Stimulus control for insomnia: A systematic review and meta‐analysis

Stimulus control (SC) is commonly viewed as an evidence‐based treatment for insomnia, but it has not been evaluated comprehensively with modern review and meta‐analytic techniques. The aim of the current study was thus to perform a systematic review and meta‐analysis of trials that examine the efficacy of stimulus control for insomnia. A systematic search for eligible articles and dissertations was conducted in six online bibliographic databases. The 11 included studies, with the majority published between 1978 and 1998, were randomised controlled and experimental studies in adults, comparing stimulus control for insomnia with passive and active comparators and assessing insomnia symptoms as outcomes. A random effects model was used to determine the standardised mean difference Hedge's g at post‐treatment and follow‐up for three sleep diary measures: the number of awakenings, sleep onset latency, and total sleep time. A test for heterogeneity was conducted, forest plots were produced, the risk of publication bias was estimated, and the study quality was assessed. In the trials identified, stimulus control resulted in small to large improvements on sleep onset latency and total sleep time, relative to passive comparators (g = 0.38–0.85). Compared with active comparators, the improvements following stimulus control were negligible (g = 0.06–0.30). Although methodological uncertainties were observed in the included trials, stimulus control appears to be an efficacious treatment for insomnia when compared with passive comparators and with similar effects to active comparators. More robust studies are, however, warranted before stronger conclusions are possible to infer.

assessed.In the trials identified, stimulus control resulted in small to large improvements on sleep onset latency and total sleep time, relative to passive comparators (g = 0.38-0.85).Compared with active comparators, the improvements following stimulus control were negligible (g = 0.06-0.30).Although methodological uncertainties were observed in the included trials, stimulus control appears to be an efficacious treatment for insomnia when compared with passive comparators and with similar effects to active comparators.More robust studies are, however, warranted before stronger conclusions are possible to infer.
adults, behaviour therapy, cognitive behaviour therapy, efficacy, sleep disturbance

| INTRODUCTION
Insomnia disorder is characterised by difficulty in initiating or maintaining sleep accompanied by daytime impairments (Morin et al., 2015;Riemann et al., 2022).Approximately 10% of the general population meets the criteria for insomnia disorder, and the condition is often chronic for many years.Insomnia also has a high burden of disease.Individuals with insomnia commonly report a decline in cognitive abilities, mood swings, comorbid health problems, and they can develop comorbid disorders in the long-term, such as depression and heart disease (Hertenstein et al., 2019(Hertenstein et al., , 2023;;Sofi et al., 2014).The societal costs of insomnia are substantial, mainly due to absenteeism and poor work productivity (Daley et al., 2009).The high prevalence, chronic course, and negative health effects associated with insomnia disorder highlight the importance of effective treatments.
In the beginning of the 1970s, Bootzin (1972) proposed an operant analysis of insomnia.In this analysis, he suggested that falling asleep can be viewed as an operant behaviour emitted to produce reinforcement.This notion implies that stimuli associated with sleep (e.g., heavy eyelids and darkness) may become discriminative stimuli for the manifestation of reinforcement.As a result, difficulty in falling asleep may be a consequence of inadequate stimulus control.Potent discriminative stimuli may not have been established, such as sleepiness cues, and/or discriminative stimuli for behaviours incompatible with sleep may be extant, e.g.consuming news on a smartphone in bed.Further, Bootzin (1972) suggested that, based on the notion that falling asleep is an emitted response, it could be possible to increase the frequency of sleep initiation by reinforcing it.
