A comparison of sleep restriction and sleep compression on objective measures of sleep: A sub‐sample from a large randomised controlled trial

Sleep restriction therapy is a central component of cognitive behavioural therapy for insomnia, but can lead to excessive sleepiness, which may impede treatment adherence. Sleep compression therapy has been suggested as a possibly gentler alternative. The aim of this study was to compare the effects of sleep restriction therapy and sleep compression therapy on objective measures of sleep, with a focus on magnitude and timing of effects. From a larger study of participants with insomnia, a sub‐sample of 36 underwent polysomnographic recordings, before being randomised to either sleep restriction (n = 19) or sleep compression (n = 17) and receiving online treatment for 10 weeks. Assessments with polysomnography were also carried out after 2, 5, and 10 weeks of treatment. Data were analysed with multilevel linear mixed effect modelling. As per treatment instructions, participants in sleep restriction initially spent shorter time in bed compared with sleep compression. Participants in sleep restriction also showed an initial decrease of total sleep time, which was not seen in the sleep compression group. Both treatments led to improvements in sleep continuity variables, with a tendency for the improvements to come earlier during treatment in sleep restriction. No substantial differences were found between the two treatments 10 weeks after the treatment start. The results indicate that homeostatic sleep pressure may not be as important as a mechanism in sleep compression therapy as in sleep restriction therapy, and an investigation of other mechanisms is needed. In conclusion, the treatments led to similar changes in objective sleep at a somewhat different pace, and possibly through different mechanisms.


Funding information
The L.J. Boëthius' foundation; Karolinska Institutet; Stockholm University Summary Sleep restriction therapy is a central component of cognitive behavioural therapy for insomnia, but can lead to excessive sleepiness, which may impede treatment adherence. Sleep compression therapy has been suggested as a possibly gentler alternative. The aim of this study was to compare the effects of sleep restriction therapy and sleep compression therapy on objective measures of sleep, with a focus on magnitude and timing of effects. From a larger study of participants with insomnia, a sub-sample of 36 underwent polysomnographic recordings, before being randomised to either sleep restriction (n = 19) or sleep compression (n = 17) and receiving online treatment for 10 weeks. Assessments with polysomnography were also carried out after 2, 5, and 10 weeks of treatment. Data were analysed with multilevel linear mixed effect modelling. As per treatment instructions, participants in sleep restriction initially spent shorter time in bed compared with sleep compression. Participants in sleep restriction also showed an initial decrease of total sleep time, which was not seen in the sleep compression group.
Both treatments led to improvements in sleep continuity variables, with a tendency for the improvements to come earlier during treatment in sleep restriction.
No substantial differences were found between the two treatments 10 weeks after the treatment start. The results indicate that homeostatic sleep pressure may not be as important as a mechanism in sleep compression therapy as in sleep restriction therapy, and an investigation of other mechanisms is needed. In conclusion, the treatments led to similar changes in objective sleep at a somewhat different pace, and possibly through different mechanisms.

