Effects of Digital Cognitive Behavioral Therapy for Insomnia in Heavy Drinkers with Insomnia: A Replication and Extension Study

Abstract

Poor sleep and problematic drinking share a bidirectional relationship. Recent studies have targeted sleep as a novel means of improving drinking outcomes, including a prior pilot trial by our group which showed that a digital insomnia intervention (Sleep Healthy Using the Internet; SHUTi) significantly improved sleep and drinking outcomes to a greater degree than a patient education (PE) program. In the current study we aimed to replicate and extend our prior findings by assessing diary-derived measures of sleep and alcohol use as well as mental health outcomes in a larger sample. 113 heavy drinkers with insomnia were randomly assigned to either the SHUTi (n = 61) or a sleep PE program (n = 52). Participants completed assessments immediately following the intervention period (9-weeks) and again at 3 and 6-months post-intervention. Linear mixed effects models showed that those in the SHUTi condition reported significantly greater improvements in sleep and greater reductions in drinking frequency relative to the PE condition. Additionally, those in the SHUTi condition reported greater reductions in depression symptoms and sleep medication use relative to the PE condition. Limitations include relatively high attrition, reliance on self-report measures, and demographic homogeneity. These findings replicate and extend the pilot trial results and support the utility of incorporating sleep interventions into alcohol use disorder (AUD) prevention and treatment efforts.

INTRODUCTION

Alcohol use disorder (AUD) and insomnia are both serious public health concerns (Colten & Altevogt, 2006; Room et al., 2005). AUD and insomnia are highly comorbid, with 60–75% of individuals with AUD also experiencing insomnia (Chaudhary et al. 2020; Koob & Colrain 2020; Zhabenko et al. 2012). This comorbidity is due in part to the bidirectional and feedforward relationship between alcohol use and sleep, such that alcohol negatively impacts sleep, and disrupted sleep contributes to increased alcohol consumption (Koob & Colrain, 2020). Indeed, longitudinal evidence in samples of adolescents (Hasler et al., 2014; Hasler et al., 2016; Miller et al., 2017) and adults (Breslau et al., 1996; Short et al., 2019) suggests that poor sleep serves as a prospective risk factor for the development of problematic drinking. As such, sleep could be a unique treatment target for preventing the onset and development of AUD in at-risk individuals.

Cognitive behavioral therapy for insomnia (CBT-I) is the recommended first-line treatment for insomnia (Edinger et al., 2021) and is highly efficacious (van Straten et al., 2018). CBT-I can be provided either in-person or virtually (digital CBT-I; d-CBT-I) and typically consists of three major components aimed at addressing problematic thoughts and behaviors related to sleep: cognitive therapy, sleep restriction, and stimulus control (Edinger et al., 2021). Previous studies have administered CBT-I or d-CBT-I in samples of non-treatment seeking heavy drinkers. In one study, binge drinking young adults with insomnia were randomly assigned to receive either 5-weeks of in-person CBT-I or a single session of sleep hygiene. Those assigned to receive CBT-I reported significantly greater decreases in insomnia severity relative to those in the sleep hygiene condition. Although both groups showed decreases in drinking outcomes, there were no significant differences in these measures between the two groups. However, mediation analyses suggested that effects of CBT-I on alcohol-related consequences were indirectly influenced by improvements in sleep (Miller et al., 2021). In another study of heavy drinking young adults, individuals were randomly assigned to complete a novel 4-week online program incorporating the major components of CBT-I or a 4-week online program dedicated to promoting healthy behaviors more globally. Although no significant group differences were found on either sleep or drinking measures, both groups displayed improvements in these measures and greater improvements in sleep were associated with greater reductions in drinking (Fucito et al., 2017). These studies established the feasibility of administering CBT-I and d-CBT-I in samples of active heavy drinkers with insomnia and provided preliminary support for targeting sleep to improve drinking outcomes.

Building on these initial studies, our group conducted a pilot study of a well-validated d-CBT-I program in a sample of heavy drinkers with insomnia (Verlinden et al. 2023). Sleep Healthy Using the Internet (SHUTi; Ritterband et al., 2017) is a fully automated version of CBT-I shown to be effective in a variety of populations (Luyster et al., 2018; Mattos et al., 2021; Moloney et al., 2020; Morin et al., 2025; Ritterband et al., 2012; Ritterband et al., 2017; Ritterband et al., 2022; Vedaa et al., 2020; Zhou et al., 2022). We conducted a single-blind, randomized pilot study to determine the efficacy of SHUTi to improve sleep and drinking outcomes in a sample of heavy drinkers who reported having insomnia. Participants were randomly assigned to complete either the SHUTi program or a participant education program over the course of 9-weeks. Participants completed follow-up assessments following the 9-week intervention period, 3-months post-intervention, and 6-months post-intervention. As with previous studies of CBT-I, individuals in the SHUTi condition reported significantly greater reductions in insomnia severity. Moreover, those in the SHUTi condition also reduced their quantity and frequency of alcohol consumption relative to individuals in the PE condition. These promising results were the first to provide evidence of group differences in drinking patterns following d-CBT-I and further highlighted the promise of d-CBT-I for treating heavy drinking individuals with insomnia (Verlinden et al. 2023).