Following from the operant analysis, Bootzin (1972) proposed a new treatment for insomnia, namely: stimulus control (SC).In the original framework, stimulus control consisted of four rules: (1) "Lie down, intending to go to sleep only when you are sleepy"; (2) "Do not read or watch television in the bedroom"; (3) "If you find yourself unable to fall asleep, get up and go into another room.Stay up as long as you wish and then return to the bedroom to sleep"; (4) "If you still cannot fall asleep, repeat Step 3. Do this as often as is necessary throughout the night".During the 1970s and 1980s, stimulus control was further developed by Bootzin and colleagues (Bootzin, 1979;Bootzin et al., 1983;Bootzin & Nicassio, 1978).While rules 1 and 4 remained the same into the 1980s, rules 2 and 3 were expanded.Rule 2 was developed into: "Do not use your bed for anything except sleep; that is, do not read, watch television, eat, or worry in bed.Sexual activity is the only exception to this rule.On such occasions, the instructions are to be followed afterward when you intend to go to sleep".Rule 3 was expanded into: "If you find yourself unable to fall asleep, get up and go into another room.Stay up as long as you wish and then return to the bedroom to sleep.Although we do not want you to watch the clock, we want you to get out of bed if you do not fall asleep immediately.Remember, the goal is to associate your bed with falling asleep quickly!If you are in bed more than about 10 minutes without falling asleep and have not gotten up, you are not following this instruction".
In addition, two new rules were introduced: (5) "Set your alarm and get up at the same time every morning irrespective of how much sleep you got during the night.This will help your body acquire a consistent sleep rhythm" and (6) "Do not nap during the day".This latter, the expanded stimulus control framework has over the years been the most often used in trials investigating the efficacy of multicomponent cognitive behavioural therapy for insomnia (CBT-I) (van Straten et al., 2018).
The efficacy of stimulus control as a standalone treatment for insomnia disorder has been explored in three reviews since the 1990s.In 1999, the American Academy of Sleep Medicine (AASM) (Morin et al., 1999) identified 12 studies investigating the efficacy of stimulus control and stated, based on the criteria developed by the American Psychological Association (Chambless & Hollon, 1998), that stimulus control is an empirically supported intervention.In 2006, the AASM based their updated report on psychological and behavioural treatments for insomnia on six studies and again classified stimulus control as a well-established treatment (Morin et al., 2006).The first two AASM reports were, however, associated with several limitations, i.e., database searches were limited up until 2004, only two databases were used, treatment efficacy was only evaluated using night-time symptoms, detailed methodological and statistical information was not provided in the quantitative assessment of stimulus control, and the criteria that were used for establishing empirically supported treatments have since been abandoned (e.g., on the grounds that studies not showing efficacy are not taken into account as part of the evaluation).
In the latest report from the AASM, two databases were used to identify trials examining the efficacy of stimulus control.Several notable revisions were made to provide more robust recommendations (Edinger et al., 2021).In particular, this report examined trials up until 2020, included one daytime symptom as an outcome (i.e., fatigue), provided detailed methodological and statistical information, and used the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) assessment system as a way to evaluate evidence of treatment efficacy, e.g., by taking into account the quality of evidence, adverse events, and patient preferences.Eight studies were included in the latest AASM report.As a general conclusion, the AASM report from 2021 rated the overall quality of evidence for SC as low for the critical outcomes (i.e., sleep quality, sleep latency, wake after sleep onset, and remission rate) because of imprecision and risk of bias.However, the report also stated that the balance of benefits versus harms is clearly in favour of stimulus control.
Apart from a limited number of databases and an almost exclusive focus on night-time symptoms in the three AASM reports, another methodological limitation is notable.Five of the included trials in the AASM reports have combined stimulus control with sleep hygiene (Epstein et al., 2012;Harris et al., 2012;Lacks, Bertelson, Gans, & Kunkel, 1983;Schoicket et al., 1988;Sidani et al., 2019), an intervention with small to medium effect sizes on sleep diary variables and global insomnia questionnaires, e.g., the Insomnia Severity Index (Chung et al., 2018).The inclusion of trials that combine stimulus control with sleep hygiene is therefore a potential threat to the overall conclusion of stimulus control in the AASM reports.As a result, it is possible that a new review and meta-analysis including all published studies on stimulus control as a standalone treatment might produce different results.
Although the efficacy of stimulus control has previously been systematically reviewed in the three AASM reports and meta-analysed in two reports (Edinger et al., 2021;Morin et al., 1999), the purpose of the current investigation was to conduct a comprehensive systematic review and meta-analysis of clinical trials that examine the efficacy of stimulus control as a standalone treatment for insomnia.The aim was to investigate the efficacy of stimulus control on both night-time and daytime symptoms.An additional aim was to review treatment-related parameters for stimulus control, such as content and delivery of stimulus control, to understand how stimulus control has been administered in published trials.