K E Y W O R D S
CBT-I, digital treatment, internet therapy, mechanisms, SRT, time in bed regularisation 1 | INTRODUCTION Insomnia is a common disorder with a prevalence in the population of around 10% (Morin et al., 2006), with as many as 30% showing symptoms of insomnia (Ohayon, 2002). Cognitive behavioural therapy for insomnia (CBT-I) is effective in treating the disorder (van Straten et al., 2018;Zachariae et al., 2015) and is recommended as the first choice (Qaseem et al., 2016;Riemann et al., 2017). One of the main components of CBT-I is sleep restriction therapy (Miller et al., 2018). Sleep restriction therapy consists of curtailing the time spent in bed to the average sleep time of the past two weeks. Rise time is fixed and bedtime is delayed weekly until sleep efficiency increases (Spielman et al., 1987). Sleep restriction therapy has been associated with improvements in sleep efficiency, sleep latency, wake after sleep onset (Maurer et al., 2021), a decrease of the amount of stage 1 sleep, and an increase of the proportion of stage 3 sleep (Maurer et al., 2022). Although effective in treating insomnia both as part of CBT-I and as a single component treatment, sleep restriction therapy comes with possible disturbing side effects such as fatigue, extreme sleepiness, and headache (Kyle et al., 2011).
A sometimes used alternative to sleep restriction therapy is sleep compression therapy, first mentioned by Lichstein in 1988. In sleep compression, the reduction of time in bed is gradual and patients approach average sleep time either by reducing the time in bed by one fifth of the wake time in bed or by reducing time in bed by 15 min each week until satisfying sleep efficiency is reached (Lichstein et al., 2001;Riedel et al., 1995;Song et al., 2019). Both sleep restriction and sleep compression entail setting a so-called sleep window and adjusting it based on sleep efficiency criteria. Sleep compression is often described as resembling sleep restriction, but as a more gentle way to achieve the same goal McCurry et al., 2007) and has mainly been used in studies in elderly populations (Martin et al., 2017;Nau et al., 2005;Riedel et al., 1995;Song et al., 2019). Research on sleep compressing is scarce, and there is a general lack of studies investigating effects of CBT-I using polysomnography (PSG), the gold-standard assessment method of sleep variables (Mitchell et al., 2019). We have found only one study investigating the effect of sleep compression on objective sleep, and this study showed no effect on objective sleep variables (Lichstein et al., 2001).
The mechanisms behind the treatments are not fully understood.
The theory behind sleep restriction therapy is that the treatment induces a mild sleep deprivation, expected to drive positive changes in sleep (Spielman et al., 1987). This is concordant with the two-process model of sleep regulation stipulating that sleep pressure builds during wakefulness, and that prolonged wakefulness leads to more intense sleep (Borbély, 1982). More specifically, sleep restriction therapy is hypothesised to build homeostatic sleep pressure, to stabilise circadian rhythm, lead to less insomnia-related perpetuating practices (e g napping, spending excessive time in bed), and to reduce hyperarousal (Spielman et al., 2011). Overall, few studies have been designed to investigate mechanisms of sleep restriction therapy (e.g. Maurer et al., 2022;, and the results are mixed with the strongest support for increased homeostatic sleep pressurereflected in increased sleepinessand fewer perpetuating practices being mechanisms involved in the treatment (Maurer et al., 2018).
Sleep compression therapy, on the other hand, builds on a theory that people with insomnia spend an excessive time in bed as a result of a mismatch between sleep need and sleep goal, and that they might benefit from gradually shortening the time in bed to match the actual sleep need so that the sleep becomes less fragmented and wake time in bed is decreased (Lichstein, 1988;Lichstein et al., 2001). The endeavour is similar to that in sleep restriction therapy: to treat problems with sleep continuity by shortening the time in bed (Lichstein et al., 2001) although in sleep compression, this is done in much smaller steps. According to the two-process model, this should lead to a slow gradual increase of homeostatic sleep pressure that never reaches the same elevated levels as anticipated during the initial weeks of sleep restriction, although this has not been studied.
To our knowledge, there are no studies directly comparing the effects of sleep restriction and sleep compression therapy on sleep patterns. Comparing the two interventions might also have the benefit of shedding additional light on the mechanisms involved in the treatments.

