Sleep-specific outcomes attributable to digitally delivered cognitive behavioral therapy for insomnia in adults with insomnia and depressive symptoms

ABSTRACT Objective Randomized controlled trials (RCTs) of digitally delivered Cognitive Behavioral Therapy for insomnia (CBT-I) have demonstrated reductions in insomnia severity, depression symptoms, anxiety symptoms, and suicidal ideation. The present study aimed to evaluate the effectiveness of self-guided, digital CBT-I to improve sleep-specific outcomes. Method An RCT of Australian adults with insomnia and depressive symptoms (N = 1149) compared SHUTi, a digital CBT-I intervention, with HealthWatch, an attention-matched control internet program, at baseline, posttest (9 weeks) and at 6-, 12-, and 18-month follow-ups. Online sleep diaries were used to derive measures of sleep-onset latency (SOL), wake after sleep onset (WASO), sleep efficiency (SE), number of awakenings, sleep quality, and total sleep time (TST). Results Participants in the SHUTi condition had greater improvements at posttest compared with control for: SOL, WASO, SE, number of awakenings, and sleep quality. These improvements were sustained at every follow-up (p < .02 for all outcomes except TST, in which statistically significant increases were observed only at 12- and 18-months). Conclusions Digitally delivered CBT-I produced lasting improvements in sleep outcomes among adults with insomnia and depressive symptoms. Findings provide further evidence of long-term improvements associated with a digital therapeutic for insomnia, compared to an attention-control condition.

CBT-I is highly amenable to digital delivery, and there is considerable evidence from randomized controlled trials (RCTs) that digitally-delivered CBT-I is effective for treating insomnia (Zachariae et al., 2016, Seyffert et al., 2016, Drake, 2016).RCTs have also demonstrated that digitally-delivered CBT-I can result in reductions in symptoms of depression, anxiety, and suicidal ideation in adults (Ashworth et al., 2015, Christensen et al., 2016, Manber et al., 2008, Wagley et al., 2013, Watanabe et al., 2011).These findings suggest that this mode of delivering CBT-I may be an important approach for scaling up treatment for insomnia in the community (Zachariae et al., 2016) and for providing a transdiagnostic approach to reducing symptoms of common mental disorders (Harvey, 2008;Peter et al., 2019, Cliffe et al., 2020).
A number of studies have investigated the specific aspects of sleep that are modified by digitallydelivered CBT-I.Key indicators of insomnia (Buysse et al., 2006;Carney et al., 2012) include sleeponset latency (SOL), wake after sleep onset (WASO), sleep efficiency (SE), number of awakenings, sleep quality, and total sleep time (TST).Investigating the effects of digitally delivered CBT-I on these sleep indicators may provide critical information on strengths and gaps of specific interventions and identify strategies for optimizing the effects of such CBT-I programs.Research may also identify whether the effects of digitally delivered CBT-I may be different for specific subgroups of the population, such as individuals with depression or anxiety symptoms.While there are digitally delivered CBT-I programs that have been shown to have positive effects on each of these sleep indicators, the focus of previous RCTs has been on adults with primary insomnia (Zachariae et al., 2016).To our knowledge, no RCTs have tested the effects of digitally delivered CBT-I on goldstandard sleep diary indicators in a sample of adults with both insomnia and elevated depression symptoms, nor done so in comparison to an attention-control condition.
This study evaluated the effectiveness of digital CBT-I -using SHUTi -to improve sleep outcomes, as measured by prospectively entered sleep diaries, in a sample of Australian adults with insomnia and depressive symptoms.The primary outcomes of this randomized controlled trial (N = 1149) have already been published and demonstrated that SHUTi was effective in reducing insomnia severity at posttest (based on the Insomnia Severity Index, ISI: d = 1.10), along with reductions in depression symptoms (d = 0.69) and anxiety symptoms (d = 0.50), with maintenance of gains demonstrated for a period of 18 months (Batterham et al., 2017;Christensen et al., 2016).Suicidal ideation was also reduced at immediate posttest.However, the ISI provides a global indicator of severity, without providing a detailed understanding of the types of sleep improvement that may have resulted from the intervention.Identifying the specific aspects of sleep that changed as a result of the intervention may provide stronger evidence for its use and identify areas for improvement.In this paper, we report the effects of SHUTi on the sleep outcomes of SOL, WASO, SE, number of awakenings, sleep quality, and TST.These outcomes were assessed from sleep diary data collected at the assessment periods, but were not the focus of the primary analyses.Although we hypothesized that participants receiving the SHUTi intervention would have significantly better outcomes on all sleep indicators at posttest relative to participants who received an attention-control program, SOL and WASO are the main sleep variables of interest.