| METHODS
A systematic review approach in combination with meta-analytical calculations to aggregate outcome measures were used to estimate the efficacy of stimulus control.The review followed the PRISMA guidelines (Moher et al., 2009) and was pre-registered at the International Prospective Register of Systematic Reviews (PROSPERO; http://www.crd.york.ac.uk/PROSPERO;CRD42020149291).

| Search strategy
In November 2022, database searches were conducted by four project-independent librarians at Karolinska Institutet University Library, Stockholm, Sweden to identify trials that have investigated the efficacy of stimulus control for insomnia.Six online bibliographic databases were used in the search process [i.e., Medline (Ovid), Psycinfo (Ovid), Embase (Elsevier), Cinahl (Ebsco), Web of Science (Clarivate), and Dissertations and Theses (ProQuest)].The librarians and the first author developed the search strategy based on several MeSH and keyword search terms.The terms were employed to identify trials in which participants with insomnia had been enrolled and stimulus control had been used as a treatment intervention.A detailed description of the search strategy can be seen in the Supporting Information (Table S1).Further, the first author hand-searched the reference lists of recent systematic reviews and meta-analyses that have evaluated the efficacy of CBT-I, and the reference lists of the studies included in the current review.

| Selection procedure
In Figure 1, the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) study inclusion flowchart is displayed.
As can be seen in the figure, the database searches returned 3352 records, of which 1644 titles were duplicates.Thus, 1708 records were the focus for further review.The inclusion criteria for the trials in the present review were: a. the study was a randomised controlled trial or used an experimental design, from which outcome data per group could be extracted; e. the participants were adults (i.e., 18 years or older); f. the participants had been diagnosed with insomnia (insomnia disorder or primary, secondary or comorbid insomnia), reported undiagnosed problems with initiating or maintaining sleep, or reported poor sleep (e.g., scoring above a cut-off on a validated insomnia scale); g. the study was published before or in November 2022; and h. the study was published in English.
Concerning criterion (f), we included studies in which the participants were classified as having insomnia symptoms or poor sleep (i.e., not meeting all criteria for insomnia disorder) based on two main reasons: (1) the mounting evidence for, and development in diagnostic systems towards, a dimensional view of mental health problems (Hankin et al., 2005;van Os et al., 1999), and (2) previous research on psychological treatments (e.g., CBT-I) showing reductions in insomnia symptoms among those without fulfilling all criteria for insomnia disorder, e.g., among those with subthreshold insomnia (Denis et al., 2020), insomnia symptoms (Swift et al., 2012), and acute insomnia (Randall et al., 2018).A priori plans also included sensitivity and moderator analyses, if considerable heterogeneity was observed concerning criterion (f).
As a first step, the 1708 records were screened using a web application designed for systematic reviews (https://rayyan.qcri.org).The records' abstracts were first independently reviewed by the first and second authors (MJF and LN) to exclude irrelevant studies.The paper was read in full text by the first and second authors if it was uncertain whether a record met the criteria after screening the abstract.Disagreements were solved with discussions between the two authors until a shared conclusion was reached.Based on the inclusion criteria described above in the abstract screening phase, 1628 papers were excluded.The remaining 80 records were reviewed by the first and second authors in full texts.The same inclusion criteria defined above were employed during the full text screening phase.In all, 11 papers were included at this stage.The excluded 69 records with reasons for exclusion are presented in Supporting Information: Table S2.