| Aims and hypothesis
The aim of this study was to compare the effects of sleep restriction therapy and sleep compression therapy on objective measures of sleep. We were interested in comparing both the magnitude and timing of effects, to investigate if sleep restriction and sleep compression had an equally large effect on sleep continuity and sleep architecture, and to investigate if the effects occurred at the same time points during treatment.
We hypothesised that sleep restriction would initially be associated with a shorter time in bed (TiB) and a shorter total sleep time (TST) compared with sleep compression, with subsequent increases in both these variables, and that sleep compression would lead to a gradual shortening of TiB and TST, without subsequent increases. This should be seen as differences in these parameters at week 2 and 5, but these differences early in treatment would not be sustained at week 10.
We also hypothesised that similar patterns would be seen for sleep continuity and sleep architecture, i.e. larger improvements in sleep efficiency (SE), sleep onset latency (SOL), wake after sleep onset (WASO), number of awakening index (NAI), and more stage 3 sleep (N3) and less stage 1 sleep (N1) at week 2 and 5 in sleep restriction compared with sleep compression. In other words, differences between the two groups would be present at week 2 and 5, but these differences early in treatment should not be sustained at week 10.
We also set out to investigate within-groups effects, effects on stage 2 sleep (N2) and REM-sleep in exploratory analyses.

| Study design
The present study is part of a larger randomised controlled trial with 234 participants and presents data from the 36 participants undergoing polysomnographic assessments (PSG) during the treatment period. The study was conducted in Stockholm, Sweden, and has been preregistered at clinicaltrials.gov (NCT02743338). It was approved by the Swedish Ethical Review Authority (Dnr 2016(Dnr /44-31/4 and 2018(Dnr /2025. Written informed consent was obtained from all participants.

| Participants and recruitment
The larger study was open to adults in Sweden. For practical reasons, the present sub-study with focus on polysomnography recruited participants from the larger study living in the Stockholm region. Initially we sought to include 20 participants per treatment arm.

| Procedure
The web-based screening questionnaires measured insomnia severity, symptoms of sleep disorders, depression, and anxiety, use of alcohol and drugs and background information. Participants also signed informed consent about participation in the larger study. A telephone interview followed, which lasted for 20-30 min and assessed the presence of insomnia disorder, alcohol and/or drug use disorder, and symptoms of sleep disorders. Information was also given about polysomnography to participants from the Stockholm area, and a preliminary decision was made about participation. Those interested in participating in the polysomnography study signed informed consent.
Baseline assessments were conducted for participants not excluded or dropping out in an earlier step. Participants were to fill out a sleep diary online for at least 7 days and return pre-treatment assessments of insomnia severity. The first polysomnography recording (including a habituation night) was also carried out during the pre-treatment period. After finishing baseline assessments, participants were randomised to either sleep restriction or sleep compression therapy consisting of 5 weeks of online delivered text-based treatment with therapist support, followed by 5 weeks of self-care, i.e. continued online treatment without therapist support. Polysomnography recordings were carried out on three additional occasions: at treatment week 2, 5, and 10.

| Allocation concealment and randomisation
After baseline assessments and the first polysomnography recording, all participants were randomised to either sleep restriction therapy or sleep compression therapy by a person not involved in the study, using www.random.org/lists/. Participants were informed that one version of the treatment had a slightly faster pace while the other would approach the goal more gradually. They were not informed as to which version they were allocated. Since randomisation was performed on the whole sample involved in the larger study, and not just the sub-group involved in polysomnography, the procedure did not guarantee equally large groups for the current study.

| Interventions
As presented in Table 1, both interventions encompassed eight textbased modules delivered via an online treatment platform. Modules one and two were to be completed during the first week of treatment, followed by four weekly modules, and the seventh module being made available to the patient at week 6 (i.e. the beginning of the self-care period) and the last module at week 10. The participants could choose to complete a module in one sitting or to work on it over several days, and at the end of each module they were to fill out and send in a report on treatment progress and difficulties.
The report was reviewed and commented on by their therapist as part of the therapist guidance. If a patient had not sent in their weekly report, the therapist would send a message to the patient and encourage them to send it in. Patients could also send written messages through the online system which their therapist would answer within 48 hours except during weekends. Therapist guidance was given individually to each patient during the first 5 weeks, and during the last week of treatment, after the self-care period. Therapist messages contained encouragement to continue work with sleep restriction/sleep compression, problem solving, and help with understanding how to carry out the treatment (e g how to calculate the first sleep window). The treatment, including therapist messages, was entirely focused on sleep restriction or sleep compression, and contained no other components of CBT-I (i.e. stimulus control, relaxation, cognitive components, or other CBT components were not included in these treatments). Therapists were either licensed psychologists or clinical psychology students under supervision of a licensed psychologist. The online sleep diary that was used for calculation and adjustment of the sleep window, corresponds to the Consensus Sleep Diary (Carney et al., 2012), translated to Swedish.

