From yawn to dawn: the effects of a sleep extension intervention on objective and subjective dimensions of sleep health in adults with habitual short sleep duration

Abstract

Study Objectives

We examined the impact of a sleep extension intervention on multiple dimensions of sleep in adults with habitual short sleep duration.

Methods

Thirty healthy participants (14 women; aged 23.1 ± 4.5 years; BMI 22.3 ± 2.2 kg/m² [mean ± SD]) with <6.5 h sleep/night completed a 2-week baseline assessment followed by a 4-week sleep extension intervention (2 h/night increased time in bed). Wrist actigraphy, at-home electroencephalography (EEG; DREEM headband), Insomnia Severity Index (ISI), Pittsburgh Sleep Quality Index (PSQI), Epworth Sleepiness Scale (ESS), and daily quality/satisfaction and alertness Likert-scales quantified sleep.

Results

Baseline total sleep time (TST) was 5.5 ± 0.7 h. During sleep extension, actigraphy time in bed and TST increased (p < .001) by 60.8 ± 46.7 and 46.6 ± 41.1 minutes, respectively, sleep onset shifted earlier (p < .001) by 51.9 ± 64.5 minutes, with regularity similar to baseline. EEG showed increases (all p < .05) in TST (61.8 ± 75.6 minutes), Stage N1 (5.09 ± 6.51 minutes), Stage N2 (39.54 ± 41.78 minutes), rapid eye movement sleep (13.58 ± 28.86 minutes), and wakefulness after sleep onset (4.86 ± 8.03 minutes), with a nonsignificant decrease (p = .07) in sleep efficiency (−1.38 ± 4.15%). Subjective alertness, ISI, PSQI, and ESS each improved (all p < .05) during sleep extension.

Conclusion

Our findings help establish efficacy of sleep extension as an experimental intervention the sleep field can leverage across diverse contexts to study potential health benefits of increasing free-living TST. During sleep extension, the largest effects were observed for improved TST and ESS. Alternatively, some sleep dimensions including sleep regularity remained unchanged, highlighting a potential need for developing multi-component interventions that can improve more dimensions of sleep as both short and irregular sleep are linked with adverse health outcomes.

Clinical Trials

Biomarkers of Increased Free Living Sleep Time. URL: https://clinicaltrials.gov/study/NCT04214184. ClinicalTRIALS.gov ID: NCT04214184.

Statement of Significance

We provide a comprehensive evaluation of the effects of a free-living sleep extension intervention on multiple dimensions of sleep health in adults with habitual short sleep duration. Using multiple objective and subjective outcomes, we show that extending time in bed for 4 weeks increases total sleep time and selectively increases N1, N2, and rapid eye movement sleep, while sleep regularity and N3 sleep remain unchanged. Sleep extension also improves subjective sleepiness and sleep quality. These findings help establish sleep extension as an experimental intervention for increasing free-living sleep duration and improving specific dimensions of sleep health. Importantly, the lack of improvement in some dimensions highlights the need for multi-component interventions to address sleep health and related disease risk more fully.

Introduction

Insufficient sleep duration is increasingly common among adults and represents a major public health concern. Guidelines from the Sleep Research Society and American Academy of Sleep Medicine recommend that adults aged 18–60 years regularly obtain 7–9 hours of sleep per night for optimal health [1, 2]. Yet, more than 33% of adults in the United States report sleeping less than 7 hours per night [3, 4], and total sleep time (TST) among adults has decreased over recent decades, especially among underrepresented ethnic and racial populations [5–8]. Evidence from observational studies consistently links insufficient sleep duration to adverse health outcomes including obesity, diabetes, cardiovascular disease, psychological disorders, poor cognitive function, and accidents [9–14]. Furthermore, findings from laboratory studies provide potential causal links showing experimental sleep restriction results in a range of adverse health risks including positive energy balance [15], impaired insulin sensitivity [16–18], and impaired cognition [19, 20].

Although TST is an important component of sleep health, sleep is a multidimensional construct, both in how it is quantified and how it can impact physiological functioning [21, 22]. For example, the RuSATED model outlines a broader view of sleep health using six key dimensions: Regularity, Satisfaction/Quality, Alertness/Sleepiness, Timing, Efficiency, and Duration (RuSATED) [22]. Additional dimensions proposed by others include sleep architecture, most rigorously quantified by electroencephalography (EEG), and the absence of clinical sleep disorders [21]. Both subjective and objective outcomes are required to fully quantify these dimensions of sleep health. Evidence more directly supporting this multi-dimensional construct demonstrates that poor sleep regularity is linked with adverse health outcomes, including all-cause mortality and risk of cardiometabolic disease [23–25]. Additionally, disrupted stage N3 sleep has been shown to worsen insulin sensitivity, a major risk factor for diabetes [26], and excessive daytime sleepiness measured by the Epworth Sleepiness Scale (ESS) has been linked to adverse cardiac outcomes [27]. Despite this evidence linking multiple dimensions of poor sleep to adverse health risks, there are limited data informing how free-living sleep-based interventions impact these multiple dimensions of sleep, especially among adults with habitual insufficient sleep but without clinical sleep disorders.

Consistent with current guidelines from the Sleep Research Society and American Academy of Sleep Medicine [1, 2] that emphasize sleep duration as a priority for health, as well as Life’s Essential 8 from the American Heart Association [28], behavioral interventions have been designed to increase nightly TST among people with insufficient sleep. Such interventions are being referred to as sleep extension [9]. Of note, earlier sleep extension studies were designed to extend sleep by using 10–14-hour time in bed conditions [29–32]; however, the translation of such findings to real-world free-living sleep extension is limited. More recent findings have demonstrated sleep extension is feasible and can improve TST, fasting insulin sensitivity, and decrease energy intake [33–38]. However, these prior findings are mainly focused on sleep duration, without analyzing the broader multi-dimensional aspects of sleep health. This represents an important gap in knowledge and limits our ability to optimize the potential health benefits of behavioral sleep-based interventions [22].