The current study aimed to replicate and extend the outcomes of our pilot trial in several important ways. First, we aimed to recruit a larger sample of heavy drinkers with insomnia with greater diversity in age, ethnicity, and race. Second, we wanted to extend the assessment of current drinking and sleeping habits in a way that may be less susceptible to recall bias. To do this, we collected daily diary measures of both alcohol consumption and sleep. These measures also enabled the evaluation of treatment effects on key aspects of sleep that were not assessed in our pilot trial [e.g., sleep onset latency (SOL) and wake after sleep onset (WASO)] as well as the use of sleep medications. Finally, the current study included measures of depression and stress to evaluate possible treatment effects on two mental health outcomes related to both sleep and alcohol use (Keyes et al., 2012; Kim & Dimsdale, 2007; McHugh & Weiss, 2019; Tsuno et al., 2005). We hypothesized that SHUTi would produce significantly greater improvements in sleep, drinking, sleep medication use, and mental health outcomes over time compared to a patient education program.

METHODS

DESIGN

The primary aims of this study were to 1) test the efficacy of the SHUTi program on sleep and drinking outcomes in an independent sample of adults with alcohol and sleep problems and 2) extend prior findings by assessing additional measures of sleep and drinking using data collected from daily sleep diaries over the course of the trial. Heavy drinkers with chronic insomnia were randomly assigned (single blind, participant) to either the SHUTi or control (patient education; PE) programs in a parallel study design with a 1:1 allocation ratio and superiority framework. Self-report measures of sleep and alcohol consumption were collected online at four assessment points: baseline, post 9-week intervention period, 3-months post-treatment, and 6-months post-treatment. We assessed the degree to which SHUTi improved sleep and decreased drinking over time relative to the PE condition. The Institutional Review Board (IRB) of the University of Kentucky approved this study and it was carried out in accordance with the Declaration of Helsinki. All participants gave informed consent and were compensated for their time.

PARTICIPANTS

Heavy drinking men and women with insomnia (n=113) were recruited through online (social media) and printed (flyer) advertisements in Florida, Indiana, Kentucky, and Ohio (where the investigators’ institutions and recruiting sites were located). Recruitment took place from January 2023 until July 2024 and a recruitment service (BuildClinical) was utilized to assist with online advertisements. Heavy drinking was defined as having an Alcohol Use Disorders Identification Test (AUDIT) score >7 and self-reported weekly binge episodes (4+/5+ drinks in one sitting for women/men) on the Timeline Follow-Back (TLFB). Volunteers met study criteria for chronic insomnia if they self-reported insomnia for three or more nights a week for the past three months and had a current Insomnia Severity Index (ISI) score >14. Additional inclusion criteria were: fluency in English, regular access to the internet, and age range of 18–50. The minimum age of 18 was selected to exclude minors and the maximum age of 50 was chosen to avoid the high rates of sleep disorders in post-menopausal women (Dancey et al., 2001; Polo-Kantola, 2008). Volunteers were excluded if they self-reported a previous diagnosis of any of the following: AUD or substance use disorder, obstructive sleep apnea, schizophrenia, bipolar disorder, or other psychotic spectrum disorders, serious medical conditions (e.g., cancer, diabetes, cardiovascular disease), and neurological problems. Additional exclusion criteria were: shift work, pregnancy, lactating, plans to become pregnant in the next three months, and high risk of obstructive sleep apnea as determined by meeting ≥ 3 criteria on the STOP-Bang (Chung et al., 2016). Participants were not excluded for using prescription or over-the-counter sleeping aids, although their use of sleep aids was recorded.

To determine sample size, a power analysis was conducted based on findings from the pilot trial (Verlinden et al., 2023). Change scores (post- minus pre-intervention) for the ISI and total drinks were calculated separately for those in the SHUTi and PE conditions. Between-group t-tests comparing ISI change scores suggested a large effect size of SHUTi for sleep (d = 0.78) and a medium effect size for drinking (d = 0.60). Power analyses using g-power indicated a sample size of n = 50 in each condition would provide over 80% power (alpha = 0.05) to detect a significant effect of SHUTi compared to PE on both sleep and drinking measures.

PROCEDURE

All screening and study procedures took place remotely. Interested volunteers completed an online pre-screen to determine eligibility. Individuals who met eligibility criteria were emailed an invitation to the study along with a link to the consent form and baseline surveys using REDCap electronic data capture tools hosted at the University of Kentucky (Harris et al., 2009; Harris et al., 2019). Baseline surveys took approximately 30 minutes to complete and included measures of demographics, sleep, alcohol use, and physical and mental health. Next, participants completed a 30-day TLFB via video call to provide a detailed assessment of past-month alcohol consumption. Participants were then randomly assigned to either the SHUTi or PE condition via a predetermined allocation using randomizer.org. This predetermined allocation was only accessible to the study coordinator who both enrolled and allocated participants. Recruitment continued until a sample size of at least n = 100 enrolled participants was reached. After randomization, participants were given access to their assigned condition. Participants in both conditions were required to fill out daily sleep diaries for 10 days within a 14-day period before receiving access to their respective intervention.

SHUTi (TREATMENT CONDITION)

SHUTi is a fully automated online program that is interactive and features personalized components such as individual sleep windows and goal-setting. There are six cores, including an overview of insomnia (“Getting Ready”), sleep restriction (“Sleep Scheduling”), stimulus control (“Sleep Practices”), cognitive restructuring (“Thinking Differently”), sleep hygiene (“Sleep Hygiene”), and relapse prevention (“Moving On”). Interactive educational and interventional content is embedded within the cores, including didactic games, quizzes, and video testimonials. Each core requires approximately 30–45 minutes to complete and participants must log at least five sleep diaries within the previous seven days to establish and continue to receive updated sleep window recommendations in subsequent cores. Participants were given nine weeks to complete all six cores of the program before completing the post-intervention follow-up. Individuals who did not complete all six cores during this timeframe had the opportunity to complete additional cores after the post-intervention follow-up. While SHUTi automatically sends daily reminders to participants, study personnel did not contact participants during the nine-week intervention unless participants reached out with study related questions (see Thorndike et al. 2008 for more details on SHUTi).

