Effect of changes in children’s bedtime and sleep period on targeted eating behaviors and timing of caloric intake

Abstract

Short sleep is associated with obesity risk. Experimental studies with adults and observational studies with children demonstrate that changes in eating, including increased caloric intake from energy-dense foods and sugar-sweetened beverages as well as increased caloric intake in the evening, may partially account for this increased risk. We therefore examined whether experimental changes in children’s sleep period lead to changes in reported caloric intake from energy-dense snack foods and sugar-sweetened beverages, and in the evening. Thirty-seven children, 8–11 years old, completed a three-week study that used a within-subject randomized cross-over design. Children slept their typical amount for one week and were subsequently randomized to either increase or decrease their typical amount by 1.5 hours/night for one week; the alternate schedule was completed during the third week of the study, creating a 3-hour time in bed difference between the increase and decrease conditions. Sleep was monitored with actigraphy, and dietary intake was assessed with 24-hour dietary recalls. Participants reported consuming 35 calories per day more from sugar-sweetened beverages during the decrease sleep than the increase sleep condition, p = .033. There were no reported differences between conditions from energy-dense snack foods. Although no differences in reported intake were observed earlier in the day, from 2000 h (8:00PM) and later, children reported consuming 132 calories more during the decrease sleep condition than the increase condition, p < 0.001. Shortened sleep achieved by delaying bedtimes led to increased caloric intake in the evening and from sugar-sweetened beverages.

Introduction

Insufficient sleep is associated with obesity risk with observational studies suggesting potentially stronger associations in pediatric populations than in adults.^1–3 Several pathways through which sleep influences weight status have been proposed,^3,4 including those that favor excess energy intake with reduced sleep. A meta-analysis of experimental adult studies demonstrated that compared to a rested condition, mean energy intake increases following partial sleep restriction and that increased caloric (kcal) intake is due in part to increased intake from fat.⁵ Extant experimental studies with adults demonstrate that changes in kcal intake resulting from sleep curtailment may be due to changes in the distribution of energy intake across the day that favor increased consumption late in the evening.^6–8

Findings regarding changes in caloric intake from adolescent experimental studies have been mixed yet point to the potential impact of short sleep on greater appetite for sweets/desserts, increased intake of high glycemic index foods and sugar sweetened beverages (SSBs), and greater overall food reinforcement (i.e., motivation for a food reward).^9–13 Observational studies with children support such findings and suggest an association of short sleep with increased intake of energy-dense foods and snacks,^14–17 and SSBs.^14,16–19 Additional studies also observed associations between later bedtimes, greater kcal intake after dinner,²⁰ less healthy eating behaviors, and increased obesity risk.^21–24 To our knowledge, only one experimental study of the effect of sleep restriction and extension on eating and activity behaviors has been conducted with school-aged children.²⁵ Consistent with adult experimental studies, families reported that compared to long sleep, children consumed 134 kcal/day more during one week of sleep restriction and weighed approximately 0.22 kg more. No reported differences in macronutrients were observed, but data indicated that change in reported kcal intake may have been driven by additional hours awake during sleep restriction.²⁵ To better inform treatment development, it is important to determine the underlying changes in eating behaviors that led to these observed changes in reported kcal intake and weight.

The present study conducted secondary analyses to determine whether observed increases in kcal intake during the decrease sleep condition in the above-noted study were due to changes in intake of energy-dense snack foods (EDSF) and sugar-sweetened beverages (SSBs) and/or to increased intake later in the day. We hypothesized that compared to the rested condition, when sleep restricted, children would report greater intake of calories from EDSF and SSBs and from intake that occurred later in the day.

Methods

Subjects

Thirty-nine children 8–11 years old were enrolled and randomized into a study to determine how changes in sleep impacted eating behaviors.²⁵ Child eligibility criteria included reported time in bed (TIB; time between reported lights off/trying to fall asleep and waking the next day) of approximately 9.5 hours per night and body mass index (BMI) at or above the 5th percentile BMI for age/sex, but not more than 100% overweight. Children were considered ineligible if diagnosed with a medical/psychiatric condition, including a sleep disorder (e.g., ADHD, parasomnia) or reported current medication use that could affect sleep or weight status.