Given the overall limited data on free-living sleep extension interventions, especially in adults, additional research is needed to help establish feasibility and understand the effects of sleep extension on the multiple dimensions of sleep, including sleep architecture. As such, our aim was to examine the effect of a behavioral free-living sleep extension intervention on the multiple dimensions of sleep among adults with insufficient sleep. To accomplish our aim, we conducted a single-arm 4-week free-living sleep extension intervention in adults who report habitually sleeping less than 6.5 h per night and who did not have clinically diagnosed sleep disorders.

Materials and Methods

Participants

The University of Utah Institutional Review Board approved the full study protocol, and all participants provided written informed consent prior to initiating study procedures. Inclusion criteria were self-reported habitual TST <6.5 h per night, age 18–35 years, weight stable (<5 lb change in body weight over past 6 months), body mass index (BMI) 18.5–24.9 kg/m², self-reported less than 150 minutes of moderate to vigorous physical activity (MVPA) per week, and willingness to increase nightly time in bed. Additionally, prior to enrollment, participants must have lived at Salt Lake City altitude (4265 feet) or higher for ≥3 months to control for physiological adaptations to high altitude, similar to our prior work [17, 39]. Ovulating women were included based on a history of regular menstrual cycle ranging from 25 to 32 days with a maximum of 3 days variation month-to-month. Exclusion criteria: any clinically diagnosed medical or surgical condition within the prior year, any clinically significant psychiatric condition, any clinically diagnosed sleep disorder, history of shift work in year prior to study or travel more than one time zone in 3 weeks prior to study, symptoms of active illness (e.g. fever), use of prescription medications/supplements or illicit drugs for 30 days prior to or need of these medications at any time during the study, reporting consumption of >500 mg per day on average of caffeine, consuming >14 alcoholic units per week or >5 alcoholic units per day, and current nicotine use.

Recruitment and screening

Participants were recruited from the greater Salt Lake City, UT, geographical area using passive advertising, mainly through social media and flyers. Interested people completed an online pre-screening survey, and if they met initial eligibility criteria, they were invited for an in-person informed consent visit. Additional screening to confirm eligibility at the informed consent included: Sleep Disorders Questionnaire, Horne–Östberg Morningness Eveningness Questionnaire [40], Epworth Sleepiness Scale, Beck Anxiety Inventory [41], Beck Depression Inventory [42], and a drug habits survey. Beck Depression Inventory scores >15 or Beck Anxiety Inventory scores >19 were exclusionary. Final eligibility was confirmed by a medical screening consisting of health history, vitals, complete blood cell count, comprehensive metabolic panel, and pregnancy (for people who menstruate) and urine toxicology tests.

Protocol

We conducted a single-arm 6-week study with baseline and intervention segments. Data collection started on December 2, 2019, and was completed on September 10, 2024. Baseline consisted of a 2-week (weeks 1–2) ambulatory monitoring segment where participants were instructed to maintain their habitual sleep onset and offset (Figure 1). Within the first 12 days of beginning the baseline segment, participants underwent an in-laboratory attended polysomnography (University of Utah Health Clinical Sleep Wake Center) or at-home sleep test (ApneaLink device, [ResMed Corporation, Poway, California]) [43] to quantify their apnea hypopnea index. Sleep was monitored continuously by wrist actigraphy (Philips Actiwatch Spectrum Plus) [44] and daily paper and electronic time-stamped sleep logs, similar to our prior work [45–48]. Because we were not able to access real-time wrist-actigraphy data, we used the electronic sleep logs to track real-time self-reported daily bed and waketimes. The paper sleep logs included additional detailed questions about daily habits such as time in bed and caffeine and alcohol use (reported in Supplementary Materials), as well as sleep onset/offset. During all sleep opportunities for week 2 of baseline, participants were instructed to wear the DREEM 2 headband (research version; Dreem; Paris, France) [49], a dry mobile sleep EEG device. After baseline, participants completed a 4-week sleep extension intervention (study weeks 3–6) and sleep was monitored as in baseline, including use of the DREEM 2 headband during week 6. Participants completed the Insomnia Severity Index (ISI) and Pittsburgh Sleep Quality Index (PSQI) at the end of baseline and intervention segments and the Epworth Sleepiness Scale (ESS) weekly throughout the study.

Figure 1.

Protocol. Yellow bars indicate example of baseline habitual short sleep timing (week 1 & week 2), light blue (week 3 & week 4) and dark blue bars (week 5 & week 6) indicate theoretical targeted time-in-bed during the sleep extension intervention with increased nightly time-in-bed by 2 hours. During the intervention segment participants could choose how they increased their time-in-bed.

Sleep extension intervention

Two members of the study team were trained in delivering the intervention for all participants, with the focus of increasing nightly time-in-bed by 2 hours. Typical sleep habits and timing were discussed with each participant individually using a semistructured interview with the goal of assigning new individualized bed and wake times to fit their personal schedule. Increased time-in-bed could be achieved by moving bedtime earlier, waketime later, or a combination of both. Participants were instructed to increase time-in-bed gradually by 30 minutes each night until they achieved their targeted bedtimes and waketimes. Although we prescribed bedtimes and waketimes, the emphasis was on increasing time-in-bed. So, if a participant missed their targeted bedtime, they were instructed to adjust their waketime to still achieve their total time-in-bed goal. To help participants achieve their targeted time-in-bed, we suggested the following sleep hygiene tips: maintain consistent bedtimes and waketimes every day, including weekends; get ≥30 minutes of sunlight each morning; finish exercise 1–2 hours before bedtime (participants were told to maintain their current exercise routine); avoid caffeine in the afternoon, and avoid alcohol and food 1–2 hours before bedtime; avoid naps unless essential; maintain bedroom at 60–68°F; avoid electronic light exposure for one hour before bedtime; and take a hot shower/bath before bedtime. Study staff checked the electronic sleep log daily during the intervention segment. If the participant incurred 2 days of not completing the sleep log or 3 days of not achieving sleep extension, the study staff reached out to discuss barriers to sleep extension. At the midpoint of the intervention, participants completed an in-person check-in visit where actigraphy data were downloaded and reviewed, and any barriers were discussed and addressed.

