Earlier bedtimes and more sleep displace sedentary behavior but not moderate-to-vigorous physical activity in adolescents

Abstract

Objectives:

Correlational models suggest increased cardiometabolic risk when sleep replaces moderate-to-vigorous (but not sedentary or light) physical activity. This study tested which activity ranges are impacted by experimentally altering adolescents’ bedtime.

Method:

Adolescents completed a 3-week within-subjects crossover experiment with 5 nights of late bedtimes and 5 nights early bedtimes (6.5- and 9.5-hours sleep opportunity, respectively). Experimental condition order was randomized. Waketimes were held constant throughout to mimic school start times. Sleep and physical activity occurred in the natural environments, with lab appointments following each 5-day condition. Waist-worn accelerometers measured physical activity and sedentary behavior. Wrist-worn actigraphs confirmed sleep condition adherence. Wilcoxon tests and linear mixed effects models compared waking activity levels between conditions and across time.

Results:

Ninety healthy adolescents (14–17 years) completed the study. When in the early (vs. late) bedtime condition, adolescents fell asleep 1.96 hours earlier (SD = 1.08, d = 1.82, p < .0001) and slept 1.49 hours more (SD = 1.01, d = 1.74, p < .0001). They spent 1.68 and 0.32 fewer hours in sedentary behavior (SD = 1.67, d = 1.0, p < .0001) and light physical activity (SD = 0.87, d = 0.37, p = .0005), respectively. This pattern was reflected in increased proportion of waking hours spent in sedentary and light activity. Absolute and proportion of moderate-to-vigorous physical activity did not differ between conditions (d = 0.02, p = .89; d = 0.14, p = .05, respectively).

Conclusions:

Inducing earlier bedtimes (allowing for healthy sleep opportunity) did not affect moderate-to-vigorous physical activity. Alternatively, later bedtimes (allowing for ≤ 6.5 hours of sleep opportunity, mimicking common adolescent school night sleep) increased sedentary behavior. Results are reassuring for the benefits of earlier bedtimes.

Introduction

Sleep deprivation is widespread among American adolescents, with less than one-third meeting clinical recommendations on school nights for 8–10 hours of sleep.¹ Adolescent research to date has largely focused on the negative consequences of short sleep, including strong links to obesity development.² In response, researchers and public health experts have called for efforts to encourage longer sleep to attenuate obesity rates.^3,4 While there has been some success with shifting morning demands later to encourage longer sleep, US high schools continue to start early (around 8 AM, on average⁵). Thus, when it comes to individual adolescents, earlier bedtimes are the primary venue for longer school night sleep. However, there are concerns that earlier bedtimes might limit time adolescents engage in other health behaviors, such as physical activity.

Findings from empirical studies that address the interplay among sleep, physical activity, and sedentary behavior in adolescents are mixed and often interrelated. Recent review articles^6,7 have variably found variable and seemingly contradictory associations between adolescent sleep duration and daytime physical activity. Furthermore, physical activity, sedentary behavior, and sleep health contain a myriad of bidirectional relationships,⁸ making it challenging to infer causality, particularly with cross-sectional and observational approaches. Importantly, few studies have accounted for the co-dependency of sleep and waking hours. Within the 24-hour day, sleep and movement behaviors (sedentary behavior, light physical activity [LPA], and moderate to vigorous physical activity [MVPA]) inherently displace one another. When one increases, one or more of the others must decrease. While this contributes to our understanding of the interrelatedness of these constructs, it also raises questions about the possible opportunity cost of sleep; could earlier bedtimes (and thus longer sleep) “steal” valuable waking hours that could be spent on other healthy behaviors, like MVPA?