Participants and procedure
The details of the study methodology are provided in the previously published protocol paper (Gosling et al., 2014) and primary outcome paper (Christensen et al., 2016).The trial was prospectively registered with the Australian New Zealand Clinical Trials Registry, number ACTRN12611000121965.
Australian adults aged 18-64 were recruited through the internet using online advertisements (e.g., Facebook), websites of sleep or mental health associations, and media releases.Initial inclusion criteria, via online screening, were subclinical depression based on scores of 4-20 on the Patient Health Questionnaire-9 (PHQ-9 (Spitzer et al., 1999) and the presence of insomnia assessed using the Bergen Insomnia Scale (Pallesen et al., 2008).Subsequently, a telephone-administered Mini International Neuropsychiatric Interview (MINI (Sheehan et al., 1998) was conducted to confirm absence of current major depressive disorder or lifetime bipolar disorder, and diagnostic confirmation of insomnia.Participants also had to meet criteria for insomnia on Morin's modified diagnostic insomnia interview (Morin, 1993) during the telephone interview.
Exclusion criteria included shift work, pregnancy, or work, family, or other commitments that interfered with regular nighttime sleep patterns; time of awakening outside the hours of 0400 h and 1000 h; bedtime outside the hours of 2000 h and 0200 h more than twice a week; absence of reliable internet access; difficulty reading English; diagnosis of psychosis, schizophrenia, or bipolar disorder by a psychiatrist (based on self-report); current involvement in a non-medication treatment program for insomnia with a health professional; presence of an untreated sleep disorder other than insomnia; medication changes in the past three months; a medical disorder accounting for insomnia; or suicidal plans or attempts in the previous two weeks identified in the MINI interview.Participants were then required to complete ten online sleep diaries within a 14-day period, within 21 days of sleep-diary commencement, in order to proceed to randomization.From the 574 and 575 participants randomized to SHUTi and HealthWatch respectively, 124 (22%) and 131 (23%) participants from the two conditions completed posttest diary data respectively, 161 (28%) and 201 (35%) completed 6-month diary data, 130 (23%) and 191 (33%) completed at 12 months, and 77 (13%) and 98 (17%) completed 18 month diary data.The study was approved by the Australian National University Human Research Ethics Committee (protocol number 2011/041) and all participants provided informed consent.
Immediately after completion of the pre-assessment sleep-diary, computer-generated randomization(stratified by age and sex was conducted within the trial management software.Participants were randomly assigned (1:1) to receive the active intervention, SHUTi, or an attention-matched control, HealthWatch (respective web programs described below).Telephone-based interviewers, statisticians, and chief investigators were masked to group allocation.Once a participant was randomly assigned, an automated e-mail contact regimen was implemented to remind participants about completion of sleep diaries, availability of program modules or "cores", and assessments.In addition, a telephone call was made at the time of the 6-month follow-up to conduct an additional MINI assessment and provide a further reminder to complete the assessment.

Interventions
SHUTi is a fully automated, interactive online insomnia treatment program based on CBT-I.The intervention period for this trial was 9 weeks.SHUTi was accessible via a web-browser on the patient's internet-connected device.A native mobile application version of SHUTi (Somryst Ⓡ ) has since been authorized by the US Food and Drug Administration (FDA) as the first prescription digital therapeutic for treating chronic insomnia.SHUTi provides six sequential treatment modules or "cores" consisting of an overview of insomnia, two behavioral cores focusing on sleep restriction and stimulus control, cognitive restructuring, environmental and lifestyle factors known to disrupt sleep (sleep hygiene), and relapse prevention.Completion of an 11-item daily sleep diary (Carney et al., 2012) for at least five of seven days was required to advance from the first to the second core, so the system can establish an algorithmically defined sleep window.
HealthWatch is an interactive, attention-matched, internet-based placebo control program, providing public-domain health information that is not specific to mental health or sleep-related content.The program has previously been shown to provide no therapeutic reductions in depression, and can therefore be regarded as a highly plausible control condition (Glozier et al., 2013).Modules provide information about environmental health, nutrition, heart health, activity, medication, oral health, blood pressure and cholesterol, calcium, and back pain, in addition to weekly surveys on these topics.
The approximate time taken to complete the program was matched to SHUTi, and similar completion rates were obtained (58% SHUTi vs 52% HealthWatch) (Christensen et al., 2016).