| Background information
Background data were collected including age, sex, education, economic situation, occupation, and medication use. Depression symptoms were assessed with the Montgomery Åsberg Depression Rating Scale Self-rated version (Montgomery & Asberg, 1979), the Alcohol Use Disorder Identification Test (Babor et al., 1989) to examine use of alcohol, and the Insomnia Severity Index (Bastien et al., 2001) to assess insomnia symptoms at baseline and at week 10. See Table 2 for participants' characteristics.

| Use of hypnotic medication
Information about use of hypnotic medications was collected after each night with polysomnography-recording.

| Sleep parameters
Sleep parameters were derived from polysomnographic assessment (PSG). Polysomnography was carried out by one of the authors, P.D., who was not blinded to the allocation of the participants. All participants underwent four polysomnographic ambulatory nights during a period of 10 weeks; at baseline, week 2, week 5, and week 10. All participants also underwent a habituation sleep registration, to reduce the likelihood of a "first-night effect". To rule out sleep apnea, all participants were screened for breathing pauses of ≥10 s with desaturations of ≥3%.
A standard sleep registration was performed with scalp EEG derivations, standard filters, and visual scoring in accordance with sleep scoring guidelines (Berry et al., 2012), and arousals from sleep were defined according to Iber et al. (2007). The electrode montage and the impedance test, with a 5 kΩ maximum impedance, were performed in the participant's home, by an experienced researcher approximately 120 minutes before the usual bedtime. Ag/AgCl electrodes were used, and sleep data were recorded on portable Embla recorders (Flaga HF ® /Medcare) with a sampling rate of 256 Hz and collected in the participant's home in the morning following the sleep registration.

Total sleep time and time in bed
TST was defined as minutes of sleep in N1-3 and REM, and TIB as minutes between getting in and out of bed. Participants were considered to have a Good economic situation if they answered "Good" or "Very good" to the question "How is your current economic situation?". Other response options were "Neither good nor bad", "Bad", and "Very bad". b Depression symptoms were assessed with MADR-S (Montgomery Åsberg Depression Rating Scale -Self rating version).

| Treatment adherence
The participants were to set their first sleep window during the first week of treatment. If the sleep window was not set during the first week, the participant was prompted to set it as soon as possible. The sleep window was reviewed every week, after which the next sleep window was set. In order to assess adherence, the mean sleep window length for each group was calculated and compared with the mean time in bed from polysomnography and mean time in bed from sleep diaries. All polysomnography data were visually inspected for normality, homoscedasticity, and influential outliers. Variables that did not meet these assumptions were transformed using log transformations and, if that did not help, square root transformation according to recommendations in the literature (Field, 2018). Transformed variables were SE, WASO, SOL, NAI, and N1. Results from analyses using transformed and non-transformed variables were compared, and no substantial differences were found. The original data were therefore kept to facilitate the interpretation of results. Age was included as a covariate in a sensitivity analysis, since the two groups differed in mean age. 3 | RESULTS

| Information about participants
In total, 36 patients were included in the present study based on convenience selection of 153 eligible participants taking part of the larger study ( Figure 1). Participants were randomised to sleep restriction (n = 19) or sleep compression (n = 17). One participant dropped out of treatment during the self-help period, but continued assessments and is included in analyses. As seen in Table 2, a majority of participants were women, and the ISI score at baseline was within a clinical range. The mean age was statistically significantly higher in the sleep restriction group, while the sleep compression group scored higher on depression. The sleep compression group displayed a higher baseline SE and shorter SOL, but this difference was not statistically significant.
The mean ISI score at week 10 was 9.4 (5.3) and 8.8 (SD 3.8) in the sleep restriction group and sleep compression group, respectively.
This sub-group of participants were comparable to the participants in the larger study regarding age (mean in larger study: 44.8), sex (proportion of women in larger study: 74.2%), and ISI score at screening (mean in larger study: 19.4) with no statistically significant differences between the sub-sample and the full sample (p = 0.185-0.837).