Recent analytic advances have opened the door to more sophisticated ways of answering time displacement questions, at least on a cross-sectional basis.^9,10 Matricciani and colleagues¹¹ used compositional analyses to examine how the reallocation of sleep time to awake/active time was associated with cardiometabolic health outcomes in children and adults. Their statistical models among children indicated that reallocating 20–60 minutes of sedentary behavior or LPA to sleep had neutral-to-positive associations with healthier body mass index (BMI), blood pressure, blood lipids, inflammatory markers, and overall metabolic health. However, their models predict significantly worse outcomes across these variables when sleep displaces even 20 minutes of MVPA. These findings suggest that if public health efforts to increase sleep come at the expense of MVPA, negative cardiometabolic health outcomes could be an unintended consequence. While that study used sophisticated compositional analyses to model what might happen to daytime activity levels after increasing sleep time, they were cross-sectional and could not infer causality.

Experimental studies are needed to causally identify what actually happens to physical activity with increased sleep. In the only published adolescent experimental study on the topic, our lab found that shifting 18 short-sleeping adolescents’ bedtimes ~1 ¹/₂ hours earlier resulted in longer sleep, less sedentary behavior during waking hours, and no impact on MVPA.¹² However, findings were considered preliminary due to the small sample and modest sleep manipulation (adolescents averaged less than the recommended 8–10 hours per night¹³). If, as that preliminary work suggests, earlier bedtimes exclusively offsets sedentary behavior, this could be leveraged to help attenuate high rates of adolescent obesity.¹⁴ However, if earlier bedtimes paradoxically impinge upon MVPA opportunities, this could have sobering implications for sleep optimization interventions. Further, as sleep scientists increasingly focus on the timing of sleep and obesity-promoting behaviors, there have been calls to explicate how sleep restriction impacts when adolescents engage in various movement behaviors.¹⁴ For example, shorter sleep might lead adolescents to be broadly less energetic across the day or might impact activity levels more at certain times of day.

This study aimed to use experimental methods in a more definitive sample and with a more robust sleep manipulation to examine the effect of longer sleep (via earlier bedtimes) on displacement of sedentary behavior, LPA, and MVPA among healthy adolescents. Adolescents’ sleep and activity levels were measured at home as they engaged in a within-subjects crossover experiment. Each condition was 5 consecutive nights with either late bedtimes set to allow 6.5 hours sleep opportunity (designed to mimic many adolescents’ school night schedules) or early bedtimes set to allow 9.5 hours sleep opportunity (designed to allow healthy sleep duration¹³). Our primary aims were to determine the effect of sleep length manipulation on 1) the proportion of waking hours spent in sedentary behavior, LPA, and MVPA, and 2) the absolute accumulated daily minutes spent in each of these activities. Based on preliminary findings, we hypothesized that during the late (vs. early) bedtime condition, adolescents would spend a greater proportion of waking hours and more accumulated minutes in sedentary activity; they would show no significant changes in absolute or relative time spent in LPA or MVPA. Secondarily, we also explored the timing of when sedentary behavior, LPA, and MVPA accumulated across the day in each bedtime condition.

Participants and methods

The Institutional Review Board at Cincinnati Children’s Hospital Medical Center approved and supervised all study procedures. Data were collected exclusively during the nonschool summer months of 2015–2018 to avoid ethical concerns arising from the negative impact of restricted sleep on school performance. Adolescents and parents provided verbal and written informed assent and consent to participate, respectively.

Recruitment and participants

Healthy adolescents (ages 14–17) were recruited via online advertisements, community flyers, and emails sent to employees within a pediatric health care network. Interested participants were initially screened over the phone for study eligibility, which was later confirmed in-person during the first study visit. Participants were deemed ineligible if they: had selfor parent-reported unusually short or long habitual sleep (< 6 or > 10 hours on school nights) or symptoms of ongoing sleep disorders measured via the Michigan Pediatric Sleep Questionnaire (ie, obstructive sleep apnea, period limb movement disorder^16,15); obligations that would preclude them from adhering to the study protocol sleep timing (eg, vacation mid-protocol); daily caffeine consumption > 1 serving of coffee/energy drink or > 2 servings of soda; current psychiatric conditions sensitive to sleep restriction or adherence to study procedures (eg, bipolar disorder, depression, psychosis, eating disorder); intellectual disability by history or IQ < 70 on a brief screener¹⁷; history of a neurological injury or illness (eg, traumatic brain injury, epilepsy); or a BMI > 30 (to reduce risk of comorbid sleep apnea and to take an obesity prevention approach). Participants were compensated up to $285 for completing all aspects of the study.