Measures
Participants prospectively entered daily sleep diaries during a two-week period at each of five assessment periods.The 11-item diary included questions as recommended by the Consensus Sleep Diary Panel (Carney et al., 2012).As noted elsewhere, prospectively collected selfreported sleep dairies are a proven methodology to assess insomnia and track treatment, as they include patient perceptions of sleep and impairment which are a necessary part of diagnosis and treatment progress tracking (Buysse et al., 2006).Prospective sleep diaries, along with the Insomnia Severity Index (Bastien et al., 2001;Morin et al., 2011) are considered the standard assessments for tracking insomnia treatment progress (Buysse et al., 2006;Carney et al., 2012).
Five common sleep variables were derived from sleep diaries to yield a single averaged metric for each assessment period across the 14 days of sleep diaries.Sleep onset latency (SOL) is the minutes spent between trying to fall asleep and actually falling asleep.Wake after sleep onset (WASO) represents the minutes spent awake during the night after falling asleep for the first time.Sleep efficiency (SE) is the amount of time asleep, divided by amount of time in bed, multiplied by 100.This yields a percentage between 0-100%, with greater scores reflecting better sleep efficiency.As the term applies, number of awakenings (NOA) counts the number of times a person wakes up during the night between the time they fell asleep and time they woke up for the day.Total Sleep Time (TST) represents all of the time spent in bed and then subtracts minutes awake.

Analysis
Mixed model repeated measures ANOVA (MMRM) were used to estimate the effects of the intervention on change in sleep outcomes from baseline, accounting for intervention condition, time, and the critical interaction between time and condition.MMRM analyses account for all available data for participants in the trial.This approach yields unbiased estimates of intervention effects under the missing-at-random assumption.An unstructured variance-covariance matrix was used and degrees of freedom (df) were estimated with Satterthwaite's correction.SAS Ⓡ , Version 9.3 or higher (SAS Institute, Cary NC) was used for all analyses.

Results
The sample consisted of approximately three quarters women, with a mean age of 43, and approximately 60% married or in a de facto relationship.By design, symptoms of depression and anxiety were elevated but typically below clinical cut-points on the PHQ-9 and GAD-7, while mean insomnia severity scores (ISI) were in the "moderate" severity range.Approximately 9% reported suicidal ideation and 6% reported a suicide attempt in the six months before the baseline assessment.
Further details of the sample characteristics are available in the primary outcomes paper (Christensen et al., 2016).
The outcomes of the MMRM models are presented in Table 1.All sleep diary outcomes were significantly improved in the intervention group relative to baseline, compared to the attention control group, at all-time points.The only exception was for TST, where the difference was only significant at 12-and 18-months (d = 0.25, 0.33 respectively), but not at posttest or 6-months (d = 0.21, 0.12 respectively).Between-group effect sizes (Cohen's d) for observed changes were small to large at posttest: 0.68 for SOL, 0.55 for WASO, 0.82 for SE, 0.38 for number of awakenings, 0.83 for sleep quality and 0.21 for TST (not significant); and largely maintained at 12-months: 0.45 for SOL, 0.37 for  WASO, 0.49 for SE, 0.23 for number of awakenings, 0.83 for sleep quality and 0.25 for TST. Figure 1 shows the change over time for each of the sleep diary outcomes (see Supplemental Table 1 for the related MMRM ANOVA estimates).