Dimensions of sleep health

Although there is no strong consensus on definitions and metrics to individually quantify each dimension of sleep, we implemented the following framework based on prior data and a recently published scientific statement from experts in the sleep field [21, 22]. We quantified the following sleep dimensions using a single main outcome for each dimension, except timing, where we used two metrics. Sleep duration measured as TST by wrist actigraphy, regularity measured by Sleep Regularity Index (SRI) [50], satisfaction measured by PSQI [51], alertness measured by ESS, timing measured by sleep onset and offset, efficiency measured by DREEM headband, and sleep architecture measured by DREEM headband.

Sleep duration—wrist actigraphy

Wrist-actigraphy data were collected using the Philips Actiwatch Spectrum Plus and analyzed using Philips Actiware Software (Version 6.3.0) [44]. Participants were instructed to wear the watch continuously throughout the entire 6-week study, except when showering or completing activities that could damage the watch. Participants were instructed to press the “event-marker” button at their bedtime and waketime daily. The Actiwatch device collected data at 32 Hz in epoch lengths of 60s. Participant data are missing from baseline segment (1 participant), intervention weeks 3–4 (5 participants), and intervention weeks 5–6 (1 participant) due to device malfunction.

Event-marker, paper and electronic sleep diaries, activity counts, and ambient light data were used to identify rest intervals in the actigraphy data using the following workflow. Two laboratory staff members independently scored all data, specifically looking at the rest-start and rest-end time-points. The y-axis for activity levels was set to 100 counts. First, if the timing of the event marker and activity drop-off (zero activity counts) was within 15 minutes of each other, the event marker was considered valid and used to define the start/end of rest intervals. If there was no marker, or if the marker was >15 minutes from the activity drop-off, we evaluated whether three out of four criteria (light drop-off, activity drop-off, paper and electronic sleep diaries) were within 20 minutes of each other. Whichever of the three out of four criteria were within 20 minutes, the latest timepoint was marked as rest-start and the earliest timepoint was marked as rest-end. If three of four criteria were not met, 5 minutes of zero activity was used to determine rest-start and rest-end time-points. If the Actiwatch was off-wrist for greater than 45 minutes of the rest-interval, the night was excluded from analyses. Nights were also flagged as invalid if there were no paper or electronic sleep logs and the activity drop-off was unclear (i.e. multiple potential activity drop-offs). Four members (A.P.S., M.K., G.A.Z., and C.M.D.) of the study team met to resolve any disparities in the timing of the rest interval between the two independent scorers. TST was then derived from the Philips Actiware Software.

Naps were identified using the paper sleep logs. Self-report naps were assumed to be true and accurate as performance of the Philips Actiware Software algorithm is inadequate to rigorously identify and score naps. Primary actigraphy sleep metric analyses were conducted without naps included; analyses with naps included are presented in Supplemental Materials. In total, there were 170 naps reported by 23 participants.

Regularity—Sleep Regularity Index

Sleep regularity was calculated based on wrist-actigraphy data from the Philips Actiwatch Spectrum Plus, then exported from the Philips Actiware Software (Version 6.3.0) [44] for calculation of SRI [52] using the Sleepreg R package (Version 1.3.5) [50]. Additional details for actigraphy data analysis within the Philips Actiware Software are located in the Sleep Duration—Wrist Actigraphy, Methods section.

Satisfaction—Pittsburgh Sleep Quality Index

Sleep satisfaction was quantified using the global score from the Pittsburgh Sleep Quality Index (PSQI) [51], which is based on sleep over the past 30 days. The PSQI was completed at two time points (1) after the baseline segment and (2) after the intervention segment. A threshold >5 to define significant sleep problems using PSQI was defined as previously established [51]. Missing data from the PSQI are due to computer malfunction.

Alertness—Epworth Sleepiness Scale

Alertness was quantified using the Epworth Sleepiness Scale (ESS), sent via email weekly during baseline and intervention segments [53]. A threshold >10 to define mild excessive daytime sleepiness using ESS was defined as previously established [53]. Missing data from the ESS are due to an error in sending the survey or the participant not completing the survey as scheduled.

Timing—sleep onset and offset

Sleep onset and sleep offset were collected through wrist-actigraphy. Details for actigraphy data analysis within the Philips Actiware Software are located in the Sleep Duration—Wrist  Actigraphy, Methods section.

Sleep architecture and efficiency—DREEM 2 headband

Sleep staging and efficiency were derived from the DREEM 2 headband [49], a wireless multi-channel dry electrode sleep EEG headband. DREEM 2 data were collected in 30-second epochs and auto-scored for sleep stages N1, N2, N3, and rapid eye movement (REM) sleep using the proprietary DREEM algorithm [49] based on American Academy of Sleep Medicine scoring guidelines [54]. In a performance assessment study published, in part, by the company, the proprietary DREEM headband scoring algorithm was tested against in-laboratory PSG in a sample of healthy adults and using five independent certified scorers [49]. Findings showed DREEM had an overall accuracy of 83.8 ± 6.8% (mean ± SD) across all sleep stages and 74.0 ± 18.1% accuracy for detecting wakefulness. The five PSG scorers achieved a consensus of 86.4 ± 7.4% across all sleep stages; thus, the proprietary DREEM headband scoring algorithm showed similar performance as the certified PSG scorers. Notably, the highest performance accuracy for the DREEM headband was for stage N2 (82.9 ± 8.1%) with the lowest accuracy for stage N1 (47.7 ± 15.6%). The DREEM proprietary algorithm has a built-in data quality assessment based on electrode signal quality, expressed as a percentage from 1 to 100 with 100 being perfect quality. Our main analyses focused on nights with data quality ≥50. Additional analyses including all nights of data, nights with data quality ≥25, ≥50, and ≥ 80 are reported in Table S3. Additional detailed methods for DREEM 2 are in Supplemental Materials. Data analysis was conducted on participants who completed DREEM 2 data collection at both the baseline and intervention segments (N = 27). Three participants’ data are missing due to device malfunction or unavailability of the device.