Sleep manipulation procedures

The sleep manipulation protocol was previously described in Beebe and colleagues,^18,19 who reported on studies with smaller subsets of the current sample and/or with unrelated outcome measures (eg, performance on a concussion screener; reward-related responses to food stimuli). Adolescents slept at home. At the start of the experimental protocol, participants were instructed to wake up at the time that they would need to arise to consistently attend 8:00am meetings at the study site. This wake-time was held constant throughout the course of the study. This was intended to mimic the early rise times for US adolescents during the academic year. Participants were instructed to avoid napping and to minimize caffeine (≤ 1 cup of coffee or energy drink/day, and/or ≤ 2 cups of soda/day). The protocol was broken into three 5-night periods, separated by 2-night “washout” periods (see Fig. 1 for study flow).

Figure 1.

Sleep manipulation design. Note: Wake times were held constant throughout the entire study procedures, set at a time that would allow adolescents to attend 8 AM Saturday study visits. Most (75%) of wake times fell between 6:30 AM and 7:30 AM for both early and late bedtime conditions.

Participants spent the first 5 nights in a “sleep stabilization” period, which was designed to provide the opportunity to acclimate to the monitors and shift to a school-night sleep schedule (early wake time, bedtime determined by the adolescent). The morning after the fifth sleep stabilization night, adolescents and their caregiver attended their first 8 AM study appointment. During that visit, study staff uploaded objective sleep and activity monitors (detailed below). Staff reviewed adolescents’ adherence to wearing the units and following wake time instructions, along with the adolescents and their caregiver. Adolescents whose wake time deviated from the sleep instructions by >1 hour were deemed ineligible.

Eligible participants then entered a within-subjects cross-over protocol involving 2 experimental conditions, with condition order randomly counterbalanced across participants. Each condition was 5 consecutive nights with bedtimes set to achieve a sleep opportunity of either 6.5 hours (late bedtime condition) or 9.5 hours (early bedtime condition). After completing the first experimental condition and before entering the second, adolescents underwent a 2-night washout period. During the washout, bedtimes were more flexible, but adolescents were asked to schedule at least 8 hours sleep opportunity. Participants attended an 8:00am appointment after each 5-night experimental condition, during which study staff uploaded and reviewed sleep and activity monitors and provided instructions for the subsequent sleep condition (see Fig. 1 for study flow).

Measures

Adolescents wore 2 separate accelerometers to objectively monitor sleep and waking activity levels, placed according to the best available validation evidence.

Sleep monitoring

Sleep was assessed with a wrist-mounted actigraph (SleepWatch; Ambulatory Monitoring Incorporated, Ardlsey, NJ), worn on the non-dominant wrist from bedtime to awakening. Adolescents completed a daily sleep diary that asked about bedtimes and rise times. When used in conjunction with sleep diaries, actigraphy is a validated way of determining objective, nonbiased estimates of healthy adolescents’ sleep-wake patterns in their home environment.²⁰ At each 8 AM visit, study staff downloaded actigraphy data in 60-second epochs. After reconciling any discrepancies between the sleep diary and visual inspection of actigraphy data with the adolescent and caregiver, staff set designated windows in the actigraphy record to run a pediatricvalidated algorithm.²¹ This resulted in the following sleep parameters: sleep onset, sleep offset, sleep period (sleep onset to offset), sleep duration during that sleep period (excluding bouts of wakefulness), and sleep efficiency (sleep duration divided by sleep period). There were rare occasions when participants had no actigraphy data after a condition (eg, watch malfunction; 0 occasions - stabilization week, 3 occasions - early bedtime, 1 occasion - late bedtime). They were considered adherent to sleep instructions if they had plausible self-report sleep diary data that was corroborated with physical activity monitor wear time patterns (0 occasions — stabilization week, 1 occasion - early bedtime, 1 occasion - late bedtime).