| Study quality assessment
The study quality assessments were carried out using standardised criteria (Kmet et al., 2004) by the third and the fourth authors (SA and BB).The assessment focussed on the degree to which design, conduct, and analyses reduce errors and biases for randomised controlled trials and experimental designs.The assessment involved 14 items: objective, design, method, subjects, intervention, blinding of investigators to intervention, blinding of subjects to intervention, outcome, sample size, analytic methods, variance, confounds, description of results, and conclusion.The 14 items were rated on a three-point rating scale (yes = 2, partially = 1, and no = 0).The items are in line with recommendations from the Centre for Reviews and Dissemination for systematic reviews (Akers et al., 2009) and with previous trial assessment approaches in the area of insomnia disorder (Miller et al., 2014;Morin et al., 2006).To enable comparisons across trials, a total score and a percentage were calculated for each study.To improve the reliability of the quality scores, the two raters were introduced to the items by reading the criteria and discussing them together with the first author.Also, the two authors conducted their ratings independently.Disagreements between the two raters (on average 6% per item) were discussed until a negotiated score was decided upon for the included studies.

| Data analysis
Data from the systematic review were used in meta-analyses for each respective outcome.Standardised mean differences at post-treatment (k = 5) and follow-up (k = 2), divided by the pooled standard deviations, were used to calculate the aggregated effects Hedge's g between stimulus control and its comparators.However, since the types of comparators varied in terms of their features, meta-analyses were performed separately for those that were considered passive (i.e., wait-list and no treatment) and active (e.g., paradoxical intention, placebo, and stimulus control parts, e.g., temporal and spatial components).Because some studies included different parts of stimulus control or comparators in the same trial, these were combined into one condition by summing the number of participants and pooling means and standard deviations.Furthermore, 95% confidence intervals (CIs) are presented for both the individual and aggregated effects.According to Cohen (1988), gs 0.20-0.50are considered a small effect, 0.50-0.80 a medium effect, and >0.80 a large effect.The effect sizes were, however, compared with other estimates in the literature to determine their clinical utility (Cumming & Finch, 2001).
A test for heterogeneity was performed to investigate betweenstudy variation, using the I 2 -statistic to assess statistical variation (25%, 50%, and 75% indicating low, medium, and high heterogeneity), as well as the Q-statistic to determine if this heterogeneity was significant (Borenstein et al., 2011).Because the I 2 -statistic is considered imprecise when the number of studies is low, 95% CIs were calculated for the I 2 .Forest plots were produced to display the between-group effect sizes for each study when deemed relevant, and the potential risk of publication bias was determined using the Egger's tests (Egger et al., 1997).All statistical analyses were conducted using R and the metafor package (Viechtbauer, 2010), implementing a random effects model as between-study variation was expected.

| Systematic review
After the screening stage, 11 studies were included (see Table 1).All All included trials enrolled participants from the community, and two studies also recruited from clinical settings (Lacks, Bertelson, Sugerman, & Kunkel, 1983;Puder et al., 1983).The sample sizes in the studies ranged from 15 to 70 (total sample size: 363).Diagnostic status was slightly different across trials; in eight studies, the participants were categorised as fulfilling diagnostic criteria for primary insomnia, whereas in the remaining trials, the participants were classi- In the original study, there were four conditions; the two groups using sleep medication and being administered medication withdrawal as part of the study were excluded from the current review to avoid concurrent interactions between two active interventions (i.e., stimulus control and sleep medication withdrawal).
with initiating sleep (one study).None of the 11 trials evaluated psychiatric or somatic comorbidity using a structured procedure.There was variation in the participants' age and gender across trials; the mean age ranged from 19 to 67 years, and most of the participants were women (47%-100%).
Ten (5) the sample size was not deemed appropriate.The ratings of each of the 14 items for each included study are displayed in Supporting Information Table S3.
In Table 2, the treatment-related parameters for stimulus control are described for each included study.The content of stimulus control varied between the original and expanded frameworks across studies.The most common approach employed the expanded framework for stimulus control (8 out of 11 trials).Stimulus control was delivered as individual therapy in four studies (1-4 sessions), as group therapy in four trials (4-6 sessions), and in a self-help format in one study (8 weeks).In four out of 11 trials, stimulus control was delivered according to a manual.The therapist professional was typically a graduate or doctoral student, but in some studies it was not reported at all or not applicable (as stimulus control was delivered in a self-help format).