PATIENT EDUCATION (PE) CONTROL CONDITION

The online PE program was created by the SHUTi developers and has been consistently used as the control condition in previous clinical trials of SHUTi (Ritterband et al., 2017; Vedaa et al., 2020; Verlinden et al., 2023). To access the PE program, participants log in to their SHUTi account in the exact same manner as those in the SHUTi treatment condition. Instead of having access to the interactive SHUTi program, however, fixed, non-interactive content is presented containing static text and images regarding the causes, impacts, prevalence, and symptoms of insomnia, sleep hygiene strategies, and recommendations for when to seek help (Ritterband et al., 2017). Over the course of nine weeks, those in the PE group had unlimited access to this content. They were not required to fill out sleep diaries during the intervention period but could print out a sleep diary to complete on their own if they chose to do so.

POST-INTERVENTION

Following the intervention period, participants in both conditions completed 10 additional sleep diaries. They also completed the same online surveys and TLFB video-call administered at baseline. TLFBs were administered, scored, and double-entered either by a trained research assistant who was blinded to participant condition assignment or by the study coordinator who was unblinded.

3-MONTH & 6-MONTH FOLLOW-UPS

At 3-months and 6-months post-intervention, participants completed the same online surveys and TLFB video-call administered at baseline. At the 6-month follow-up only, participants also completed the AUDIT, 10 final sleep diaries, and an end-of-study phone interview to provide their SHUTi experiences and feedback (data not reported here).

MEASURES

SLEEP AND ALCOHOL DIARIES

Diaries consisted of 11 questions regarding the participant’s sleep, use of alcohol, and use of sleep aids from the previous night. We used a standardized protocol for the SHUTi program for diary data cleaning (Shaffer et al., 2023a; Shaffer et al., 2023b) and only timepoints with 10 valid diaries were included in analyses. Outcome variables included SOL, WASO, number of days sleep medication(s) were used, drinking days (DD), heavy drinking days (4+/5+ drinks in a single day for women/men; HDD), total number of drinks (TD), and mean number of drinks per drinking day (DPDD). SOL and WASO were selected as the primary sleep diary outcomes since they represent the two primary insomnia criteria: delayed sleep onset and disrupted sleep maintenance. Other outcome measures were sleep efficiency (SE) and total sleep time (TST). Sleep medications were defined as any over-the-counter or prescription medication used to help one sleep.

ISI

This 7-item self-report questionnaire (Bastien et al., 2001) requires participants to rate the severity of their insomnia symptoms over the previous two weeks. Scores range from 0 to 28: 0–7 - no clinically significant insomnia; 8–14 - subthreshold insomnia; 15–21 - moderately severe clinical insomnia; 22–28 - severe clinical insomnia.

PITTSBURGH SLEEP QUALITY INDEX (PSQI)

This 19-item self-report measure (Buysse et al., 1989) provides a general index of sleep quality and sleep disturbances over the previous month. Scores >5 indicate poor sleep quality.

TLFB

The TLFB (Sobell & Sobell, 1992) assesses daily patterns of alcohol consumption over the past 30 days. The measure uses “anchor points” to structure and facilitate participants’ recall of past drinking episodes to provide a more accurate retrospective account of alcohol use during that time period. The secondary outcome measures included in these analyses are percentage of heavy drinking days (PHDD), percentage of drinking days (PDD), and total drinks (TD). These outcomes were chosen in line with general recommendations for AUD treatment clinical trials (Witkiewitz et al., 2015).

AUDIT

This 10-item self-report measure (Babor et al., 1989) assesses past-year alcohol consumption and alcohol-related problems with scores ranging from 0 (no-alcohol related problems) to 40 (most severe alcohol-related problems). A score > 7 typically indicates hazardous drinking.

PERCEIVED STRESS SCALE (PSS)

This 10-item self-report measure (Cohen et al., 1983) assesses global stress levels over the previous month. All items are scored on a scale of 0 to 4, with higher scores indicating greater psychological distress.

CENTER FOR EPIDEMIOLOGIC STUDIES DEPRESSION SCALE REVISED 10-ITEM VERSION (CESDR-10)

This 10-item self-report measure (Andresen et al., 1994) gauges symptoms of depression over the past week. Each item is scored on a scale of 0 to 3, with scores ≥ 16 indicating clinical relevance.

SHUTi SATISFACTION SCALE

Participants rated their adherence, acceptability, and satisfaction with the treatment program (SHUTi or PE). Those in the SHUTi group also rated the elements of SHUTi they found most helpful.