Physical activity monitoring

Daytime activity was measured using a waist-worn accelerometer (ActiGraph wGT3x-BT; ActiGraph, Pensacola, FL), worn during waking hours on the dominant hip, fastened with an adjustable belt. Participants were instructed to put on the accelerometers immediately upon awakening and to remove them upon getting into bed for the night, removing only if the device was at risk of damage (eg, submerged in water). Waist-worn accelerometers show strong reliability and validity estimates for activity levels, can discriminate the intensity of physical activity, and have been used extensively for activity assessment in youth.²² Data were downloaded and analyzed in 30-second epoch increments using Actilife software (ActiGraph, Pensacola, FL). A day of data was considered valid with ≥ 10 hours of recorded accelerometer wear time (determined via the manufacturer’s algorithm and visually verified by condition-blind study staff). To be included in analyses, adolescents needed ≥ 3 (of 4) valid days between Tuesday-Friday from each condition, and were dropped from all analyses if they failed to meet this criterion. Outcomes derived from valid wear time included 1) average accumulated minutes across valid days spent in sedentary activity (0–100 activity counts per minute), LPA (101–2295 counts per minute), and MVPA (≥ 2296 counts per minute),²³ and 2) percentage of waking hours (spent wearing the accelerometer) spent in sedentary activity, LPA, and MVPA.

Demographic characteristics and chronotype

During the initial 8:00am office visit (immediately following sleep stabilization), caregivers provided basic demographic information including adolescent age, sex, ethnicity, and race, and yearly household income. Adolescents also completed Acebo and Carskadon’s 10-item Adolescent Chronotype Questionnaire²⁴ about their morning-ess vs. eveningness preferences. Adolescent height and weight were measured in triplicate on hospital-calibrated scales and used to calculate age- and sex-adjusted z-scores for BMI, based on US Centers for Disease Control and Prevention data.

Analytic plan

All analyses were performed using SAS version 9.4 (Cary, NC). Preliminary analyses first compared everyone who entered the study against the final sample (completers who met sleep adherence criteria and had sufficient sleep and physical activity data) on age, race, sex, family income, chronotype, and BMI z-score. We then used paired-samples t-tests to characterize differences in sleep duration between the late bedtime and early bedtime conditions to determine study adherence.

Our primary aim focused on the impact of the sleep manipulation on sedentary behavior, LPA, and MVPA. While we considered forms of compositional analyses,¹¹ we were unable to find well-characterized applications to within-subjects experimental trials. Fortunately, more traditional analyses are still helpful; if earlier bedtimes exclusively resulted in reduced sedentary behavior, this would support their utility in promoting health outcome, but if earlier bedtimes reduce MVPA, then they could have an unwelcome cost. Accordingly, we computed difference scores in both accumulated minutes and percentage of waking hours spent in the 3 primary categories of activity (sedentary behavior, LPA, and MVPA) by subtracting late bedtime values from early bedtime values. Due to substantial skew in difference scores, we used the Wilcoxon signed rank test to test whether the sedentary, LPA, and MVPA difference scores statistically differed from 0. Significance level alpha was set at .05 for these primary analyses. We also ran exploratory linear regressions on each difference score to probe for moderating effects of age, sex, race, income, or order of randomization.

Our secondary aim was to explore the temporal patterning of sedentary activity, LPA, and MVPA across waking hours between conditions. Using an approach analogous to a prior study on dietary intake in a separate sample,²⁵ minutes spent in each type of activity were broken into 9 two-hour bins (ie, 6–8 am, 8–10 am, 10–12 pm, 12–2 pm, 2–4 pm, 4–6 pm, 6–8 pm, 8–10 pm, 10 PM-12 AM; activity was minimal from 12–6am so those hours were omitted from this exploratory analysis). To investigate whether activity trajectories differed between early and late bedtime conditions, linear mixed effects models examined the cumulative minutes of each activity type on sleep condition, time block, and the time by condition interaction. P-values were adjusted for the number of outcomes in this analysis using Bonferroni’s method with alpha set at 0.017 (.05/3).