| Meta-analysis: Night-time symptoms
The aggregated findings for stimulus control at post-treatment and follow-up were separated depending on the type of comparator used, i.e., passive (Table 3) and active (Table 4)  comparator), and the smallest for TST, g = 0.06 (versus an active comparator).Compared with passive comparators, stimulus control produced significant and large improvements on SOL at post-treatment and follow-up.Also, stimulus control resulted in a non-significant and small effect size on TST at post-treatment, relative to passive comparators.Compared with active comparators, stimulus control resulted in non-significant changes, i.e., negligible to small effect sizes on SOL, and negligible effect sizes on TST and NAW.Forest plots comparing stimulus control with passive and active comparators on SOL at posttreatment are presented in Figures 2 and 3.
Heterogeneity varied from non-significant (0.00%) to high (70.96%),but no direct pattern appeared with regard to its nature.
Also, due to the limited number of comparisons, sub-group analyses were considered not viable to perform.Regarding possible publication bias, the Egger's test was only significant in one case (SOL when compared with active comparators at post-treatment).

| Narrative review: Daytime symptoms
In four of the trials, daytime symptoms were assessed.Due to the limited number of trials and heterogeneity in the type of daytime outcomes, it was not reasonable to estimate effects using meta-analysis.
Instead, a narrative approach was selected.One of the four trials were not possible to use since two groups were excluded in this review, and the presented main effect analyses in the study are thus not relevant as a source for estimating efficacy of SC on daytime symptoms (Riedel et al., 1998).In two of the remaining three trials using daytime symptoms as outcomes, stimulus control was significantly more effective in reducing depressive symptoms, relative to no treatment and imagery training (Espie et al., 1989;Morin & Azrin, 1987).However, the third trial did not show a superior effect of SC on depressive symptoms (Morin & Azrin, 1988).Relying on statistical significance in the three trials as a benchmark, there was no evidence that stimulus control was more effective than the control groups (relaxation, placebo, no treatment, imagery training, and waitlist) on anxiety symptoms.

| Summary of main results
The current investigation is the first exhaustive systematic review and meta-analysis of the efficacy of stimulus control for insomnia.The current findings indicate that stimulus control, relative to passive comparators, produces large effects on SOL ( g = 0.85-0.87)and small effects, albeit non-significant, on TST (g = 0.38).Relative to active comparators, stimulus control resulted in non-significant improvements, i.e., a negligible improvement on SOL at post-treatment, a small deterioration on SOL at follow-up, and negligible improvements on TST and NAW at post-treatment.Compared with previous reviews, the current investigation extends the quantitative assessment of stimulus control as an evidence-based treatment as it compared stimulus control with passive versus active comparators, excluded trials that combined stimulus control with sleep hygiene, and included both night-time and daytime symptoms (Edinger et al., 2021;Morin et al., 1999Morin et al., , 2006)).
To understand the significance of the effect sizes, it is advisable to compare such sizes with other relevant estimates in the literature (Cumming & Finch, 2001).One highly relevant source of estimates is Another limitation concerns the characteristics of the participants.The total sample size was limited to 363 participants, and only two trials reported that power calculations were made before the study started (Epstein et al., 2012;Harris et al., 2012).Type-2-errors are therefore a risk in the majority of trials.All studies recruited participants from the community, meaning that the current results may be less generalisable to patients seeking care in health settings (Davidson et al., 2009).Further, in nearly all of the trials, we classified the participants as meeting the criteria for primary insomnia.As a result, it is unclear whether SC should be regarded as an efficacious therapy for other forms of insomnia, particularly comorbid insomnia.Another limitation concerning diagnostic status is that comorbidity was not formally assessed in the included trials.Although a few studies employed criteria to assess comorbidity or to exclude comorbid participants, the lack of validated assessment of psychiatric and somatic conditions restricts the generalisability, as comorbid problems are more prevalent than insomnia without comorbidity (Stepanski & Rybarczyk, 2006).