STATISTICAL ANALYSES

Group differences (SHUTi vs PE) in demographic and baseline measures were tested using between-group t-tests and chi-square tests in IBM SPSS Statistics for WindowsOS, Version 30.0. All other statistical analyses were performed using R software Version 4.4.2 (R Core Team, 2024). We used an intention-to-treat approach as recommended by Witkiewitz et al. (2015) for randomized controlled trials of alcohol treatments, such that all participants who were enrolled in either intervention were included in analyses, regardless of their completion status. Diaries of participants were averaged at each timepoint to create summary scores for the diary-derived outcomes. Due to non-normal distributions, SOL, WASO, diary-derived TD and DPDD, and all TLFB variables were log transformed prior to analysis. SE was square transformed prior to analysis due to a non-normal distribution that was not corrected by log transformations. Next, WASO and TLFB variables were winsorized to account for the presence of outliers (>3 SD from mean). To predict all outcomes, we conducted separate linear mixed effects models for repeated measures with maximum likelihood parameters (Hedeker & Gibbons, 2006) to determine the degree to which condition (SHUTi vs. PE) interacted with timepoint (baseline, post-intervention, and 6-month follow-up for diary-derived outcomes; baseline, post-intervention, 3-month follow-up, and 6-month follow-up for all other outcomes). Intraclass correlation coefficients (ICCs) for all outcome measures are reported in the Supplementary Materials. Timepoint (level 1) was nested within participant (level 2), and the models included random subject intercepts to account for the correlation between repeated measures. SHUTi and baseline assessments were coded as the reference conditions for condition and timepoint, respectively. For diary-derived outcomes, timepoint was treated as a categorical variable to account for non-linear effects, and Type III ANOVA tests were conducted to test condition x timepoint interactions. For all outcomes, significant interactions were probed by 1) comparing within-group effect sizes at each timepoint compared to baseline and 2) conducting between-group t-tests at each timepoint. To determine within-group effect sizes for the PE condition, all models were conducted with the PE condition recoded as the reference condition. Significance was set at a 2-sided alpha level of 0.05 for all comparisons.

TRANSPARENCY AND OPENNESS

We report how sample size, data exclusions, manipulations, and measures used were determined. The trial was registered on 11/17/2022 at clinicaltrials.gov (ID: NCT05630118; URL: https://clinicaltrials.gov/study/NCT05630118) and all of the pre-registered outcomes are reported below. A full protocol (including statistical analysis plans and code) can be provided upon request.

RESULTS

PARTICIPANTS

Study enrollment is outlined in Figure 1 and shows that 323 volunteers were deemed eligible based on their responses from the pre-screen. Of these, 156 consented to participate. After consenting, 43 participants were excluded for either reporting less than weekly binge drinking on the TLFB (n = 20) or STOP-Bang responses indicating high risk for obstructive sleep apnea (n = 23). Eligible participants were randomly assigned to either the SHUTi (n = 61) or PE condition (n = 52). Of these, n = 41 (SHUTi) and n = 39 (PE) completed at least one follow-up measure. Those who did not complete any follow-up measures had significantly lower ISI scores (p = 0.016) at baseline (n=33, M = 19.0, SD = 2.8) compared to participants who completed at least one of the follow-up measures (n=80, M = 20.7, SD = 3.5). No other differences at baseline were found between completers and non-completers of follow-up assessments.

Figure 1.

Consolidated Standards of Reporting Trials (CONSORT) flowchart of study enrollment.

Demographics and baseline measures are provided separately by condition in Table 1. Across both groups participants were predominantly White, Non-Hispanic, and female. However, in line with our aim for a more diverse sample, participants in the current study were significantly older (age: M = 32.2, SD = 9.1, p < 0.001) and the sample had a significantly larger proportion of non-White (30%) and Hispanic or Latino individuals (14%) relative to our pilot study (ps ≤ 0.029). Individuals in both groups were well above the threshold for moderately severe clinical insomnia (ISI > 14), reported poor sleep quality (PSQI > 5), and had mean AUDIT scores well above the threshold for hazardous drinking (AUDIT > 7). Participants also reported moderate levels of stress and were slightly below the threshold for clinically significant symptoms of depression (CESD-R < 16). Individuals in the SHUTi condition were slightly but significantly older than individuals in the PE condition (p = 0.048). The groups did not differ on any other demographic or baseline measure. No participants in either group reported any adverse events during the study.

Table 1.

Demographic and baseline measures for the SHUTi and PE conditions.

	SHUTi (n=61)	PE (n=52)	Contrasts
Sex (M/F)	20/41	18/34	ns
Race			ns
American Indian or Alaska Native	1	0
Asian	1	0
Black or African American	13	13
White	41	37
Multiracial	4	2
Unknown	1	0
Ethnicity			ns
Hispanic or Latino	12	3
Not Hispanic or Latino	47	45
Unknown	2	4
Age	33.8 (9.1)	30.4 (9.0)	p = 0.048
Highest Level of Education			ns
Unknown	3	2
Some High School	1	1
High School	8	9
Some College	18	22
Bachelor’s Degree	16	12
Postgraduate Degree	15	6
Annual Household Income			ns
$50,000 or less	29	28
$50,000 to $100,000	22	13
$100,000 to $150,000	7	5
$150,000 or more	8	6
ISI	20.3 (3.5)	20.1 (3.4)	ns
PSQI	13.8 (2.9)	13.7 (2.6)	ns
PSS	22.3 (6.5)	24.4 (6.6)	ns
CESD-R	15.1 (5.2)	15.9 (5.3)	ns
AUDIT	18.5 (7.0)	19.1 (6.6)	ns
TLFB Measures
Heavy Drinking Days (Raw)	13.7 (8.3)	11.7 (6.7)	ns
Heavy Drinking Days (%)	45.7 (27.7)	38.9 (22.2)	ns
Drinking Days (Raw)	19.3 (8.1)	17.7 (7.9)	ns
Drinking Days (%)	64.2 (26.8)	59.1 (26.2)	ns
Total Drinks	114.6 (80.2)	120.9 (111.5)	ns
Diary Measures	n = 50	n = 43
SOL (min)	52.1 (36.6)	55.5 (39.2)	ns
WASO (min)	71.3 (53.8)	69.2 (41.7)	ns
SE (%)	73.2 (14.5)	74.7 (12.4)	ns
TST (hrs)	5.9 (1.7)	6.3 (1.3)	ns
Sleep Medication Days	3.0 (3.5)	2.0 (2.8)	ns
Heavy Drinking Days	4.3 (3.2)	3.7 (2.5)	ns
Drinking Days	6.7 (3.0)	6.4 (2.7)	ns
Total Drinks	35.8 (24.2)	31.5 (17.1)	ns
DPDD	5.2 (2.7)	5.3 (3.3)	ns

