group.

The 2 × 4 ANOVA of TST revealed a significant Time effect (F= 40.33, p <.001) but a non‐significant Group × Time interaction (F= 0.44, p=.724). For BT, TST was significantly increased from pre‐ to mid‐treatment (t=−4.65, p=.001) and from mid‐ to post‐treatment (t=−4.19, p=.002), but no change occurred from post‐treatment to follow‐up (t=−1.04, p=.321) (percentage change: 5–10% increase; within‐group effect sizes: d= 1.14–1.42). For BT + CW, TST was significantly increased from pre‐ to mid‐treatment (t=−2.61, p=.028) and mid‐ to post‐treatment (t=−3.54, p=.006), but no change occurred from post‐treatment to follow‐up (t=−1.70, p=.123) (percentage change: 3–9% increase; within‐group effect sizes: d= 1.11–1.41). Between‐groups contrasts demonstrated that there were no significant differences on TST between the two groups at mid‐treatment (t= 0.51, p=.616), at post‐treatment (t= 0.47, p=.644), and at follow‐up (t= 0.25, p=.809). The between‐group effect sizes were less than small to small (d: 0.11–0.21), in favour for the BT group.

The descriptive statistics for the ISI across the two groups are shown in Table 3. The 2 × 4 ANOVA of ISI revealed a significant Time effect (F= 67.15, p <.001) and Group × Time interaction (F= 5.29, p=.003). The ISI score for the BT group was significantly reduced from mid‐ to post‐treatment (t= 6.71, p <.001), but not from pre‐ to mid‐treatment (t= 2.18, p=.055) and not from post‐treatment to follow‐up (t= 1.52, p=.160) (percentage change: 14–31% decrease; within‐group effect sizes: d= 1.29–2.62). For BT + CW, the ISI score was significantly decreased from pre‐ to mid‐treatment (t= 5.57, p <.001), from mid‐ to post‐treatment (t= 3.12, p=.012), but not from post‐treatment to follow‐up (t= 2.05, p=.070) (percentage change: 30–53% reduction; within‐group effect size: d= 3.24–6.09). Between‐groups contrasts confirmed that the BT + CW group had a lower ISI score than the BT group at mid‐treatment (t= 2.46, p=.023), at post‐treatment (t= 2.61, p=.017), and at follow‐up (t= 3.63, p=.002). The between‐group effect sizes were large (d: 1.10–1.68), in favour for the BT + CW group.

3 The questionnaire outcome measures at four assessment points across the two groups

3 Note. ISI, Insomnia Severity Index; APSQ, Anxiety and Preoccupation Questionnaire; WSAS, Work and Social Adjustment Scale. *Significant group difference: BT + CW > BT. aSignificant improvement from previous assessment point. =No significant change from previous assessment point.

The descriptive statistics for the APSQ across the two groups are displayed in Table 3. The 2 × 4 ANOVA of APSQ revealed a significant Time effect (F= 25.75, p <.001) and Group × Time interaction (F= 3.99, p=.012). For BT, APSQ was significantly reduced from mid‐ to post‐treatment (t= 2.75, p=.020), but not from pre‐ to mid‐treatment (t=−0.11, p=.916) and not from post‐treatment to follow‐up (t=−0.97, p=.356) (percentage change: from 0.3% increase to 16% decrease; within‐group effect sizes: d= 0.89–0.91). For BT + CW, APSQ was significantly decreased from pre‐ to mid‐treatment (t= 3.07, p=.012), from mid‐ to post‐treatment (t= 7.12, p <.001), and from post‐treatment to follow‐up (t= 3.62, p=.006) (percentage change: 14–36% reduction; within‐group effect sizes: d= 4.58–7.82). Between‐groups contrasts confirmed that the BT + CW group had a lower APSQ score than did BT participants at mid‐treatment (t= 2.34, p=.030) and at follow‐up (t= 3.21, p=.008), but not at post‐treatment (t= 1.59, p=.135). The between‐group effect sizes were medium to large (d: 0.76–1.64), in favour for the BT + CW group.

The descriptive statistics for the WSAS across the two conditions are shown in Table 3. The 2 × 4 ANOVA of WSAS revealed a significant Time effect (F= 32.50, p <.001) and Group × Time interaction (F= 4.46, p=.007). For BT, WSAS was significantly reduced from pre‐ to mid‐treatment (t= 5.06, p <.001), but not from mid‐ to post‐treatment (t= 0.08, p=.936) and not from post‐treatment to follow‐up (t= 0.475, p=.645) (percentage change: 16–18% decrease; within‐group effect sizes: d= 0.57–0.72). For BT + CW, WSAS was significantly decreased from pre‐ to mid‐treatment (t= 4.32, p=.002), from post‐treatment to follow‐up (t= 3.20, p=.011), but not from mid‐ to post‐treatment (t= 0.64, p=.536) (percentage change: 26–37% reduction; within‐group effect sizes: d= 1.27–1.96). Between‐groups contrasts confirmed that there were no significant differences on the WSAS between the two groups at mid‐treatment (t= 0.26, p=.797), at post‐treatment (t= 0.38, p=.712), and at follow‐up (t= 1.34, p=.196). The between‐group effect sizes were less than small to medium (d: 0.16–0.59), in favour for the BT + CW group.