Another area of methodological uncertainty regards the administration of SC.Although the expanded stimulus control framework was used in the majority of trials, three studies used the original approach (Bootzin, 1972).Further, the delivery format was diverse, with individual, self-help, and group formats used in the trials, and the number of sessions varied (1-6 sessions).Also, it was not the norm that sufficient information was provided regarding other treatment-related parameters; this concerned whether a manual for stimulus control administration was employed, who the therapists were delivering stimulus control, training and supervision of the therapists, and whether treatment integrity was checked.Due to the limited number of trials in the current review, an examination of potential differences between administration formats was not possible.As in our recent review of paradoxical intention for insomnia (Jansson-Fröjmark et al., 2022), other treatment-relevant features that might have importance for the interpretation of findings was non-existent in the included studies, such as acceptability, adherence, credibility and expectancy ratings, perceived usefulness, and adverse events.
An inclusion criterion for the current review was that studies must report night-time, insomnia symptomatology as outcomes.While 10 studies reported on sleep diary outcomes, none of the trials presented objective sleep outcomes.Other study outcomes related to mental health symptoms and daytime functioning were only reported upon in four studies.It is also noteworthy that generic insomnia scales, such as the Insomnia Severity Index (Bastien et al., 2001), were only used in one of the included studies, due to that extant insomnia scales were not available when the large majority of stimulus control part, it is easy to view stimulus control as an integral component.Or should the findings be interpreted as if stimulus control could be used as a standalone treatment for those with insomnia disorder?The current findings could be interpreted as evidence for the efficacy of stimulus control as a standalone therapy.However, due to uncertainties that have been underscored in the current review concerning the effects for stimulus control, we believe that more research is needed in more well-powered trials with a wide array of clinical outcomes before stimulus control should be viewed as a standalone treatment.
At the same time, stimulus control might play a role in some cases as a sole therapy.For example, if a patient describes poor discriminative stimuli (e.g., regularly allowing the bed to be a context in which nonsleep activities are performed), stimulus control could be a main option in treatment.Future research could investigate whether there are subgroups of patients who benefit more from stimulus control; based on Richard R Bootzin (1972), patients displaying weak discriminative stimuli for sleep and potent discriminative stimuli for behaviours incompatible with sleep could be one important subgroup.

| CONCLUSIONS
Although previous research is limited and methodological limitations are apparent, it is reasonable to view stimulus control as an effective treatment approach for insomnia, particularly so for decreasing sleep initiation difficulties.We recommend that stimulus control be used in clinical settings for patients with insomnia disorder, but also that more research is devoted to exploring the efficacy and implementation of stimulus control.
study reported treatment outcomes using at least one outcome assessing night-time, insomnia symptomatology [e.g., sleep onset latency (SOL), wake after sleep onset (WASO), number of awakenings (NAW), and early morning awakening (EMA)].If available, other secondary outcomes were also considered in the data extraction and effect estimations [i.e., quantitative self-report scales assessing insomnia symptomatology and daytime symptoms, other sleep diary variables (e.g., total sleep time), and objective sleep assessment]; c.SC was evaluated as a standalone treatment in at least one group; d.SC was compared with other CBT-I therapies and components (e.g., relaxation training), various control conditions (e.g., wait-list control or treatment-as-usual), or other interventions (e.g., pharmacotherapy); During the data extraction phase, the 11 papers were reviewed independently by the third and the fourth authors (SA and BB).The two authors used a standard extraction sheet to extract information about each study: country of origin, type of publication (peer-reviewed paper or thesis), study design (RCT or experimental study), sampling (community sample, clinical sample, or other), sample size, diagnosis [i.e., insomnia disorder/primary insomnia/secondary insomnia/ comorbid insomnia, insomnia symptoms (not fulfilling all criteria for a diagnosis), or poor sleep (no evidence of insomnia symptoms)], psychiatric and somatic comorbidity, mean age of participants, percentage of women, treatment-related parameters for SC and other active treatments (content, number and duration of sessions, administration, manual, therapists, training, supervision, and treatment integrity), outcome measures, and findings.Quantitative night-time, insomnia symptoms (i.e., sleep diary assessment of SOL, WASO, NAW, and EMA) were considered as the primary outcome.Self-report scales assessing insomnia symptomatology (e.g., the Insomnia Severity Index), daytime symptoms (e.g., tiredness and mood), other sleep diary variables (e.g., total sleep time [TST], sleep efficiency [SE], and sleep quality [SQ]), and objective sleep assessment (e.g., polysomnography) were employed as secondary outcomes.Discrepancies between the two authors (on average 5% per variable) were resolved through discussion with the first author (MJF).