Alternate outcomes to quantify sleep timing, regularity, and additional self-reported metrics

Additional methods of calculating sleep timing (midpoint of sleep) and sleep regularity (standard deviation of midpoint of sleep [midpoint SD], standard deviation of TST [TST SD], and social jetlag; Figure S2), and insomnia symptoms, daily sleep quality, and alertness are all reported in Supplementary Materials.

Statistical analysis

Statistical analyses were conducted with R (version 4.3.2). Presented data per segment are mean ± SD, whereas change data represent the average within-participant differences across segments, using participants with complete data. For outcomes with repeated observations within a study segment, we used linear mixed-effects models with a random intercept for participants and study segment (baseline vs. sleep extension) as a fixed factor. Outcomes analyzed with this approach included SRI, TST, Weekday TST, Weekend TST, Sleep Onset, Sleep Offset, Sleep Stages (N1, N2, N3, REM), DREEM TST, sleep efficiency, and wakefulness after sleep onset (WASO). TST, Sleep Onset, Sleep Offset, and SRI data were analyzed as weeks 1–2 (baseline) versus weeks 5–6 (sleep extension), whereas DREEM 2 data were analyzed as week 2 (baseline) versus week 6 (sleep extension). Effect sizes for the single main outcome for each dimension were calculated by Cohen’s d. Additional exploratory comparisons between TST with and without naps for each segment (baseline, intervention weeks 3–4, and intervention weeks 5–6), and between TST on weekdays versus weekends within each segment (baseline, intervention weeks 3–4, and intervention weeks 5–6) were conducted by paired t-tests.

For outcomes with single observations within a study segment, we used paired t-tests. Outcomes analyzed with this approach include the PSQI and ESS. The PSQI was calculated from surveys completed after baseline and after the sleep extension intervention. ESS was calculated as the average of weeks 1–2 (baseline) versus week 6 (sleep extension). Effect sizes were calculated by Cohen’s d.

Because the intervention was designed to extend time in bed and thereby increase TST, we conducted exploratory association analyses between TST and other sleep dimensions to generate hypotheses about potential interrelationships among these domains. For the correlation between TST and SRI, data were grouped by segment (baseline weeks 1–2, intervention weeks 3–4, and intervention weeks 5–6) and analyzed by linear mixed effects regression with a random intercept for participants. For the other dimensions of sleep, exploratory analyses tested the associations between change in TST and change in PSQI, ESS, efficiency, WASO, and sleep staging, using linear regression.

Results

Participants

Thirty-eight participants initiated data collection (Figure 2), with 30 participants (16 men/14 women) aged 23.1 ± 4.5 years (mean ± SD), with a normal body mass index 22.3 ± 2.2 kg/m² completing the protocol (Table 1). All participants reported sleeping less than 6.5 hours per night at screening with an average of 5.7 h ± 0.77. For chronotype, 16 of 30 participants were neither morning nor evening. One participant was an evening type, and no participants were extreme morning or evening chronotypes. AHI was 2.5 ± 3.5 events per hour, indicating, on average, that our participants did not meet diagnostic criteria for OSA. However, one participant had an AHI of 17.3, yet the main outcome of TST was not statistically different with or without this participant’s data; thus, we included this participant in all subsequent analyses. Detailed data with and without this participant included are reported in Supplemental Materials. Additionally, we assessed baseline demographic data between completers and noncompleters (Table S1). Analysis showed a lower (p < .01) AHI in noncompleters (0.25 ± 0.3 events/hour) versus completers (2.53 ± 3.5 events/hour). Although statistically significant, this relatively small difference is not considered physiologically relevant because both groups are below the minimum threshold of five events per hour for mild OSA.

Figure 2.

Consort diagram.

Table 1.

Participant demographics

	Participants (N = 30)
Age [mean ± SD]	23.13 ± 4.49
Sex	14F/16M
BMI [mean ± SD]	22.35 ± 2.24
Habitual total sleep time, self-report	5.71 ± .77
Chronotype
Neither	16
Moderately morning	4
Moderately evening	9
Evening	1
AHI* [mean ± SD]	2.53 ± 3.51
Epworth Sleepiness Scale at consent visit	10.27 ± 4.68
Race/ethnicity
White/not Hispanic or Latino	11
Asian/not Hispanic or Latino	9
White/Hispanic or Latino	3
Black or African American/not Hispanic or Latino	2
Other/Hispanic or Latino	2
American Indian or Alaska Native/White/not Hispanic or Latino	1
Asian/ White/not Hispanic or Latino	1
Other/not Hispanic or Latino	1

^*AHI is missing data for four participants due to device malfunction.

Sleep duration—wrist actigraphy

Baseline TST was 5.5 ± 0.7 hours, and intervention weeks 5–6 TST was 6.3 ± 0.6 hours, excluding naps (Figure 3). The average increase in TST from baseline to intervention weeks 5–6 (p < .001) was 46.6 ± 41.1 minutes with a very large effect size (d = 1.78), which parallels increased (p < .001) time in bed of 60.8 ± 46.7 minutes. Secondary analyses also showed TST was significantly (p < .001) increased during intervention weeks 3–4 (6.2 ± 0.6 h) versus baseline, with similar TST between intervention weeks 3–4 and intervention weeks 5–6. When analyzing TST with naps included, baseline TST with naps was 6.3 ± 14.9 minutes longer (p < .05) versus baseline TST without naps. Regarding intervention weeks 3–4 and intervention weeks 5–6, TST was similar with naps included versus without naps included. Despite the small difference in TST at baseline when naps were included, the baseline segment remained significantly different (both p < .001) versus intervention weeks 3–4 and intervention weeks 5–6 when naps were included in the analyses (Figure S1).