Note: Significant contrasts are highlighted in bold. Group contrasts tested by between-group t-tests (Age, ISI, PSQI, PSS, CESD-R, AUDIT, and all TLFB measures) and chi square tests (sex, race, ethnicity, education, and income). Raw values indicate 30-day totals rather than percentages. ISI = Insomnia Severity Index; PSQI = Pittsburgh Sleep Quality Index; PSS = Perceived Stress Scale; CESD-R = Center for Epidemiologic Studies Depression Scale Revised 10-Item Version; AUDIT = Alcohol Use Disorders Identification Test; TLFB = 30-day Timeline Follow-back; SOL = Sleep Onset Latency; WASO = Wake After Sleep Onset plus Early Morning Awakenings; SE = Sleep Efficiency; TST = Total Sleep Time; DPDD = Mean Drinks Per Drinking Day.

PRIMARY OUTCOMES

Primary Sleep Outcomes.

The results of the primary sleep outcomes are presented in Figure 2 and Table 2. As hypothesized, greater improvements were observed in the SHUTi condition compared to the PE condition in all but one primary sleep outcome (significant condition × timepoint interactions: ISI (p = 0.003), PSQI (p < 0.001), SOL (p = 0.015), and sleep medication days (p = 0.005); Table 2]. By contrast, the interaction was not significant for WASO (p = 0.105). Follow-up tests showed that within-group effects from baseline to post-intervention were larger in the SHUTi condition (Cohen’s d range = 0.47–2.90) compared to the PE condition (Cohen’s d range = 0.10–1.68) for all measures (Figure 2; Table S1). This pattern extended to the 6-month follow-up as well (SHUTi Cohen’s d range = 1.14–3.38; PE Cohen’s d range = 0.03–2.28). Additionally, between-groups tests showed that those in the SHUTi condition reported significantly lower ISI and PSQI scores and reduced SOL compared to those in the PE condition at all follow-ups (Table S2; ps ≤ 0.035). No significant group differences were observed for sleep medication days (Table S2).

Figure 2.

Mean (SD) ratings on the Insomnia Severity Index (ISI), Pittsburgh Sleep Quality Index (PSQI), Sleep Onset Latency (SOL), and Sleep Medication Days. At baseline, post-intervention, and 3-month (ISI; PSQI) and 6-month follow-ups. There were significant group × time interactions for the ISI, PSQI, SOL, and Sleep Medication Days. Each bar represents the raw mean and has total number of participants listed at the bottom. Within-group effects for each timepoint from baseline to each follow-up are shown in the top bars and error bars show standard error. Cohen’s d values are calculated using the estimate (beta) of each timepoint (compared to baseline) divided by the residual standard deviation (sigma).

Table 2.

Linear mixed effects models testing effects of condition and timepoint on primary sleep and alcohol consumption outcomes. Type III ANOVA results are reported for diary outcomes.

Primary Sleep Outcomes
		Estimate	SE	|t|	p
ISI
	Intercept	18.30	0.58	31.54	<0.001
	Timepoint	−4.04	0.30	13.34	<0.001
	Condition	0.58	0.85	0.69	0.493
	Timepoint x Condition	1.28	0.43	2.96	0.003
PSQI
	Intercept	12.96	0.45	29.12	<0.001
	Timepoint	−2.57	0.21	12.54	<0.001
	Condition	0.25	0.65	0.39	0.699
	Timepoint x Condition	1.16	0.29	3.97	<0.001
		F		p
SOL (Diary; min)
	Timepoint	F2, 137.32 = 23.59		<0.001
	Condition	F1, 93.74 = 8.61		0.004
	Timepoint x Condition	F2, 136.01 = 4.37		0.015
WASO (Diary; min)
	Timepoint	F2, 135.71 = 22.05		<0.001
	Condition	F1, 90.61 = 3.59		0.061
	Timepoint x Condition	F2, 134.30 = 2.29		0.105
Sleep Medication Days (Diary)
	Timepoint	F2, 129.40 = 10.60		<0.001
	Condition	F1, 94.84 = 0.00		0.987
	Timepoint x Condition	F2, 128.69 = 5.52		0.005
Primary Drinking Outcomes
		F		p
Heavy Drinking Days (Diary)
	Timepoint	F2, 127.87 = 26.67		<0.001
	Condition	F1, 86.61 = 0.36		0.548
	Timepoint x Condition	F2, 126.68 = 3.35		0.038
Drinking Days (Diary)
	Timepoint	F2, 132.98 = 21.63		<0.001
	Condition	F1, 95.56 = 1.02		0.315
	Timepoint x Condition	F2, 132.07 = 2.81		0.064
Total Number of Drinks (Diary)
	Timepoint	F2, 137.91 = 21.20		<0.001
	Condition	F1, 94.20 = 1.24		0.268
	Timepoint x Condition	F2, 136.60 = 0.95		0.391
Mean Drinks Per Drinking Day (Diary)
	Timepoint	F2, 136.47 = 15.03		<0.001
	Condition	F1, 91.04 = 0.43		0.512
	Timepoint x Condition	F2, 135.04 = 0.15		0.865

Note: Bold font indicates significant effects (p ≤ 0.05). ISI = Insomnia Severity Index; PSQI = Pittsburgh Sleep Quality Index; SOL = Sleep Onset Latency; WASO = Wake After Sleep Onset.