| Missing data
All participants fulfilled all five nights of polysomnography, including a habituation night. Concerning the polysomnography-variable SOL, data are missing at baseline for one participant in the sleep compression group.

| Use of hypnotic medication during PSG recordings
In the sleep restriction group, three participants (16%) had used sleep medication during the baseline recording, four participants (21%) had used it during the second recording at week 2, two (11%) had used it during the third recording at week 5, and one (5%) had used hypnotics at the fourth recording during week 10. In the sleep compression group, three participants (18%) had used hypnotic medication at all four assessment points. Please see Table S2 in supplementary for more information about hypnotic medication. Table 3 shows the raw means and standard deviations. Figure 3a As can be seen in Figure 4, effect sizes comparing change in TST and TiB from baseline to the different time points, indicated large between-group differences at week 2, medium differences at week 5 and very small differences at week 10.

| Effects on sleep continuity
Raw means and standard deviations are found in Table 3.  There are small sized differences in SE, SOL, and WASO at week 2, implying somewhat larger improvements in the sleep restriction group from baseline to week 2. Table 3 presents raw means and standard deviations. Figure 3g,h illustrates the results for the two groups visually.
Sensitivity analyses with age included as a covariate did not change the results.
However, the effect sizes indicated small differences at week 5 for both N1 and N3, indicating larger decrease of stage 1 sleep and larger increase of stage 3 sleep in the sleep compression group. Please see Figure 4 for a visual presentation of effect sizes.