Figure 3.

Actigraphy-derived total sleep time (TST) across study segments. Yellow dot, box, and density plot indicate baseline. Light blue plots indicate intervention weeks 3–4, and dark blue plots indicate intervention weeks 5–6. Dots represent the average TST for each individual per segment. ^*** = p < .001, indicated by connecting lines. Complete actigraphy data were obtained for 29/30 participants for baseline, 29/30 participants for intervention weeks 3–4, and 29/30 participants for intervention weeks 5–6.

Exploratory analyses show TST was similar between weekend and weekdays within each study segment (Figure 4). When comparing TST across segments for weekends and weekdays separately, results showed similar increases in TST as observed for the overall analyses. Specifically, for weekend days, TST increased (p < .001) from baseline (5.7 ± 0.9 h) compared to intervention weeks 5–6 (6.4 ± 0.7 h) by 54.9 ± 98.3 minutes. For weekdays, TST increased (p < .001) from baseline (5.4 ± 0.8 h) to intervention weeks 5–6 (6.3 ± 0.6 h) by 63.2 ± 85.2 minutes.

Figure 4.

Weekdays versus weekends. Figure represents the average TST for each study segment during weekends and weekdays. Yellow indicates the baseline segment, light blue indicates intervention segment weeks 3–4, and dark blue indicates intervention segment weeks 5–6. Grey diamonds represent individual data points. ^*** = p < .001, ^** = p < .01 and connected with lines.

Regularity—Sleep Regularity Index

The SRI [52] was similar across all segments with a small effect size (d = 0.48) (Figure 5). Our exploratory association analysis showed TST was not associated (p = .109) with SRI (Figure S3). Similar findings show no differences from baseline to sleep extension for regularity measured by midpoint SD, TST SD, and social jetlag (Figure S2).

Figure 5.

Regularity. Shows the sleep regularity index across each study segment. Yellow indicates the baseline segment, light blue indicates intervention segment weeks 3–4, and dark blue indicates intervention segment weeks 5–6. Grey diamonds represent individual data points.

Satisfaction—PSQI

The PSQI improved (p < .01) during intervention week 6 versus baseline with a medium effect size (d = 0.56) (Table 2). At baseline, 14 participants had PSQI >5, and following sleep extension, this number dropped to 8 participants. No significant correlations were detected between the change in TST and the change in PSQI.

Table 2.

Subjective sleep outcomes

Survey: administration	Cutoff	Baseline		Week 6		P-value
		Mean ± SD	N > cutoff	Mean ± SD	N > cutoff
PSQI	>5: significant sleep problems	5.5 ± 2.2	14	4.1 ± 2.7	8	p < .01
ESS	>10: mild excessive daytime sleepiness	8.3 ± 4.6	11	3.4 ± 3.1	0	p < .001

Table 2 shows the mean ± standard deviation and the number of participants above the stated cutoff values “N > cutoff” for each survey of the Pittsburgh Sleep Quality Index (PSQI), and the Epworth Sleepiness Scale (ESS). Completed data include 28/30 participants for the PSQI pre-intervention survey, 27/30 participants for the postintervention survey. Twenty-eight out of 30 participants completed at least one of the ESS surveys during the baseline weeks; 21/30 participants completed the ESS surveys during week 6.

Alertness—ESS

ESS improved (p < .01) during intervention week 6 versus baseline with a very large effect size (d = 1.39) (Table 2). At baseline, 11 participants had ESS >10, and following sleep extension, this number dropped to zero. Furthermore, exploratory analyses showed a significant (p < .01) correlation between the change in TST and the change in ESS scores, where a 1-minute increase in TST was associated with a reduction/improvement of 0.056 in ESS (R² = 0.34, β = −0.056, SE = 0.018). Additional exploratory analyses showed improvements in the ISI and daily alertness during sleep extension versus baseline (Table S2).

Timing—sleep onset and offset

The increased TST in our primary analysis for intervention weeks 5–6 versus baseline was mostly due to participants going to bed earlier (Figure 6, A). Average sleep onset shifted earlier (p < .001) from baseline (01:29 ± 1:16 h) compared to intervention weeks 5–6 (00:37 ± 1:17 h) by 51.9 ± 64.5 minutes. Average sleep offset did not significantly change; at baseline, sleep offset was 07:43 ± 1:17 h, and during intervention weeks 5–6, sleep offset was 07:52 ± 1:24 h (Figure 6, B). Sleep onset showed a large effect size, and sleep offset showed a small effect size (sleep onset d = 0.97, sleep offset d = 0.32).

Figure 6.

Change in sleep onset and offset. (A) represents the average change in sleep onset comparing the baseline two weeks to the final 2 weeks of the intervention. (B) represents the average change in sleep offset comparing the baseline 2 weeks to the final 2 weeks of the intervention. Each individual participant is represented by the dark-blue bars, and the total mean change ± standard deviation is shown with the yellow line. The subjects are matched with the same letter(s) in each plot. Twenty-eight out of 30 participants are included in this analysis as two participants were missing either the baseline or intervention weeks 5–6 data.

Efficiency—DREEM 2 headband

Sleep efficiency showed a nonsignificant (p = .07) decrease from 91.19 ± 3.29 at baseline to 89.26 ± 5.70 at intervention with a very small effect size (d = 0.19) (Figure 7). When including all nights of data or nights with data quality ≥25 or ≥80, sleep efficiency showed a similarly small yet significant decrease (p < .05) (Table S3). No significant correlations were detected between the change in TST and the change in efficiency.

Figure 7.

Sleep efficiency. Figure shows the change from baseline week 2 to intervention week 6. Twenty-seven out of 30 participants completed both baseline and week 6 DREEM data collection. Yellow indicates the baseline segment, and dark blue indicates intervention segment weeks 5–6. Grey diamonds represent individual data points.