Primary Drinking Outcomes.

Results of the primary (diary-derived) drinking outcomes are presented in Figure 3 and Table 2. Greater reductions in drinking were observed for 2 of the 4 primary outcomes [significant condition x timepoint interaction for HDD (p = 0.038) and a trend level effect for DD (p = 0.064)]. Significant interactions were not observed for diary-derived TD (p = 0.391) or DPDD (p = 0.865). Follow-up analyses showed that within-group effects from baseline to post-intervention were larger in the SHUTi condition for both HDD (SHUTi: Cohen’s d = 1.12; PE: Cohen’s d = 0.64) and DD (SHUTi: Cohen’s d = 1.12; PE: Cohen’s d = 0.34). This pattern was also observed at the 6-month follow-up (HDD: SHUTi Cohen’s d = 1.71; PE Cohen’s d = 0.64; DD: SHUTi Cohen’s d = 1.50; PE Cohen’s d = 0.34; for TD and DPDD see Table S1). No significant group differences were observed between any of the primary drinking measures at any timepoints (see Table S2).

Figure 3.

Mean (SD) ratings for diary-derived Heavy Drinking Days (HDD) and Drinking Days (DD) and 30-Day TLFB-derived Percentage of Heavy Drinking Days (PHDD) and Percentage of Drinking Days (PDD). There were significant group x time interactions for diary-derived HDD and TLFB-derived PHDD and PDD and a trend-level effect for diary-derived DD. Each bar represents the raw mean and has total number of participants listed at the bottom. Within-group effects for each timepoint from baseline to each follow-up are shown in the top bars and error bars show standard error. Cohen’s d values are calculated using the estimate (beta) of each timepoint (compared to baseline) divided by the residual standard deviation (sigma).

SECONDARY OUTCOMES

Secondary Sleep Outcomes.

Results of the secondary sleep outcomes are presented in Table 3. Greater improvements in SE over time were observed in the SHUTi condition compared to the PE condition [significant condition × timepoint interaction (p = 0.010)]. In contrast, this interaction was not observed for TST (p = 0.245). Follow-up tests revealed that within-group effects from baseline to post-intervention were larger in the SHUTi condition for the SE (SHUTi: Cohen’s d = 1.57; PE: Cohen’s d = 0.64; Table S1). This pattern was also observed at the 6-month follow-up (SHUTi: Cohen’s d = 1.78; PE: Cohen’s d = 0.94). Finally, between-groups tests showed that those in the SHUTi condition reported significantly greater average SE at the 6-month follow-up (p = 0.021; Table S4).

Table 3.

Linear mixed effects models testing effects of condition and timepoint on secondary outcomes. Type III ANOVA results are reported for diary outcomes.

Secondary Sleep Outcomes
		F		p
SE (Diary; %)
	Timepoint	F2, 134.91 = 35.48		<0.001
	Condition	F1, 92.18 = 2.84		0.095
	Timepoint x Condition	F2, 133.65 = 4.77		0.010
TST (Diary; hrs)
	Timepoint	F2, 136.43 = 10.61		<0.001
	Condition	F1, 95.92 = 0.01		0.916
	Timepoint x Condition	F2, 135.31 = 1.42		0.245
Secondary Drinking Outcomes
		Estimate	SE	|t|	p
Percentage of Heavy Drinking Days
	Intercept	0.34	0.03	16.91	<0.001
	Timepoint	−0.10	0.01	−11.03	<0.001
	Condition	−0.04	0.03	1.33	0.185
	Timepoint x Condition	0.04	0.01	3.49	<0.001
Percentage of Drinking Days
	Intercept	0.47	0.03	18.73	<0.001
	Timepoint	−0.08	0.01	8.85	<0.001
	Condition	−0.03	0.03	0.79	0.431
	Timepoint x Condition	0.03	0.01	2.15	0.032
Total Number of Drinks
	Intercept	4.48	0.10	45.51	<0.001
	Timepoint	−0.49	0.05	10.20	<0.001
	Condition	−0.01	0.14	0.04	0.971
	Timepoint x Condition	0.12	0.07	1.85	0.065
Secondary Mood and Stress Outcomes
		Estimate	SE	|t|	p
PSS
	Intercept	21.77	0.82	26.58	<0.001
	Timepoint	−1.78	0.30	5.88	<0.001
	Condition	2.19	1.20	1.82	0.071
	Timepoint x Condition	0.64	0.43	1.48	0.140
CESDR-10
	Intercept	14.56	0.69	21.17	<0.001
	Timepoint	−1.91	0.27	6.96	<0.001
	Condition	0.85	1.01	0.84	0.400
	Timepoint x Condition	0.86	0.39	2.21	0.028

Note: Bold font indicates significant effects (p ≤ 0.05).

Secondary Drinking Outcomes.

Results of the secondary (TLFB-derived) drinking outcomes are presented in Figure 3 and Table 3. Greater reductions in drinking over time in the SHUTi compared to the PE condition were observed for all of the TLFB-derived drinking outcomes [significant condition × timepoint interactions for TLFB PHDD (p < 0.001) and TLFB PDD (p = 0.033), and a trend-level effect for TLFB TD (p = 0.065); Table 3]. Follow-up analyses revealed that within-group effects from baseline to post-intervention were larger in the SHUTi condition (Cohen’s d range = 1.01–2.33) compared to the PE condition (Cohen’s d = range 0.63–0.78) for all measures (Figure 3; Table S1). This pattern extended to the 6-month follow-up as well (SHUTi Cohen’s d range = 1.97–2.51; PE Cohen’s d range = 1.27–1.54). No significant between-group differences were observed for any secondary TLFB measures (Table S3). A trend level effect showed lower AUDIT scores at the 6-month follow-up for those in the SHUTi condition (M = 10.29, SD = 5.71) compared to those in the PE condition (M = 13.24, SD = 6.69; p = 0.056; Cohen’s d = 0.47).