Results

Preliminary analyses

Of the original 149 recruited adolescents, 32 were excluded prior to randomization due to nonadherence to study instructions during the sleep stabilization week or to in-person discovery of ineligibility (eg, symptom report on a structured psychiatric interview). Of the 117 randomized adolescents, 10 dropped in the first experimental condition and 6 dropped in the second experimental condition. An additional 11 were removed from analyses due to having ≤ 3 days of valid physical activity data during 1 or both experimental conditions. The final sample included 90 adolescents who completed both sleep conditions with ≥3 days of valid physical activity data during each condition. See Table 1 for full demographic characteristics. The final sample did not differ significantly from the excluded sample on age, race, sex, family income, chronotype, or BMI z-score.

Table 1.

Sample demographics at time of randomization

	Excluded (n = 27)	Final sample (n = 90)	p
Age (years, mean ± SD)	15.81 ± 0.92	15.72 ± 1.08	.69
Sex (female)	59%	67%	.48
Race/ethnicity			.65
White/Caucasian (non-Hispanic)	52%	61%
Black/African American (non-Hispanic)	33%	29%
Multiracial/other	15%	10%
Family income			.73
<$50,000	33%	26%
$50,000 – $100,000	30%	32%
>$100,000	37%	42%
BMIz (z-score, mean ± SD)	0.32 ± 0.99	0.34 ± 0.87	.94
Chronotype questionnaire	25.15 ± 3.74	25.67 ± 3.20	.48
Stabilization week actigraphy sleep outcomes
Sleep onset (SD in h)	24:02 ± 0.89	24:12 ± 1.06	.47
Sleep offset (SD in h)	7:10 ± 1.03	7:01 ± 0.65	.35
Sleep period (h)	7.13 ± 1.15	6.81 ± 0.97	.16
Sleep duration during sleep period (h)	6.47 ± 1.19	6.17 ± 1.00	.19
Sleep efficiency	91% ± 5.29	91% ± 6.49	.74

Note: BMI, body mass index; chronotype questionnaire, Acebo and Carskadon’s 10-item Adolescent Chronotype Questionnaire; stabilization week, 5 nights prior to condition randomization with wake times set to accommodate 8 AM study appointments; sleep onset, actigraphy algorithm-defined time of first sleep onset after bedtime; sleep offset, actigraphy algorithm-defined final wake time; sleep period, sleep onset to sleep offset; sleep duration during sleep period, sleep period excluding wake after sleep onset; sleep efficiency, sleep duration/sleep period.

There were no significant differences (p > .05) on any of these variables when comparing the final sample to those who were randomized but either dropped out or had inadequate data.

Time of sleep onset was significantly later during late bedtime (12:42 AM, SD = 0.33) compared to early bedtime (10:44 PM, SD = 0.61), with nearly identical average sleep offset times (7:04 AM and 7 AM, respectively). This resulted in a statistically significant difference in average nightly sleep period of 1.90 hours between conditions (late bedtime sleep period M = 6.35 hours, SD = 0.61; early bedtime period M = 8.27 hours, SD = 0.90; p <.0001, d = 1.81; See Table 2). That difference in sleep period was nearly identical to daily wear time differences for the physical activity monitor (1.98 hours longer in late vs. early bedtime conditions), providing reassurance that any effects on physical activity metrics were not due to removal of the monitors beyond intended differences in time spent awake.

Table 2.