Sleep architecture—DREEM 2 headband

The DREEM 2 headband results for sleep staging, TST, and wake after sleep onset during the last week of baseline and the last week of intervention are in Figure 8. Our main analyses showed increases in TST, stage N1, stage N2, and REM sleep (all p < .001) and WASO (p < .05), from baseline week 2 to intervention week 6. Stage N3 was similar across baseline week 2 and intervention week 6. The effect sizes are as follows: TST (d = 0.885, large effect size), stage N1 (d = 0.863, large effect size), stage N2 (d = 0.664, medium effect size), Stage N3 (d = 0.114, very small effect size), REM (d = 0.447, small effect size), and WASO (d = 0.241, small effect size). Exploratory analyses showed significant (all p < .05) correlations between the change in TST and the change in stage N1 (R² = 0.21, β = 4.1, SE = 1.7), N2 (R² = 0.17, β = 24.4, SE = 11.3), N3 (R² = 0.19, β = 12.8, SE = 5.5), and REM sleep (R² = 0.23, β = 19.9, SE = 7.5). Each of these β values represents the change in sleep stage, in minutes, for a 1-hour increase in TST. No significant correlations were detected between the change in TST and the change in efficiency or WASO.

Figure 8.

Sleep architecture. All figures show the change from baseline week 2 to intervention week 6. (A–D) show the stages of sleep, while (E) and (F) show TST and WASO. ^*** = p < .001, ^** = p < .01, ^* = p < .05, and connected with lines. Twenty-seven out of 30 participants completed both baseline and week 6 DREEM data collection. Yellow indicates the baseline segment, and dark blue indicates intervention segment weeks 5–6. Grey diamonds represent individual data points.

Discussion

We conducted a single-arm trial to examine the impact of a 4-week behavioral sleep extension intervention on the dimensions of sleep health [21, 22] in young adults who report habitually sleeping less than 6.5 hours per night. Our findings show the sleep extension intervention increased time in bed by 60.8 ± 46.7 minutes, which led to an increased average nightly TST of 46.6 ± 41.1 minutes, quantified by wrist actigraphy. Such an increase exceeds the 30-minute threshold suggested for improved “clinical well-being," [55] yet TST was still below the recommended 7 or more hours of sleep per night. Our results demonstrate feasibility to experimentally increase TST over 4 weeks in the free-living environment, helping create the foundation for larger randomized controlled trials. From a research perspective, such experimental sleep extension has significant promise to expand our knowledge of the potential health benefits of increasing TST in people with habitual insufficient sleep [9]. Regarding other dimensions of sleep health, most prior sleep extension research has not comprehensively characterized the multiple dimensions of sleep health as we report here. Our findings demonstrated that a 4-week behavioral sleep extension intervention resulted in increased TST, satisfaction (PSQI), and alertness (ESS) and earlier timing of sleep onset, with changes in TST and ESS showing the largest effect sizes. Alternatively, we detected no significant changes in sleep regularity (SRI) or sleep efficiency (DREEM 2 Headband). Regarding sleep architecture, our findings showed increased stage N1, N2, and REM sleep, with no changes in N3 sleep, and increased WASO during the final week of sleep extension. Together, our findings show some, but not all, dimensions of sleep health were altered after 4 weeks of behavioral sleep extension. Thus, for specific target populations and health outcomes of interest, more tailored interventions are likely required to create positive changes, most notably for sleep regularity, which is strongly associated with all-cause mortality [23]. Importantly, improved ESS and sleep stages N1, N2, N3, and REM sleep showed correlations with change in TST, suggesting improvements in these dimensions of sleep were at least partly linked to increased TST. These exploratory analyses raise the hypothesis that people who achieve greater increases in TST may also achieve greater improvements in these dimensions. Critically, without a control group, we cannot rule out that observed improvements reflect study participation effects, including sleep tracking and attention from research staff. Larger randomized controlled trials with control groups are needed to confirm these findings and provide more precise estimates of effect size.

Sleep duration

As research on sleep extension in adults is still emerging, a key design aspect of our eligibility criteria was to enroll otherwise healthy young adults who reported sleeping less than 6.5 hours per night but without clinically diagnosed sleep or psychological disorders. At this early stage of research, our relatively strict eligibility criteria were designed to minimize potential confounding and help establish efficacy of sleep extension as an experimental intervention. Our participants had generally poor overall sleep at baseline. Specifically, baseline TST was ~5.5 hours with average PSQI, ISI, and ESS scores above thresholds for clinically significant sleep problems, subthreshold insomnia [56], and high daytime sleepiness [53], respectively. Although our participants were healthy young adults, these findings indicate they were obtaining insufficient sleep. The sleep extension intervention worked rapidly and was sustained through 4 weeks as participants significantly increased TST by ~40 minutes in the first 2 weeks and by ~47 minutes during the final 2 weeks of the intervention. These observations highlight the utility of sleep extension as an experimental intervention. This achieved increase in TST is similar to prior sleep extension studies with increases of ~30 to 60 minutes across different populations, ranging in duration from 1 to 6 weeks [9, 33, 34, 37, 38, 57–62]. Among this prior research, differences between weekday and weekend TST were reported in adults for only two studies [33, 34]. Our results show TST increased consistently on both weekends and weekdays across intervention segments. This pattern aligns with one prior sleep extension study [34] but not with another [33] where increases occurred primarily on weekdays. Importantly, across these sleep extension studies, the collective trend suggests a shift away from relying on weekend recovery sleep. We consider this a positive outcome given data suggesting weekend recovery sleep is inadequate to mitigate the adverse health consequences of insufficient sleep [17, 63, 64].