Procedures

Recruitment occurred in southeastern New England using mailers, advertisements, and posted flyers throughout the community/on the Center website from March 2009 to December 2011. Those eligible after completing an initial phone screen attended individual orientations during which written consent was obtained from parents and assent from children. Children then completed a baseline assessment that had a two-fold purpose. First, it determined final eligibility for the study based on family reported TIB, which was confirmed with actigraph-estimated sleep period (i.e., the time between actigraph-estimated sleep onset and offset). If the actigraph sleep period supported that, on average, reported TIB was approximately 9.5 hours/night, the child was deemed eligible. Second, thresholds for prescribing changes in TIB during the two experimental conditions were established from sleep during the baseline week.

Following final eligibility determination, a variably sized permuted blocks randomization procedure (stratified by weight status: healthy weight or with overweight/obesity) was used to randomize children to increase TIB by 1.5 hours/night for 1 week or decrease TIB by 1.5 hours/night for 1 week with the goal of creating a 3 hour/night TIB difference. Order of conditions was counterbalanced. The alternate schedule was completed the subsequent (i.e., final week of the study). Changes in TIB focused on nocturnal sleep (i.e., naps were not permitted) and were achieved by changing bedtimes; wake times remained constant. Adherence was measured throughout with children wearing actigraphs on their non-dominant wrist, completing sleep diaries, and calling the research center twice daily to leave a time-stamped voice message reporting on their bedtime/wake time and any aberrations (e.g., medication use) that may affect study participation. As was previously reported,²⁵ children demonstrated high levels of adherence with an average 141 (CI: 133–148; i.e., 2 hours 21 minutes) minute/night difference in the actigraph-estimated sleep period between conditions.

Families were also asked to complete three 24-hour dietary recalls each week. At the end of each week children returned to the Center to be weighed and measured for height; actigraphs were downloaded and associated sleep-wake data were reviewed. Staff were blind to experimental condition when assessing study outcomes (e.g., dietary intake, weight). Families were compensated for time completing study activities. All procedures were approved by the Institutional Review Board at the Miriam Hospital.

1.1.1. Measures

Demographic and Anthropometric Measures.

Parents reported on basic demographic data. Trained staff measured height and weight in duplicate with children dressed in street clothes without shoes. Weight was measured using a calibrated digital scale (Tanita BWB-800; Arlington Heights, IL) to the nearest 0.1 pounds. Height was measured to the nearest mm with a wall-mounted stadiometer (Perspective Enterprises, Portage, MI). Normative age- and sex-appropriate data from the Centers for Disease Control and Prevention (CDC) were used to determine child weight status (healthy weight versus overweight/obese).²⁶

Sleep.

The actiwatch 2 (AW2; Phillips Respironics, Bend, OR), which has demonstrated reliability and validity for assessing sleep compared polysomnography in children,²⁷ was used to confirm adherence to prescribed changes in sleep. Children wore the AW2 on their non-dominant wrist 24-hours/day throughout the study. Devices were configured to collect data in one-minute epochs using a medium sensitivity threshold and were scored for sleep versus wake using Actiware software, version 5.59.0015. Sleep diaries helped establish sleep onset and wake²⁸ using standard procedures²⁸ to resolve discrepancies between self-report and actigraph-estimated sleep by reviewing with families; issues remaining after this review (e.g., families could not recall) were rectified during consensus meeting with CNH, MAC, and study staff.

Food Intake.

Child food intake was assessed on two randomly selected weekdays and one weekend day during each study week using the USDA automated multiple pass method for 24-hour dietary recalls, considered the most accurate approach in determining child energy intake when compared to the doubly labeled water method.²⁹ Instructions and aids for portion size estimations were provided to families who completed recalls together with blinded staff by phone. The Nutrition Data System for Research (Nutrition Coordinating Center, University of Minnesota, Minneapolis, MN) was used to compute dietary data. Start time of each family-reported eating occasion, foods (and quantities) consumed, and associated caloric intake at each occasion were documented. For the purpose of the present study, we defined an eating occasion similar to that of previous work: the consumption of at least 25 kcal; all foods and beverages consumed within 30 minutes of the start time of such consumption was defined as one eating occasion.³⁰ Consistent with our previous work,³¹ calories from EDSFs (i.e., cakes, cookies, muffins, ice cream, candy, chips) and SSBs (i.e., any beverage with added sugar, excludes 100% fruit juice) were calculated.