| Explorative analysis: REM, stage 2 sleep, and within-group analyses
Please see Table 3 for raw means and standard deviations, Figure 3 illustrates the results for the two groups visually. Including age as a covariate in a sensitivity analysis did not change the  Within-group analyses were performed as explorative analyses since a control group was missing. Please see Table S1 in  in line with our hypotheses and undoubtedly a surprise. The differences are small, but the tendency is interesting. According to the twoprocess model (Borbély, 1982), the sleep deprivation in sleep restriction therapy should have caused a similar change, and previous research has, indeed, found a decrease of N1 sleep and increase of N3 sleep (Maurer et al., 2022) following sleep restriction therapy.
From this perspective, the data from the current study do not support homeostatic sleep pressure as a mechanism for sleep restriction therapy. One possible explanation is the timing of the polysomnography measurements; Maurer et al. (2022) recorded the changes in sleep architecture during the first week with sleep restriction. It is possible that the changes in the sleep restriction group had occurred before our second polysomnography assessment in week 2, but that the later changes in sleep compression group were timed to this assessment. A more precise day-to-day assessment with polysomnography could provide answers to this.
Interestingly, the tendency for sleep restriction to show faster improvements in sleep continuity variables is consistent with the hypotheses that increased homeostatic sleep pressure caused by initial sleep deprivation drives positive changes in sleep. However, these differences were only present at treatment week 2, and the effect of the treatments on sleep continuity variables were comparable for the two treatments at week 5 and 10. At that time point, the sleep compression group had not yet experienced any decrease of total sleep time and it could therefore be argued that they should not be experiencing a higher sleep pressure. This implies that something other than homeostatic sleep pressure is involved in the improvements of sleep continuity variables, most notably in the sleep compression group, but possibly for both groups.
A possible explanation is that other proposed mechanismse g.
prevention of perpetuating practices, stabilising circadian rhythm, and reduction of hyperarousalare driving the changes in the sleep continuity variables in the sleep compression group. For example, Maurer et al. (2018) found some support for the prevention of perpetuating practices, such as spending too much time in bed and taking naps, being a mechanism in sleep restriction therapy. Since the sleep compression procedure is in fact directed towards decreasing time in bed (rather than total sleep time), this could be valid for sleep compression as well. Moreover, sleep restriction is believed to stabilise circadian rhythm mainly through fixed rise time and delayed bedtime (Maurer, Ftouni, et al., 2020;Spielman et al., 2011), although the evidence is still weak (Maurer et al., 2018). Sleep compression also entails a fixed rise time and a gradually delayed bedtime, so this could be a valid mechanism for sleep compression as well. Sleep restriction is also believed to reduce hyperarousal; physiological arousal through the increased homeostatic sleep pressure and cognitive arousal via the experience that is gained by the treatment (Maurer et al., 2018;Maurer et al., 2022 Another important question is whether the improvements seen in objective sleep are clinically meaningful. Total sleep time did not increase at all according to polysomnography assessments. Sleep continuity improved, but the improvements were small and not sustained at week 10. At the same time, insomnia severity improved with over 10 points on the ISI in both groups, which is considered a large and clinically important improvement  and suggests that patients were considerably less burdened after treatment. These results are in line with a recent meta-analysis that did not find any clinical meaningful improvements on polysomnography-assessed sleep variables after either sleep restriction therapy or full CBT-I but still found improvements of insomnia severity that were considered clinically meaningful (Edinger et al., 2021). This raises the intriguing question that insomnia may not fundamentally be a sleeping disorder.
However, that is a question that cannot be answered in this study, but requires much further research.
Future research should use day-to-day assessments with polysomnography to more closely examine mechanisms in sleep restriction and sleep compression therapy, preferably in combination with physiological assessments of hyperarousal, circadian rhythm, and questionnaires covering sleep related beliefs and attitudes. This would give further information on the interplay between these mechanisms, and how they might work differently in sleep compression compared with sleep restriction. With this knowledge, we would have the possibility to fine tune the treatments to obtain the desired effects.

| Strengths and limitations
The main strength with the present study lies in the randomised blind comparison of sleep compression therapy to the gold standard onecomponent treatment for insomnia, sleep restriction therapy. In addition, we use polysomnography to measure sleep, which gives us objective information about the participants' sleep as opposed to subjective assessments. Very few studies report on polysomnography variables in relation to sleep restriction therapy (Mitchell et al., 2019), and we have found no previous studies that use polysomnography assessments during ongoing sleep compression therapy. Another strength with the current study is that we included a clinical sample, allowing comorbidities and medications, which makes it similar to patients in real world clinical settings.
Our study does have limitations. No power calculations were performed a priori, and the sample size is small, so the study may be underpowered to detect small differences. Larger studies on sleep compression therapy need to be performed. The small sample size also allowed the groups to differ in mean age which may have affected the results, although we did control for age in the statistical model. However, we did not control for baseline values of outcome measures in the statistical model. Another limitation is that the sleep compression group showed a tendency to spend a shorter time in bed than their prescribed sleep window, which makes their treatment resemble sleep restriction more than if they had kept more closely to their sleep window. We allowed hypnotic medication during polysomnography assessments, allowing for a possible confounder since hypnotics may affect brain activity during sleep. The frequency of polysomnography assessments may have been too low to capture all of the changes occurring during the treatments. Preferably, polysomnography should have been performed each night during treatment, but this was not feasible due to limited resources and inconvenience for patients. In addition, the person who scored polysomnography not being blind to group allocation is also a limitation in this study. Finally, EEG spectral analysis was not performed. Neither were subjective ratings included, for instance ratings of fatigue. Finally, the generalisability of the results may be restricted since the majority of the participants in our sample were highly educated and had a high socioeconomic status.