Participants showed a range of increased TST. Although the majority increased their TST, three participants decreased TST and eight increased by less than 30 minutes. Moreover, despite increased TST, most participants still obtained less than the recommended [1] 7–9 hours of sleep per night during the final 2 weeks of the intervention, with average TST exceeding 7 hours per night for only two participants. Additionally, out of 343 total nights of data across all participants for intervention weeks 5–6, TST exceeded 7 hours for only 72 nights (~21% of all nights). Previous sleep extension studies in adults tend to focus on the increase in TST, and only some noted the amount of sleep attained in comparison to the recommended 7–9 hours of sleep per night [57, 58]. As the intervention did not increase TST for all participants, understanding barriers to success is an important aspect of future research to improve sleep extension interventions. Longer-term studies are necessary to establish sleep extension durability and effectiveness as a public health intervention, especially as it relates to potentially mitigating cardiometabolic or other chronic disease risk.

Regularity

Our intervention focused on increasing time in bed and thereby TST. While we helped participants establish goal bedtimes and waketimes, they were told that increasing their total time in bed was more important than maintaining goal bed and wake times. Compared to the population study with 60 977 UK Biobank (62.8 ± 7.8 years old) participants [23], our participants’ average SRI was in the fourth quintile with 16 participants scoring in the bottom two quintiles. These results suggest our participants had relatively poor sleep regularity and behavioral sleep extension is unlikely to improve sleep regularity in such a population. Furthermore, our exploratory analysis showed no association between TST and the SRI, providing additional evidence that interventions targeting TST have a low probability of impacting sleep regularity, at least among our young adult population. As poor regularity is associated with poor health outcomes [25, 65–68], including increased mortality risk [23], our findings suggest alternate interventions targeting sleep regularity may be required to help improve health in people with poor regularity. For example, one hypothesis is that multi-component interventions targeting sleep extension and circadian timing may help simultaneously improve multiple dimensions of sleep health, similar to prior work with multidimensional interventions in adolescents [33].

Satisfaction and alertness

Average baseline PSQI and ESS scores were above thresholds defining significant sleep problems and high daytime sleepiness, respectively. Average postintervention scores both improved and indicated good sleep quality and low daytime sleepiness. Our findings demonstrate sleep extension has the potential to improve subjective components of sleep, even with a slight increase in WASO. Interestingly, the amount of increased TST was positively correlated with improved ESS scores. Together, these findings show that some improvement in sleepiness can occur without necessarily achieving the recommended 7 hours of sleep per night and without improving sleep regularity. Moreover, these findings raise the hypothesis that improved TST in response to behavioral sleep extension is a key factor leading to improved daytime sleepiness. Thus, participants who achieve greater increases in TST are likely to experience less sleepiness. However, changes in the PSQI were not associated with changes in TST, suggesting that improvements in the PSQI may not be directly driven by increased TST in the context of this behavioral intervention. In summary, these findings highlight that different dimensions of subjective sleep health may respond differently to behavioral sleep extension and underscore the need for future research to clarify causal pathways and identify which aspects of sleep are most amenable to targeted interventions.

Some prior sleep extension studies in adults have also measured subjective outcomes. In general, findings show measures of sleep quality, sleepiness/fatigue, and insomnia symptoms improve [57–59] or remain unchanged [60], but these findings have not been specifically linked to changes in TST or other dimensions of sleep, as we report here. As subjective measures of sleep are linked with health outcomes including mortality [69], future research is needed to identify which components of behavioral sleep interventions most effectively improve subjective sleep outcomes and, ultimately, overall health. Such knowledge will be critical for designing more targeted and effective behavioral sleep interventions.

Timing—sleep onset and offset

Increased TST was mostly achieved by participants shifting sleep onset earlier without significant shifts in sleep offset at the group level. However, there was individual variability with 20 participants showing directionally later sleep offsets ranging from 06:05 to 12:34, despite the lack of a significant group-level change. These data highlight that while the most common strategy was to shift sleep onset earlier, participants use a variety of approaches involving both sleep onset and offset. Despite the importance of sleep timing, changes in sleep onset and offset are only reported in a few prior sleep extension studies in adults [33, 60]. Our data are consistent with these few prior findings that generally show earlier sleep onset as the primary method to achieve sleep extension. Although still limited, in aggregate, these data suggest that nonshiftworkers may have more flexibility to manipulate bedtimes versus waketimes. We did not specifically ask participants to explain their individual sleep schedule choices, but they were instructed to identify sleep times that best fit their existing schedules and responsibilities. Future work is needed to understand if tailored sleep extension strategies, specifically targeting sleep onset or offset, may benefit specific populations or occupational groups, an important public health consideration that is beyond the scope of our current data.

Efficiency

Our finding of no change in sleep efficiency is consistent with another 6-week sleep extension study in adults [33]. In experimental sleep restriction studies, sleep efficiency tends to improve as people are typically more efficient at sleeping when homeostatic sleep pressure is increased [70]. On the other hand, worse sleep efficiency is linked to detrimental effects on cardiometabolic health [10, 22, 71, 72]. In our study, we observed a nonsignificant trend toward slightly lower sleep efficiency, but the effect size was very small and is not considered physiologically meaningful for this population. Moreover, as a change in TST was not associated with a change in efficiency, this suggests that within the range of changes in TST in our study, there is little risk for greater sleep extension to adversely impact sleep efficiency. In aggregate, our data suggest that moderate sleep extension in healthy adults is unlikely to impair sleep efficiency. However, if similar interventions were applied to older adults or other populations, larger changes in TST could potentially impact sleep efficiency, highlighting the need to monitor this outcome when designing behavioral sleep extension interventions.