Discussion
The findings of this study indicate that SHUTi, a self-guided digitally-delivered CBT-I intervention, was effective, based on statistically significant improvements in all sleep diary indicators of sleep disturbance with moderate to large effect sizes.The single exception was TST, which showed a delayed effect, typical of CBT-I outcome studies (Scott et al., 2022;Trauer et al., 2015).As the effect sizes for all outcomes except TST and number of awakenings were moderate to large, this may indicate that changes were clinically meaningful.In fact, 30 minutes of SOL and WASO are used in clinical guidelines for determining the presence of insomnia (i.e., <30 minutes of SOL and WASO are considered normal sleep patterns).As can be seen in Figure 1, those who were in the SHUTi group experienced less than 30 minutes of SOL and WASO at all post and follow-up time-points (except WASO at 18-months).Similarly, 85% sleep efficiency is considered in the "normal" range, and, again, in Figure 1, SHUTi participants averaged above 85% at all post and follow-up time points.In addition, the SHUTi participants had significantly better outcomes than those in the control condition for SOL, WASO, and SE at every post and follow-up time-point.These results suggest that SHUTi effectively reduced insomnia by modifying all forms of sleep disturbance, including reduced sleep latency and reduced wake after sleep onset -two of the primary indicators of treatment improvement.Participants in the SHUTi condition also reported greater improvements in sleep quality and sleep efficiency at posttest and all follow-up assessments relative to participants in the control condition.These findings are broadly consistent with the outcomes for some, but not all, digitally-delivered CBT-I programs (Zachariae et al., 2016).A previous systematic review reported moderate effects at follow-up for insomnia severity, sleep efficiency and sleep quality but no significant improvements in SOL, WASO and TST (Zachariae et al., 2016).The SHUTi program may be particularly effective in reducing sleep disturbance, with active components that are needed to reduce problems with sleep onset, sleep maintenance, sleep efficiency and subjective sleep quality.This clinical trial was the first to test the effect of digitally delivered CBT-I on sleep diary-derived outcomes in an indicated sample with elevated depressive symptoms.The findings are comparable to those of previous trials examining the efficacy of SHUTi in samples selected on the basis of insomnia symptoms but not depression symptoms (Ritterband et al., 2009(Ritterband et al., , 2012)).This equivalence suggests that digitally delivered CBT-I is effective irrespective of whether patients have symptoms of depression or not.The findings are consistent with a growing body of literature that suggests face-to-face CBT-I is effective regardless of cooccurring mental health symptoms (Okajima et al., 2011;Trauer et al., 2015, Geiger-Brown et al., 2015, Wu et al., 2015).Further research would benefit from using an implementation science framework to expand the delivery of CBT-I across diverse clinical and non-clinical settings, to maximize the public health benefit of this therapy (Koffel et al., 2018).In addition, identifying other clinical outcomes related to the potential benefits of digitally-delivered CBT-I should also be explored (Vitiello et al., 2013) given the associations of insomnia with an extensive range of physical and mental comorbidities (Fernandez-Mendoza & Vgontzas, 2013;Sivertsen et al., 2009).
While this was the first study to examine multiple sleep outcomes of digitally delivered CBTI in a sample with insomnia and depression symptoms, there are some limitations to be noted.First, sleep outcomes were only assessed using self-report sleep diaries.However, this method is robust (Carney et al., 2012;Haythornthwaite et al., 1991) and feasible in the context of a digitallydelivered public health intervention.Moreover, sleep actigraphy may be inaccurate due to challenges in detecting wake patterns across diverse populations, while subjective perception of sleep provides valuable and clinically meaningful data (Carney et al., 2012;Marino et al., 2013).Second, attrition from the trial was high, as is often the case in digitally delivered trials.While the statistical analyses accounted robustly for all available data under the missing-at-random assumption, differential attrition may have influenced outcomes.Finally, while the sample had subclinical depressive symptoms, direct conclusions cannot be drawn about effectiveness for groups with more severe levels of depression.

Conclusions
Fully automated digitally delivered CBT-I was effective in improving multiple sleep outcomes in adults with insomnia and subclinical symptoms of depression.These findings reinforce the primary outcomes of the trial, indicating that digitally delivered CBT-I offers effective and highly accessible treatment that addresses a range of sleep variables, including sleep onset latency and wake after sleep onset, with subsidiary prevention effects for depression symptoms.

Figure 1 .
Figure 1.Change in sleep diary-derived outcomes over time t .t The dotted lines in the SOL, WASO and SE figures indicate clinical guidelines for determining the presence of insomnia (i.e., <30 minutes of SOL and WASO are considered normal sleep patterns; >85% is considered normal SE).

Table 1 .
Mixed model repeated measures ANOVA estimates of intervention effects on sleep outcomes at each time point relative to baseline.