Changes in sleep and activity levels across experimental conditions (N=90)

Sleep outcomes-actigraphy	Early bedtime Mean ± SD	Late bedtime Mean ± SD	Effect size	p value
Sleep onset (time)	22:44 ± 0.61	24:42 ± 0.33	1.81	<.0001
Sleep offset (time)	7:01 ± 0.26	7:04 ± 0.27	0.16	.12
Sleep period (h)	8.27 ± 0.90	6.35 ± 0.61	1.74	<.0001
Sleep duration during sleep period (h)	7.33 ± 1.01	5.84 ± 0.59	1.45	<.0001
Sleep efficiency	89% ± 6.67	92% ± 5.01	0.71	<.0001
Activity outcomes – accelerometry	Mean ± SD	Mean ± SD
Average daily sedentary behavior (h)	9.72 ± 1.34	11.39 ± 1.80	1.00	<.0001
Proportion of waking hours	70.3% ± 9.3	71.8% ± 8.7	0.21	.02
Average daily LPA (h)	3.67 ± 1.09	3.99 ± 1.12	0.37	.001
Proportion of waking hours	26.3% ± 7.0	25.2% ± 7.0	0.20	.05
Average daily MVPA (h)	0.49 ± 0.56	0.47 ± 0.53	0.02	.89
Proportion of waking hours	3.5% ± 3.9	3.0% ± 3.3	0.14	.05
Overall accelerometer wear time (h)	13.87 ± 1.08	15.85 ± 1.51	4.61	<.0001

Note: Early bedtime, bedtime set to allow 9.5 hours/night sleep opportunity for 5 nights; late bedtime, bedtime set to allow 6.5 hours/night sleep opportunity for 5 nights; sleep onset, actigraphy algorithm-defined time offirst sleep onset after bedtime; sleep offset, actigraphy algorithm-defined final wake time; sleep period, sleep onset to sleep offset; sleep duration during sleep period, sleep period excluding wake after sleep onset; sleep efficiency, sleep duration/sleep period; LPA, light physical activity; MVPA, moderate to vigorous physical activity; proportion of waking hours, % of waking hours spent in each activity level. Wake times held constant across all study conditions and set to approximate school year wake times for 8 AM school start time. Effect size (Cohen’s D) for sleep outcomes calculated from t-test results; effect sizes for activity outcomes calculated from Wilcoxon signed rank test.

Impact of sleep manipulation on overall activity (primary aim)

As noted in Table 2, compared to late bedtime, adolescents averaged 1.68 fewer hours/day of sedentary behavior (SD = 1.67, p = .0001, d = 1.00) and 0.32 fewer hours/day of LPA (SD = 0.87, p = .001, d = 0.37) during early bedtime. Neither time spent in MVPA (p = .89, d = 0.02) nor proportion of waking time in MVPA differed between conditions in any meaningful way (see Table 2). The difference in the 3 activity measures between conditions was not significantly moderated by age, sex, race, family income, or order of the experimental conditions (p > .05, alpha set at .017).

During the early (vs. late) bedtime condition, adolescents’ waking time use shifted slightly, such that they spent 1.1% more waking time in LPA and 0.5% more waking time in MVPA, which approached significance (p = .05; see Table 2). Conversely, there was a statistically significant decrease of 1.5% of waking hours spent in sedentary activity (p = .02, d = 0.21). These effects were not moderated by age, sex, race, family income, or experimental condition order (p > .05, alpha set at 0.017).

Temporal patterning of activities across waking hours (secondary aim)

Using linear mixed effect models, we found a significant condition by time interaction for accumulated time spent in sedentary behavior (F(8, 1839) =15.81, p < .0001) and LPA (F(8, 1839) = 2.48, p = .0114), indicating differing trajectories between the late and early bedtime conditions. Follow-up analyses revealed that cumulative minutes of activity differed between conditions in the final time bin only (ie, 10 PM-12 AM) for both sedentary activity (p < .0001) and LPA (p = .002), see Fig. 2. No time bin effects were evident for MVPA (p > .05). See Table 3 for accumulated totals.

Figure 2.

Accumulation of time spent in various activity levels in late vs. early bedtime. Note: Early bedtime = bedtime set to allow 9.5 hours/night sleep opportunity for 5 nights; Late bedtime = bedtime set to allow 6.5 hours/night sleep opportunity for 5 nights; Sedentary = sedentary behavior (0–100 movement counts/minute); Light = light physical activity (101–2295 counts/minute); Moderate-to-vigorous physical activity (≥ 2296 counts/minute). *p < .05 comparing values between late bedtime and early bedtime conditions during the 10 PM-12 AM bin only

Table 3.