Sleep architecture

DREEM data showed an increase in TST of ~62 minutes, whereas actigraphy data showed an increase of ~47 minutes, though fewer nights of DREEM data were analyzed because it was only worn during the final week of baseline and sleep extension. DREEM data also showed participants increased their time in stage N1 (large effect), stage N2 (medium effect), and REM sleep (small effect). One prior sleep extension study in adults incorporated ambulatory EEG [33] and detected increases in stage N1 and N2, whereas we detected increased Stage N1, Stage N2, and REM. This may have been due to differences in the length of the study as their sleep extension was 6 weeks, differences in data collection method (DREEM versus PSG), or the number of nights of data collection (1 night of data per participant versus 1 week in our study). The increase in REM is noteworthy, given that disrupted REM sleep is linked to poor cardiovascular, metabolic, and neurocognitive health outcomes [73, 74]. Although the effect size was small, the directional increase in REM supports the hypothesis that sleep extension may help restore REM sleep, given the association between change in TST and change in REM. This hypothesis warrants testing in larger randomized controlled trials. Related to N3, increases in N3 during recovery sleep tend to dissipate as homeostatic sleep pressure is relieved [70, 75]. It is possible our participants experienced increased N3 during the first weeks of intervention; however, we did not quantify sleep staging during these segments. Finally, the positive associations between increased TST and each sleep stage suggest a dose–response relationship in which larger gains in TST may yield proportionally greater improvements across multiple aspects of sleep architecture. Whether these relationships persist with smaller or larger TST changes, or extend over longer intervention periods, remains unknown. Overall, these findings highlight behavioral sleep extension as a practical method to shift sleep architecture toward patterns associated with better health while also underscoring the need for future work to define the magnitude, timing, and durability of such physiological benefits [76, 77].

Limitations

Our study has some important limitations to consider. There was no control group, and it is possible that the “Hawthorne effect” [78] contributed to some improvements, especially subjective sleep outcomes, as participants knew that they were in a sleep extension study. Additionally, our findings unlikely translate to older adults or people with anxiety, depression, sleep disorders, or chronic disease. Future research focused on sleep extension interventions in these populations is an important area to help advance the field. Although studying people across several weeks in their home free-living environment enhances generalizability, this approach necessarily introduces some limitations. Notably, ambulatory monitoring must balance methodological rigor with participant burden and research costs. We therefore chose to quantify sleep using the Philips Actiwatch Spectrum Plus (a common wearable device in the sleep field until its discontinuation in December 2022) because of its long battery life and established use in the sleep field. We also chose the DREEM headband to capture EEG data because of its ease of use, comfort, and feasibility for multi-night in-home recordings. Both these devices minimized participant burden, were cost-effective, and provided longitudinal data. However, both devices use proprietary algorithms to analyze the raw data, which limits transparency and can impede direct comparison with other analytic platforms used in other studies. Despite this limitation, our findings are consistent with prior data [9] and the EEG-derived analyses represent an important advance in the context of behavioral sleep extension. Nonetheless, replication in future trials with open-source algorithms is a critical step to build on our current findings.

Conclusion

In our study of young healthy adults, a 4-week behavioral sleep extension intervention impacted some, but not all, dimensions of sleep. Importantly, average TST increased by ~47 minutes per night, mainly achieved by earlier sleep onset. This change in TST was strongly correlated with improved daytime sleepiness measured by ESS. Sleep staging analyses showed increases in stage N1, N2, and REM sleep. On the other hand, our participants had generally poor sleep regularity that did not change with the intervention, suggesting alternate strategies are needed to improve sleep regularity. Together, our findings help establish efficacy of sleep extension as an experimental intervention in adults and identify which dimensions of sleep are most likely to be impacted by sleep extension, which will help open new doors of research. An important consideration for future work will be to understand the potential health benefits of such sleep extension interventions using larger randomized controlled trials. Such knowledge could help refine and improve the health and safety benefits of sleep extension interventions and inform the need for simultaneously targeting sleep regularity, especially for the goal of improving other clinically relevant outcomes including cardiometabolic disease and all-cause mortality.

Author contributions

Audrey Patrice Stegman (Data curation, Formal analysis, Investigation, Methodology, Project administration, Visualization, Writing—original draft, Writing—review & editing [equal], Funding acquisition, Resources [supporting]), Michelle Kubicki (Data curation, Investigation, Methodology, Project administration, Writing—review & editing [equal], Formal analysis [supporting]), Zachary Mallender (Data curation, Project administration [supporting], Writing—review & editing [equal]), Grace A. Zimmerman (Data curation, Formal analysis [supporting], Investigation, Writing—review & editing [equal]), Krishna M. Sundar (Formal analysis, Funding acquisition, Methodology, Resources [supporting], Investigation [equal], Project administration, Writing—review & editing [equal]), Kelly Glazer Baron (Funding acquisition, Investigation, Methodology [supporting], Writing—review & editing [equal]), Nichole Reisdorph (Conceptualization, Resources [supporting], Funding acquisition, Writing—review & editing [equal]), Kenneth P. Wright Jr (Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Writing—review & editing [equal]), and Christopher Michael Depner (Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision [lead], Data curation, Formal analysis, Investigation, Validation, Visualization, Writing—original draft, Writing—review & editing [equal])

Funding

NIH-UM1TR004409; NIH-K01HL145099; NIH-T32DK110966; NIH-R01HL166733; University of Utah Seed Grant-10060570; Ben B. and Iris M. Margolis Foundation; University of Utah Vice president for research-Research Instrumentation Fund; and University of Utah Diabetes and Metabolism Research Center Graduate Student Fellowship.

Disclosure statement

Financial disclosure: A.P.S. receives research support from Labfront, unrelated to this work.

Dr. Wright received Research support/donated materials from DuPont Nutrition & Biosciences, Grain Processing Corporation, and Friesland Campina Innovation Centre, unrelated to this work.

Dr. Depner received funding from the Ben B. and Iris M. Margolis Foundation unrelated to this work.

Non-financial disclosure: Dr. Sundar is a co-investigator for investigator-initiated research on cancer outcomes with CPAP (Sponsor - Phillips -Respironics USA), unrelated to this work and for which no monies have been received. Dr. Sundar's other industry relationships are for cough-related research, advisory board at Trevi Therapeutics, Oct 2025, and consultant at GSK Pharma.

Data availability

Data will be available upon request.