In Table 4, the results on treatment response and remission based on the ISI are displayed. The treatment response and remission criteria for the ISI were taken from previous research ([40]). Treatment response was operationalized as follows: ISI change score, from baseline, mid‐treatment, or post‐treatment to follow‐up, equivalent to one category on the ISI (7 points). Remission was defined as follows: ISI score less than 8 points at follow‐up. In the BT group, 18–27% of the participants fulfilled criteria for treatment response at the three time points. In the BT + CW group, 80–100% of the participants met criteria for response at the three assessment points. Comparisons between the two conditions showed that there were significantly more participants in the BT + CW than in the BT condition that responded to treatment at mid‐treatment (χ2= 8.03, p=.005), post‐treatment (χ2= 14.32, p <.001), and at follow‐up (χ2= 5.84, p=.016). The table also shows that none of the participants remitted according to the ISI. Chi‐square statistic was therefore not executed.

4 Treatment response and remission based on the insomnia severity index

4 Note. ISI, Insomnia Severity Index. aISI change score, from baseline to mid‐treatment, post‐treatment or follow‐up, equivalent to one category on the ISI (7 points). bISI score less than 8 points.

Taken as a whole, this study showed that CW added to the effects of sleep restriction and stimulus control. This effect was evident by larger reductions on insomnia severity and sleep‐related worry. This supports the idea that CW could be an augmenting strategy for cognitive‐behavioural therapy for insomnia.

Consistent with previous research, the current findings showed that BT resulted in subjective sleep improvements ([39]). TWT was reduced from 134 to 82 min and TST was increased from 5.5 to 6 h. To investigate whether the current study's implementation of BT was similarly effective as previous findings, a benchmark procedure was used. From one of the latest reviews of the CBT literature on insomnia ([39]), only one study was identified that was similar to the current study ([49]). In the previous study, the treatment approach was short and consisted only of sleep restriction and stimulus control. When comparing the results of this study and the previous investigation on TWT, the efficacy of our BT implementation is comparable to previous findings (TWT reduction: 38% and 31%, respectively). The sleep improvements were also substantial for the group who received BT plus CW; TWT was decreased from 143 to 66 min and TST was increased from 5.5 to 6 h. The percentage change and the effect sizes for TWT indicated visually that CW added to the effects of BT, but this was not confirmed when statistical tests were applied. To investigate whether the size of the sample was too small to detect differences, post hoc analysis was performed. Based on the follow‐up scores, 96 (TWT) and 282 (TST) participants in each group were found needed to detect a significant difference between the two conditions (one‐tailed hypothesis; alpha level: 0.05; desired statistical power level: 0.8). It should also be mentioned that the previous study investigating the effects of CW did not find significant effects on subjective sleep parameters ([9]), a parallel finding with this study, though the percentage change and effect sizes were slightly higher in the present study. Taken together, although the sleep improvements were substantial in both groups, no significant group difference was found.

Besides sleep diaries, three questionnaires were used to evaluate whether CW has an augmenting effect for insomnia. Based on the ISI, both groups reduced their insomnia severity over time. While BT led to a 14–31% reduction in insomnia severity, the addition of CW resulted in a 30–53% decrease. The addition of CW showed statistically superior outcomes on insomnia severity at mid‐treatment, post‐treatment, and follow‐up, and the between‐group effect sizes were large. In terms of sleep‐related worry, which was our primary outcome, both groups reduced their scores on the APSQ. While BT resulted in a 0.3% increase to a 16% reduction in worry, the addition of CW led to a 14–36% decrease. The addition of CW showed statistically superior outcomes on sleep‐related worry at mid‐treatment and follow‐up, and the between‐group effect sizes were medium to large. The difference at mid‐treatment was non‐significant (post hoc analysis indicating a minimum of 23 participants in each group to detect a difference between the conditions). Regarding sleep‐related dysfunction, both groups decreased their scores on the WSAS. While BT led to a 16–18% decrease, the addition of CW resulted in a 26–37% reduction. The percentage change and the effect sizes for WSAS indicated visually that CW added to the effects of BT, but this was not confirmed when statistical tests were used (post hoc analysis indicating a minimum of 37 participants in each group to detect a significant difference between the two conditions).