Accumulated minutes in activity across the day by group

	Sedentary behavior		LPA		MVPA
	Early bedtime	Late bedtime	Early bedtime	Late bedtime	Early bedtime	Late bedtime
6 AM	0	0	0	0	0	0
8 AM	27	26	8	8	1	1
10 AM	100	100	31	30	5	4
12 PM	181	181	61	59	8	7
2 PM	261	263	92	90	12	11
4 PM	342	345	123	119	16	14
6 PM	421	424	155	151	20	17
8 PM	497	495	186	183	25	23
10 PM	560	570	213	213	28	28
12 AM	573	650	216	233	29	29

Note: Values are accumulated totals (ie, 12 AM represents the total number of minutes spent in sedentary, LPA, or MVPA between 6 AM and 12 AM).

Early bedtime, bedtime set to allow 9.5 hours/night sleep opportunity for 5 nights; late bedtime, bedtime set to allow 6.5 hours/night sleep opportunity for 5 nights; LPA, light physical activity; MVPA, moderate to vigorous physical activity.

Discussion

This study experimentally tested whether re-allocating more time to sleep via earlier bedtimes would displace sedentary behavior, LPA, or MVPA. We found that compared to when going to bed later, adolescents fell asleep nearly 2 hours earlier (~10:45 PM), woke around the same time (~7 AM), and slept 1.49 more hours/night in the early bedtime condition. When in the late bedtime condition, adolescents benefitted from a slight increase in sleep efficiency, likely a homeostatic response to sleep restriction.²⁶ However, sleep quality fell within the broadly healthy range across all conditions. When going to bed earlier, adolescents accumulated 1.68 fewer hours of sedentary behavior and 0.32 fewer hours of LPA — but maintained nearly identical low levels of daily MVPA across both sleep conditions. This pattern was also reflected in an increased proportion of waking hours spent in sedentary behaviors in the late bedtime condition. Of the additional 1.68 hours spent in sedentary behavior during late bedtime, 76% occurred in the late evening, during the awake time that displaced sleep. Taken together, results suggest that increasing sleep via earlier bedtimes primarily displaced sedentary behavior and small amounts of LPA but did not impose on MVPA.

These experimental findings are consistent with prior observational work connecting later adolescent bedtimes with more time spent in specific sedentary behaviors (eg, screen time).^27—31 Although prior correlational studies have yielded inconsistent results, our finding that later bedtimes and less sleep has negligible impact on MVPA fits with our preliminary adolescent experimental work¹² and a previous experimental study with school-aged youth.³² Importantly, the current study expanded this work in several important ways. It was more highly powered than prior pediatric research due its large sample and within-subjects design (differences between conditions for MVPA were nonsignificant and effect sizes were negligible) and used objective accelerometers to monitor both sleep and physical activity. Further, it compared a realistically late bedtime against an earlier bedtime which, when faced with early school start times, is the only realistic option most US adolescents have to get the recommended duration of sleep on school nights.

We are unaware of well-characterized compositional data analyses appropriate for within-subjects experimental studies, but results of these more traditional temporal analyses bolster the argument for simple time displacement of sedentary behavior vs. sleep. The effects of our bedtime manipulation on sedentary behavior and LPA were not evenly distributed across the day. Instead, they emerged in the evening, close to when bedtimes were changing. This finding is not entirely surprising, as MVPA-rich recreational activities (eg, playing basketball) are often limited in the late evening hours. Though we did not assess which specific sedentary behaviors adolescents engaged in during the study, it is reasonable to assume they had easy evening access to materials that promote sedentary behaviors, such as screened devices. Time use studies with college students suggest that media (and particularly smartphone) use is the most common activity that displaces sleep.³³ Other studies show that 86% of American adolescents keep phones in their bedrooms overnight.³⁴ However, it will be interesting to see how prebedtime activities may differ during summer vs. school year in future studies. Although in some ways unsurprising, these findings provide reassurance that going to bed earlier does not result in displacement of MVPA, which prior epidemiological models suggest would have adverse cardiometabolic effects.¹¹ Rather, going to bed earlier resulted in adolescents sleeping more and having less available time to be sedentary, an effect that prior models suggest is either neutral or beneficial for cardiometabolic health in youth.¹¹