Data Analysis

A priori power calculations determined sample size needs for the primary aims of the study. Specifically, we presumed two-sided hypothesis testing with power of 80%, conservatively assumed zero correlation for repeated measures across time, and used a 2-sided type I error (alpha) of .0125 to account for multiple comparisons. Based on preliminary studies, we estimated needing 24 to 42 participants to detect anticipated effect sizes across primary study aims. Specifically, if we were to assume a standard deviation of 1.0, 36 participants would be needed to detect an effect size of .8.

We used repeated measures methods to account for multiple measurements within and between conditions (i.e. a cross over design). Children had meal measurements on up to three days and sleep measurements on up to seven days per condition. To investigate associations of condition with sleep differences and meal time, we used multiple linear regressions estimated by Generalized Estimating Equations (GEE).⁵ We assumed an exchangeable working correlation matrix to account for within subject correlation. We used robust standard errors to account for some skew (i.e., non-normality) in the regression response variables. In the investigation of meal time’s relation with food consumption more broadly, we similarly used a GEE-estimated linear regression of kcal consumption in which we entered time into the model using a linear spline with three knots at 1030 h, 1430 h, and 2000 h. These knots were empirically driven, and seemed to define natural cut-points between breakfast, lunch, dinner, and late evening snacks. We also included the arm indicator and the interaction between arm and each of the mealtime spline terms in this fuller model. We restricted some analyses to data observed at 2000 h or later. We explored the potential role of child gender and condition order on outcomes, but found no evidence of an effect of either variable on the intervention effect so did not include either in our models. We used paired t-tests and McNemar’s tests for data at the individual level as appropriate. All analyses were conducted using STATA (StataCorp, College Station, TX).

Results

Thirty-nine children were enrolled and randomized; 37 (95%) children completed the study (Figure 1). Most children were non-Hispanic White (81%), male (57%), and with a healthy weight status for their age and sex (73%). Mean (SD) age of participants was 9.6 (1.0) years. As reported previously, mean reported caloric intake differed between the increase and decrease TIB conditions by −134 kcal/day, p = .038, 95% CI [−261 −8], d = .52; no differences in macronutrient consumption were observed.²⁵

Figure 1.

Consort flow diagram of progress through stages of the randomized trial.

Note. Primary reasons for exclusion from study participation included being unable to contact families after initial inquiry regarding the study and refusal to participate. Primary reasons for not meeting inclusion criteria included baseline TIB that was not approximately 9.5 hr/night, suspected/diagnosed sleep disorder, or multiple inclusion criteria not being met.

Differences in Energy Dense Snack Foods and Beverages

Children reported consuming more kcal from SSBs during the decrease TIB (129 ± 119 kcal/day) than the increase TIB (94 kcal ± 89 kcal/day) condition, t_paired (1, 35) = 2.21, 95% CI: −67.5 to −2.9, p = .033 (Figure 2), but reported no differences in kcal intake from EDSFs (407 ± 249 kcal/day in increase vs. 380 ± 208 kcal/day in decrease, p = .485).

Figure 2.

Differences in Mean Daily Reported Caloric Intake from Energy Dense Snack Foods (EDSF) and Sugar Sweetened Beverages (SSB) during the Increase and Decrease Sleep Confitions (N = 36).

Mean daily reported caloric intake from energy dense snack foods and sugar sweetened beverages during each experimental condition. Values are means + SEM. The asterisk (*) denotes significant differences in reported intake between conditions. One participant was missing complete dietary data and was therefore not included in analyses focused on changes in EDSF and SSB.

Differences in Timing of Energy Intake

The reported timing of eating occasions differed between conditions. Specifically, the mean reported time of eating occasions during the decrease TIB condition occurred 35 minutes later than during the increase TIB condition, 95% CI 16 – 53, Z =3.69, p<0.001. Differences in the timing of eating occasions were reported for both dinner (mean time 1758 h in increase vs. 1813 h in decrease, p = .027) and snacks (inclusive of any food item reportedly consumed as a snack, not just EDSF; mean time 1450 h in increase vs. 1537 h in decrease, p = 0.009) but not breakfast or lunch. However, there were no reported differences in the number of eating occasions during the decrease TIB condition (5.45, Standard Deviation [SD] 1.41) relative to the increase TIB condition (5.14, SD 1.20), 95% CI of difference (−0.01, 0.65), Z = 1.90, p = 0.058.