In terms of treatment response and remission, based on the ISI, more participants in the group who also received CW (80–100%) than in the BT group only (18–27%) responded to treatment at mid‐treatment, post‐treatment, and at follow‐up. However, none of the participants remitted according to the ISI. Compared with a previous study on insomnia comparing cognitive behaviour therapy (sleep restriction, stimulus control, cognitive restructuring, and sleep hygiene) with cognitive behaviour therapy plus medication, the current findings on remission differ ([40]). The studies' results are comparable in terms of treatment response; while 48–57% of the patients in the current study showed ISI response, 59% did so in the CBT treatment arm in the previous investigation. Also, the percentage change was similar across the two studies; while the current study's treatment arms resulted in an ISI reduction of 31% in the BT condition and a 53% decrease in the BT + CW condition, there was a 48% ISI reduction in the previous study. However, while none in the current study met remission criteria, 39% did so in the previous investigation. It should however be underscored that the pre‐treatment ISI levels differ across the two studies. While the current study's pre‐treatment level was 21.4, the previous investigation was 17.3. In other words, this means that the participants in the current investigation had a higher insomnia severity before the interventions started, which in itself makes the remission criteria for ISI (score less than 8 points) more difficult to meet.

Although CW needs to be tested more in future research, the current findings indicate that there may be clinical benefits above and beyond sleep restriction and stimulus control. Although post hoc analyses indicated that with a larger sample group differences could have been detected on TWT and sleep‐related dysfunction, the more robust findings were that insomnia severity and sleep‐related worry were more largely reduced when adding CW. Compared with suppression strategies, CW has also potentially the advantage of addressing the intrusive and worrisome thinking in a problem‐solving way, but not leading to longer sleep onset latency and poorer sleep quality ([19]). Also, since many patients with insomnia hold positive beliefs about the utility of pre‐sleep worry ([20]), CW offers an alternative route for addressing sleep‐related worry. In all, CW could be an augmenting strategy for cognitive‐behavioural therapy for insomnia ([24]).

Despite the promising results for CW in the realm of insomnia, it should be underscored that there might also be other specific interventions that can hypothetically effectively address sleep‐related worry. One intervention type that might be worth exploring within an insomnia context are methods that replace cognitive avoidance with an approach towards thoughts and feelings, thereby providing habituation, for example, worry exposure, worry periods, and Pennebaker‐style writing intervention ([22]; [26]; [28]). Other intervention types that should be investigated are components that allow patients to shift processing to a more concrete level or to identify and challenge the positive and negative meta‐cognitive beliefs hypothesized to contribute to intrusive and worrisome thinking ([25]).

The study has a number of methodological shortcomings. An initial shortcoming is the relatively small sample that was used. An associated problem might also be that participants recruited via advertisements might differ from patients seeking care at sleep clinics ([11]). Hence, it is difficult to generalize the current findings to a broader patient group with insomnia. Second, due to that the inclusion criteria required a relatively high level of sleep‐related worry, there might be an interaction between selection bias and treatment. However, to explore the potential use of a worry intervention, it is vital to include patients with a significant level of worry to avoid floor effects. In addition, sleep‐related worry is often reported by individuals with insomnia. For example, 33% of individuals with insomnia disorder in our latest epidemiological study reported a significant level of sleep‐related worry ([29], submitted for publication). A third shortcoming is the short follow‐up period. Fourth, the procedure by which participants returned the measures to their assigned therapist might have introduced a reporting bias. Fifth, other worry measures, for example, the Insomnia Daytime Worry Scale or the Penn State Worry Questionnaire, could have provided additional information about the influence of the interventions on daytime worry and trait worry, thereby capturing a wider worry construct. Sixth, although the BT + CW group was generally superior to the BT condition, a clinical note that should be mentioned is that, even though the patients were instructed to use CW on a regular basis, some patients in the BT + CW condition reported using the intervention on an as‐needed‐basis. Hypothetically, CW might have its most potent clinical effect in the beginning of treatment to alleviate the initial side effects of particularly sleep restriction. This idea is supported by our finding that while BT led to only a minor change on insomnia severity and a very marginal deterioration on sleep‐related worry in the beginning of treatment, the addition of CW resulted in an early, more dramatic improvement on insomnia severity and sleep‐related worry. Given the shortcomings of this study, future research is clearly warranted.

In sum, this investigation showed that a short and easy‐administered CW intervention added to the effects of BT on insomnia severity, sleep‐related worry, and number of treatment responders. Given the elevated worry levels in patients with insomnia and theoretical accounts that highlight worry as a maintaining factor for insomnia, the current findings are promising.

We would like to express our appreciation to Mikael Bermås and Andreas Kjellén for collaboration on recruiting the participants.