11 papers were published in peer-reviewed journals between 1978 and 2022, with only one study being published since 1998.Nine trials were carried out in North America and two in Europe.All 11 trials were randomised controlled trials.Seven of the trials compared stimulus control with a passive condition (i.e., waitlist or no treatment) and nine included an active comparator (i.e., EMG biofeedback, imagery training, paradoxical intention, placebo, pseudo-biofeedback, relaxation, cognitive refocussing, awareness-based control, and stimulus control parts, i.e., temporal and spatial components, counter-control, and non-contingent control).
trials used sleep diary outcomes, consisting of items assessing daily SOL, difficulty falling asleep, WASO, NAW, TST, SE, SQ, sleep enjoyment, and restedness/restfulness on awakening.None of the studies reported on objective sleep outcomes.Five of the trials presented data from questionnaire outcomes, assessing insomnia severity, anxiety and depressive symptoms, daytime functioning, and sleepiness.None of the studies reported on adverse effects of stimulus control.Regarding the study quality, the total score for the included trials ranged from 15 to 22 points (M = 19.4,SD = 2.0) out of 28.The five most predominant methodological limitations were: (1) insufficient description of method of subject selection, (2) lacking description of random allocation to groups, (3) inadequate reporting of blinding of investigators, (4) lacking reporting of blinding of subjects, and , and by outcome.The number of comparisons varied between three (SOL at follow-up, TST at post-treatment, and NAW at post-treatment) and five (SOL at posttreatment).As a whole, the positive effects for SC were greater in comparison with passive rather than active comparators, with the largest being attained for SOL, Hedge's g = 0.86 (versus a passive T A B L E 3 Meta-analyses comparing stimulus control with passive comparators one of the largest and more recent meta-analysis, in which cognitive and behavioural interventions (e.g., CBT-I, relaxation, psychoeducation, and sleep restriction) were compared with passive controls(van Straten et al., 2018).Comparing the effect sizes fromvan Straten et al. (2018) for cognitive and behavioural treatments with the current study's effect sizes for stimulus control relative to passive comparators, the effects were larger in the current study for stimulus control on SOL (0.57 vs. 0.85) andTST (0.16 vs. 0.38).A tentative summary would therefore be to conclude that stimulus control has a slightly larger effect on two key outcomes than other cognitive and Forest plots for sleep onset latency, comparing stimulus control with passive comparators at post-treatment Forest plots for sleep onset latency, comparing stimulus control with active comparators at post-treatment behavioural interventions, but a comparison is not possible on other important outcomes (e.g., WASO, EMA, Insomnia Severity Index, and objective sleep assessment).This conclusion is, however, limited by several methodological issues in the trials exploring efficacy for stimulus control.The relatively small number of studies, the limited number of study participants and outcomes, and other methodological features of the trials hamper the possibility of drawing strong conclusions about the efficacy and generalisability of stimulus control.It is also important to underscore that the effect of stimulus control on TST was small and non-significant (0.38).Finally, since the current study did not estimate the efficacy of stimulus control in a dismantling study or network meta-analysis, comparing different components and/or combinations of components of CBT-I, the conclusion regarding the efficacy of stimulus control must be interpreted with caution.4.2 | Methodological considerations and quality of evidenceIn the 11 studies included in the current review, there were several noteworthy methodological limitations.The study quality assessment displayed that the quality of the studies ranged from 15 to 22 points out of 28, implying a modest study quality.The study quality was predominantly poor in four areas: participant selection, randomisation, blinding of participants, and inappropriate sample sizes.Concerning the study quality, it should be noted that 10 of the 11 stimulus control trials were published up to 1998, an era in the insomnia research field when there was less spotlight on reporting standards and methodological quality.In terms of study design, there was variation between the passive and active comparators used in the comparisons with stimulus control.It is important to mention that seven of the trials compared stimulus control with a passive condition, and that these comparisons do not control for non-specific factors (e.g., attention from a therapist).More specifically, this means that the superiority of stimulus control over passive comparators could be due, at least partly, to factors not related to stimulus control itself.The aggregation of active comparators into one active condition category was created on the rationale that they provided active treatment content.This could, however, have caused misclassification if active comparators with different effects were combined and masked meaningful effect differences.If more studies on stimulus control are published in the future, additional evaluations could investigate the efficacy of different clusters of active comparators against stimulus control.