Figure 3 depicts reported kcal intake across the day during both conditions. As shown in the Figure, the only difference in kcal intake occurred at 2000 h or later; no other differences were observed earlier in the day. Specifically, a higher proportion of children reported kcal intake at 2000 h or later during decrease (75.0%) relative to increase TIB (11.1%, Chi-squared, 1 degree of freedom [DF]= 21.16, p<0.001). Further, compared to the increase TIB condition, children reported consuming 132 kcal more during decrease TIB from 2000 h on (β = 131.51 per day beyond 2000 h, 95% CI: 87.47–175.56, Z=5.85, p < 0.001). When considered as a percent of total calories consumed in a given condition, kcal consumed from 2000 h on was 7% higher in the decrease TIB condition relative to the increase TIB condition (β = 0.070, 95% CI: 0.047–0.093, Z=5.95, p < .001).

Figure 3.

Reported Caloric Intake Throughout the Day During the Increase and Decrease Sleep Conditions (N = 37).

Focusing specifically on significant differences in reported consumption at 2000 h or later, a greater proportion of children reported EDSF consumption during decrease TIB (6% during increase versus 58% during decrease, Chi-squared, 1 DF=17.19, p<0.001), and SSB consumption during decrease (0% during increase versus 28% during decrease, Chi-squared, 1 DF=10.00, p=0.002). Specifically, children reported eating 65 kcal more from EDSF per day at 2000 h or later during the decrease TIB condition (β = 64.93/day, 95% CI: 39.89 – 89.98, Z=5.08, p < 0.001) and 6 kcal more SSB during decrease (β = 6.10/day, 95% CI 1.59 – 10.60, Z=2.65, p=0.008).

Discussion

Findings demonstrate that changes in children’s sleep opportunity (i.e., time in bed) elicit changes in both reported SSB intake and in the timing of food intake, extending previous findings from adult experimental studies, which have documented similar shifts in food intake following sleep restriction,^6–8 to a free-living sample of school-aged children. They also build upon prior observational studies with children that have demonstrated associations between short sleep and greater reported intake of SSBs¹⁷ as well as later bedtimes and later reported food intake.²⁰ Moreover, findings provide additional context for our previous observations with this sample of children, which demonstrated increased reported kcal intake within the context of sleep restriction, including during the three additional hours awake when sleep was decreased.²⁵ The present secondary analyses suggest that when children sleep less by going to bed later, timing of the dinner meal and snacks shift later, and daily kcal intake increases primarily as a function of additional intake later in the day and in a limited way to increased intake of sugar-sweetened beverages.

Observed shifts in meal timing and increased kcal intake later in the day are likely due to several biologically and behaviorally based factors. For example, there is increased recognition of the interconnectedness of circadian factors not only with sleep timing but also eating behaviors and metabolic outcomes.^32–35 Of particular relevance to the present findings, an experimental study conducted with adults demonstrated the existence of endogenous circadian rhythms for hunger with peak hunger ratings in the evening (approximately 2000 h).³⁴ Importantly, this observed rhythmicity was independent of eating episodes and time since waking.³⁴ These reported peaks in hunger coincide with the time during which we observed differences in children’s reported intake in the present study. Thus, earlier bedtimes may help to decrease kcal intake by circumventing periods when children would naturally be hungrier. It is also possible that going to bed three hours earlier (as was done during the increase sleep condition relative to decrease in the present study) phase advanced endogenous circadian rhythms. The potential underlying circadian mechanisms warrant further study in children.

Alternatively, it has been hypothesized that short sleep may impact obesity risk simply because the increased time awake provides increased opportunity to eat.³ This is particularly salient in our current environment in which excess calories are readily available. It is also possible that increased late-day food intake following sleep restriction may result from potential increased challenges with eating self-regulation as the day wears on. Experimental studies with adults demonstrate that greater kcal intake following sleep restriction is likely due to hedonic processes (e.g., greater appetitive drive),^4,7,36 and adult neuroimaging studies demonstrate activation in brain regions involved in motivation, reinforcement, decision-making, and self-control following sleep restriction.^37,38 Thus, staying awake later in the evening may place individuals at increased risk of excessive kcal intake due to decreased self-regulatory capacity.