Secondary Mood and Stress Outcomes.

Results of the secondary mood and stress outcomes are presented in Table 3. Greater reductions in depression symptoms over time in the SHUTi compared to PE conditions were observed [significant condition × timepoint interaction for CESDR-10 (p = 0.028)]. By contrast, the interaction was not significant for PSS (p = 0.140). Follow-up tests revealed that within-group effects from baseline to post-intervention were larger in the SHUTi condition for the CESDR-10 (SHUTi: Cohen’s d = 1.03; PE: Cohen’s d = 0.57; Table S1). This pattern was also observed at the 6-month follow-up (SHUTi: Cohen’s d = 1.54; PE: Cohen’s d = 0.78). Finally, between-groups tests showed that those in the SHUTi condition reported significantly lower CESDR-10 scores at the 6-month follow-up (p = 0.024; Table S4).

USE, ADHERENCE, FEEDBACK, AND BLINDING

Over the course of the 9-week intervention period, individuals enrolled in the SHUTi condition logged in between 0 and 58 times (median = 22). 51.8% of participants (29 of 56) in the SHUTi condition completed all 6 cores within the designated 9-week intervention period. In contrast, 14.2% (8 of 56) did not complete any cores, 5.4% (3 of 56) completed through the first core, 3.6% (2 of 56) completed through the second core, 12.5% (7 of 56) completed through the third core, 7.1% (4 of 56) completed through the fourth core, and 5.4% (3 of 56) completed through the fifth core. 12 participants completed additional cores after the 9-week intervention period, 7 of whom ended up completing all 6 cores. Individuals in the PE condition logged in between 0 and 6 times (median = 1) over the course of the 9-week intervention period. 95.7% of participants (44 of 46) in the PE condition scrolled through all material provided in the program. In contrast, 2.2% (1 of 46) opened the program but did not scroll through all of it, and 2.2% (1 of 46) never accessed the program.

Results of user feedback are shown in Table S5. Of the 75 participants who completed the post-treatment SHUTi Satisfaction Scale, a greater percentage of participants in the SHUTi condition reported that their treatment made various aspects of their life “somewhat” or “a lot” better. Specifically, individuals in the SHUTi condition reported greater improvements in their insomnia symptoms, energy levels, productivity, stress coping, life enjoyment, hopefulness, and mood (between-group t-tests; ps ≤ 0.031) compared to those in the control condition. Among individuals in the SHUTi condition who completed the post-treatment assessment, a majority reported adhering to all but one of the major strategies and techniques consistently or most of the time and finding all these strategies and techniques to be very or moderately helpful (Table S6). Finally, 16 participants in the PE condition incorrectly stated that they believed they had been assigned to the SHUTi condition, and 9 participants in the SHUTi condition believed that they had been assigned to the PE condition.

DISCUSSION

In this study, we replicated and extended our previous finding that SHUTi improves sleep among heavy drinkers with chronic insomnia in a larger and more diverse sample. Specifically, we showed that in addition to improving self-reported insomnia symptoms and global sleep quality, SHUTi also decreases SOL, SE, and sleep medication usage as reported on daily diaries. Additionally, we replicated our previous finding that SHUTi decreases frequency of drinking days and heavy drinking days as assessed by the TLFB, and extended these findings to a diary-derived measure of heavy drinking days (and trend level effects for diary-derived drinking days and the AUDIT). However, despite these differences in magnitude of within-group improvements, there were no between-group differences in drinking outcomes at any timepoint. Finally, we showed that SHUTi decreases self-reported depression symptoms. Together, these findings support the utility of digital sleep interventions for improving sleep and mood among heavy drinkers with insomnia and provide evidence in support of digital sleep interventions for reducing alcohol consumption.

Our findings that SHUTi significantly improved sleep outcomes are in line with previous trials of CBT-I that showed improvement in insomnia severity, sleep quality, and sleep-related impairments in active heavy drinkers with sleep problems (Fucito et al., 2017; Miller et al., 2021: Verlinden et al., 2023). We further extend these findings by showing improvements in specific aspects of sleep, including SOL and SE, and decreases in the use of sleep medications. The latter finding is particularly beneficial for this population due to potential contraindications with alcohol (Tanaka, 2002; Zosel et al., 2011; NIAAA, 2014). Contrary to prior trials of SHUTi (Mattos et al., 2021; Morin et al., 2025; Ritterband et al., 2017; Ritterband et al., 2022; Vedaa et al., 2020), we did not find a significant effect of SHUTi on WASO. However, the effect size of changes in WASO over time were greater in the SHUTi condition (Cohen’s d range = 1.28–1.33) compared to PE (Cohen’s d range = 0.62–0.92), in line with expectations. Given that our sample was not explicitly powered to detect group differences in WASO, it is possible that the lack of a significant interaction effect for WASO is due to lack of power.