studies were carried out.Due to the lack of trials determining other outcome measures, all meta-analytical calculations were estimated using sleep diary data assessing SOL, TST, and NAW.As a result, we can only draw conclusions for stimulus control regarding three sleep diary-assessed night-time symptoms.The lack of other vital outcome measures in the meta-analytical calculations is a definite limitation for the interpretation of the efficacy for stimulus control.A related limitation is also that estimations of the long-term efficacy for stimulus control were limited to only SOL and TST.Sensitivity and moderator analyses were not conducted due to the limited number of studies.With more trials, we could, for example, have examined the effects of the addition or removal of lower quality studies and whether insomnia symptoms at baseline and administration of stimulus control might moderate the efficacy of stimulus control.To improve the understanding of stimulus control as a treatment form, there are several vital areas that future research could focus on.As several limitations and uncertainties were identified in the present review, we advise future research to employ active comparators, ad hoc power calculations, samples from health care settings, a variety of insomnia types (including comorbid insomnia), validated assessments of comorbidity, different delivery formats (e.g., digital versus faceto-face and original versus expanded frameworks), and extensive assessments of night-time and daytime symptoms as outcomes.Although not a focus of the current review, it is noteworthy that mechanistic research on stimulus control is non-existent.Richard RBootzin (1972) suggested that stimulus control might reverse the patient's difficulty falling asleep by establishing potent discriminative stimuli for sleep and reducing discriminative stimuli for behaviours incompatible with sleep.However, to our knowledge, there is a lack of research showing that this operant analysis is valid.Basic research could, for example, explore whether sleepiness cues are reduced or missed in those with insomnia and that the use of sleepiness as a cue to try to fall asleep actually decreases SOL.Such research could also examine whether a reduction of reading news on a smartphone in bed (a potential discriminative stimulus for behaviours incompatible with sleep) results in shorter SOL.Not only would such research investigate the validity of Bootzin's model, but also point out which stimuli are important to assess and intervene upon.Based on the results in this review, the question is how stimulus control should be used in the treatment of insomnia disorder.The results in this study demonstrate that stimulus control has clear advantages against passive comparators and similar effects compared with active comparators.The question is, however, what this implies.Do the findings mean that stimulus control should maintain a firm place in CBT-I, a multifactorial intervention with various treatment components combined (van Straten et al., 2018)?Based on the solid evidence-base for CBT-I, with stimulus control often integrated as a Description of included studies examining stimulus control for insomnia Not applicable due to the exclusion of two of the four study groups ZDS, ARS) SC and paradoxical intention > no treatment: SOL.SC > placebo (first five treatment weeks) and paradoxical intention (first three treatment weeks): SOL.SC maintained effect on sleep diary measures at four follow-ups (the last at 17 month).SC and relaxation > no treatment: ZDS Sleep diary (WASO, NAW), scale (BDI, STAI) SC > two other groups: WASO.SC not significantly different from the two other groups on NAW.SC > imagery training: BDI 17 (61%) Sleep diary (SOL, WASO, NAW, TST), scale (BDI, STAI) SC > waitlist: WASO.SC > two other groups: TST.No significant group differences on BDI and STAI.SC not significantly different from imagery training on sleep diary measures at Sleep diary (SOL, WASO, TST, SQ, SE), scale (ESS, BDI, STAI) Description of the delivery of stimulus control in the included studies Information regarding therapist training or supervision was reported in five trials; in those instance, training behavioural procedures, readings, role-plays, and having weekly meetings were used as means.In three studies, treatment integrity was checked, e.g., via audiotape reviews.T A B L E 2