Studies with children and adolescents provide additional context for this hypothesis. In adolescents, relative to a rested condition, sleep restriction leads to greater appetite for sweets/desserts, increased intake of high glycemic index foods and sugar sweetened beverages (SSBs), and greater overall food reinforcement.^11–13 However, the impact of experimental changes in sleep on adolescents’ eating behaviors are not always consistent^9,10 nor have findings from experimental studies with children.^25,39 For example, although in a small pilot we demonstrated decreases in food reinforcement (i.e., motivation for a food reward) when sleep was extended in school-aged children,⁴⁰ we also previously observed no changes in food reinforcement in the present experimental study.²⁵ Further, a cross-over study with children 8–12 years old in which typical TIB was both extended and restricted observed no differences in objectively assessed eating in the absence of hunger (i.e., a measure of hedonic eating or food responsiveness).³⁹ Unfortunately, observed learning effects and the relatively small change in the sleep period (i.e., approximately 46 minutes/night) may have limited detection of potential changes in EAH. Taken together, extant work points to the need to continue to understand mechanisms through which sleep affects eating behaviors and weight. In children it will also be important to understand such associations within the context of their development. In contrast to adults who have greater control over their own eating behaviors, parents play an important role in determining what foods are available within the home, whether or not children have access to those foods, and the timing and context for meals.

Findings regarding both the shift in timing of the mean eating occasion and reported increased consumption later in the day are particularly interesting in light of studies documenting the contribution of meal timing in weight regulation. Animal studies demonstrate metabolic effects when fed during the biologic night versus the biologic day-with greater weight gain occurring when fed during the biologic night.^35,41 Emerging work in humans is consistent. For example, one investigation demonstrated that within the context of weight control and a prescribed isocaloric diet, women randomized to consume the majority of calories at breakfast demonstrated greater weight loss than those instructed to consume the majority at dinner.⁴² Further, within the context of sleep restriction, Spaeth and colleagues⁸ observed a greater percentage of calories consumed during late-night hours and associated overall weight gain. The present findings are consistent, and suggest that one important mechanism through which weight gain may occur due to shortened sleep may be increased kcal intake during the evening.

Beyond changes in timing of intake, changes in sleep also led to small yet significant differences in reported SSB consumption, including later in the evening (i.e., from 2000 h and on). The finding regarding SSBs is consistent with a previous experimental study with adults⁴³ and adds to mixed findings among observational studies with children. Although a meta-analysis and additional observational studies have observed associations between short sleep and increased reported intake of sugar-sweetened beverages, others have not.^14,16,17,19 Given links between sugar-sweetened drinks and obesity risk,⁴⁴ findings remain compelling.

No overall differences were observed for EDSF, although there was some evidence suggesting that when bedtimes were delayed during the decreased sleep condition, intake from snacks, and more specifically EDSF, increased after 2000 h. These findings are somewhat consistent with adult experimental studies^6–8,13 and observational studies with children¹⁷ – both of which have demonstrated associations between short sleep and increased intake from energy-dense foods, including snacks. Differences in findings may be due, in part, to different definitions for snacks and snack foods used across studies. Nonetheless, taken together, findings speak to increased risk of kcal intake due to snacks in the evening hours when sleep is shortened via delayed bedtimes.

Strengths of the present study include the randomized cross-over design and high levels of adherence to prescribed changes in sleep. However, these secondary analyses should be considered in light of limitations, including the small study sample and reliance on self-report for dietary intake. Further, there was no washout period between experimental conditions, which is particularly problematic given the potential negative impact on detecting differences in study outcomes due to carryover effects. Prescribed changes in sleep resulted not only in changes in sleep duration but also in large shifts in sleep timing. Emerging evidence supports that beyond sleep duration, the timing of children’s sleep is also important for obesity risk reduction.²³ Thus the relative influence of sleep timing and duration cannot be discerned. Future work should explore how both sleep timing and duration impact children’s eating behaviors and obesity risk.

In sum, findings suggest that changes in children’s nocturnal sleep led to changes in the timing of food intake and differential kcal intake from sugar-sweetened beverages. It will be important to further explore whether these observed changes result from shifts in the circadian timing system and/or simply reflect longer exposure to food rich environments.

Abbreviations:

kcal

calories

EDSF

energy dense snack foods

SSB

sugar sweetened beverages

TIB

time in bed

BMI

body mass index