The positive effects of SHUTi in heavy drinkers with insomnia add to the growing body of research examining the potential efficacy of sleep interventions to improve not only sleep, but also drinking behaviors and other mental health outcomes. We replicated our previous findings of larger within-group reductions in alcohol consumption in the SHUTi compared to PE group (Verlinden et al., 2023), and we extended these to show similar effects with daily diary-reported drinking and AUDIT scores. These findings are also in line with a CBT-I trial among treatment seeking individuals with AUD that reported significant improvements in alcohol-related problems (Miller et al., 2023). However, these findings conflict with the overall lack of effect of CBT-I on alcohol-related outcomes as reported in a recent meta-analysis (Turkmen et al., 2025). It is important to note that the majority of studies included in the meta-analysis (5 of 8) examined CBT-I effects in AUD treatment seekers, many of whom were currently abstinent. In contrast, the current sample consisted of non-treatment seekers who were actively drinking. Along those lines, sleep interventions may be more effective in reducing problematic alcohol consumption than in preventing relapse altogether and therefore may be more in line with a harm reduction approach. Finally, our finding that SHUTi significantly improved symptoms of depression among heavy drinkers with insomnia is novel. Although SHUTi has been proven efficacious in improving mental health symptoms in other populations (Badour et al., 2023; Batterham et al., 2017; Moloney et al., 2020), this study is the first to demonstrate this effect in heavy drinkers with insomnia. It is possible that heavy drinkers with insomnia derive mental health benefits through improvements in sleep, adding further support for the utility of sleep interventions in this population.

It is important to note that sleep, alcohol consumption, and mental health outcomes were all significantly improved in both the SHUTi and PE conditions. This is in line with prior CBT-I trials showing significant within-group improvements in sleep and reductions in drinking following both CBT-I and sleep hygiene control conditions (Fucito et al., 2017; Miller et al., 2021). Importantly, the control condition in each of these examples was sleep-hygiene focused and thus provided participants information and recommendations to help improve their sleep (although to a substantially lesser degree than the CBT-I conditions). It is possible that this information alone may be sufficient to produce significant improvements in sleep, and that these improvements in sleep are accompanied by reductions in alcohol consumption. Moreover, one study showed that improvements in sleep were directly correlated with reductions in drinking in both the CBT-I and control conditions (Fucito et al., 2017). Given the mounting evidence suggesting links between improvement in sleep and decreases in alcohol consumption (both following CBT-I and general sleep hygiene information), it will be important for future studies to probe the mechanisms underlying this association to more clearly pinpoint the level of intervention necessary to provide sleep and drinking benefits in at-risk groups.

One important limitation to the current study is the relatively high attrition rate (~30%). Although it compares favorably to other digital health interventions, where rates commonly exceed 40–65% (Meyerowitz-Katz et al., 2020; Postel et al., 2011), and is in line with prior research demonstrating that heavy drinkers are significantly more likely to drop out of digital interventions (Radtke et al., 2017), these attrition rates still raise questions as to who in this population may benefit most from this intervention. It is important to note that attrition was driven by individuals with less severe insomnia symptoms, suggesting that our results may better generalize to those with more severe insomnia symptoms. Given that no other differences between completers and non-completers were found, it is possible that those with more severe insomnia are more motivated to complete the SHUTi program due to the severity of their symptoms, and that the implementation of this intervention may be more suitable for these individuals.

There are several other limitations to our work. First, only subjective measures of both sleep and drinking outcomes were used in this study. It will be important to test these hypotheses with objective measures of sleep and alcohol consumption (e.g., actigraphy, transdermal alcohol sensors). Second, no clinical diagnoses of insomnia or AUD were conducted, nor was there any objective screening for additional sleep disorders. Therefore, it is unknown if participants met DSM-5 criteria for insomnia or AUD or what the clinical severity of their conditions may be, limiting the interpretations that can be drawn from our results. Third, the sample is predominately female and white. Although both factors limit the generalizability of findings, the racial and ethnic composition of this sample is significantly broader than our prior study and accurately reflects the demographics of three of our four recruitment states (Kentucky, Ohio, Indiana) where 76–81% of the population are white (U.S. Census Bureau, 2025). Fourth, the AUDIT assesses one’s drinking habits over the past year, yet the 6-month follow-up occurred approximately 9 months after baseline. Therefore, this overlap prevents a precise reflection of changes in AUDIT scores over time. Fifth, some of the follow-up TLFBs were conducted by the study coordinator, who was not blind to participant condition. Finally, due to the online nature of the study, we cannot rule out the possibility that some participants may have sought treatment for alcohol while enrolled in the study. However, in end-of-study phone interviews participants were asked “Since the start of the study, have you discussed your sleep or drinking habits with a health professional?”. No alcohol treatment was reported to research personnel during these interviews or the TLFB.

These findings suggest important directions for next steps in this line of research. Although the within-group effects of SHUTi on alcohol consumption are promising, it is likely that providing additional, alcohol-specific education material in SHUTi could considerably enhance efficacy for drinking behaviors (particularly for between-group effects). Additionally, future work is needed to identify factors mediating the reductions in drinking seen here. For example, does improved sleep help to reduce drinking via less motivation to use alcohol as a sleep aid or are effects driven by other factors related to sleep quality (e.g., increased cognition, decreased depression)?

In conclusion, the current study replicated and extended our prior findings by showing that SHUTi was effective in improving sleep, alcohol consumption, sleep medication use, and mental health symptoms among heavy drinking individuals with insomnia. These findings add support for the implementation of scalable, evidence-based sleep interventions that can simultaneously address two major public health concerns: insomnia and problem drinking. Further research is needed to test the addition of alcohol-related content to these interventions and to further investigate the mechanisms underlying potential improvements in drinking. Future research should also continue developing retention strategies specifically for this vulnerable population within the broader challenge of engaging and retaining individuals with substance use disorders in digital health interventions.

Public Significance Statement:

Both poor sleep and problem drinking negatively impact health and often co-occur. This study suggests that, by treating sleep, individuals may see improvements in both their sleep habits and alcohol consumption.