The small but statistically significant increase in LPA in the late bedtime condition was unexpected. The effects of LPA have been relatively understudied,³⁵ but there is some evidence supporting health benefits of LPA over sedentary behavior.³⁶ However, late bedtimes induce both sleep restriction and markedly more sedentary behavior, so any benefits related to increased LPA are likely outweighed by the cumulative negative effects of sleep deprivation³⁷ and sedentary time.³⁸ Sedentary behavior has been shown to be a high-risk time for excess calorie intake³⁹ and increases obesity and cardiometabolic risk in youth and adults.^11,40 Because being overly sedentary incurs risk for negative health outcomes, researchers have suggested that interventions that decrease sedentary activities may be more effective and sustainable than those attempting to increase physical activity.⁴¹ Intervening on bedtime may be a novel method for managing adolescents’ sedentary time.

Limitations

Present results should be understood in the context of the study limitations. First, although traditional analytic approaches adequately answered our study questions, we await the explication of compositional data analytic approaches to within-subjects experimental data. The robust effect size for sedentary behavior and negligible effect size for MVPA suggest that findings would still hold under a compositional analysis approach. Regardless, such an approach could account for the inherent collinearity among sleep and activity, better model time displacement, and promote similar advances that compositional analysis approaches have brought to cross-sectional studies.

Second, although we objectively assessed adolescent activity in the natural environment, we did not assess for specific sedentary behaviors. This information would be helpful for informing development of interventions to move bedtimes earlier, particularly when considering possible differences in sleep-displacing behaviors in the summer vs. school year (eg, prebedtime screen time vs. late evening sports or homework). Future studies would benefit from incorporating low-burden measures to provide more comprehensive activity assessments.

Third, some aspects of this study’s sample and design may limit generalizability. We intentionally set our sleep schedules to mimic the early rise times faced by most US adolescents during the school year. There is a risk that “night owls” differentially dropped out of the study or chose not to enroll after learning about it. Reassuringly, our final sample was similar in chronotype to both the excluded participants and to a large epidemiological sample,²⁴ but differential enrollment remains possible. Further, because the study was conducted with relatively healthy-weight adolescents, results may not generalize to youth with sleep problems/disorders (including delayed circadian preferences) or higher BMIs. Also, while intended to mimic feasible changes in sleep on school nights, for ethics and logistical reasons, data were collected in the summer and relied on manipulating bedtimes over a relatively brief duration (5 nights, excluding weekends) to accomplish changes in sleep period. While these findings dovetail nicely with those from a small-sample sleep extension study conducted during the school year,¹² the field awaits more definitive work in a larger sample using longer manipulations (including weekends) during the school year with exploration of additional effects on health and related behaviors. Finally, this study cannot speak to whether changing rise times (a key goal of advocacy around school start times) would produce similar effects; future work should explore this area.

Conclusions

Limitations notwithstanding, findings were reassuring for the benefit of earlier bedtimes for adolescents. When assigned to earlier (vs. later) bedtimes that allowed for healthy sleep duration, adolescents fell asleep ~2 hours earlier and slept ~1.5 hours longer (for an average sleep period of ~8.25 hours). Going to bed earlier did not impact time spent in MVPA, but rather displaced a significant amount of time spent in sedentary behavior. When going to bed later, adolescents accumulated ~1.7 hours more sedentary time, which was primarily spent in the extra evening hours awake being sedentary. Coupled with prior data,¹² findings provide reassurance that promoting longer sleep duration via earlier bedtimes does not adversely affect health-promoting physical activity, and may be an effective strategy for curbing sedentary activity.Uncited References