Independent associations of sleep timing, duration and quality with adiposity and weight status in a national sample of adolescents: The UK Millennium Cohort Study

Short sleep appears to elevate obesity risk in youth; however, sleep is a multidimensional construct, and few studies have investigated parameters beyond duration. The objective of this study was to investigate if sleep onset time, duration, latency and night waking frequency are independently associated with adiposity and weight status in UK adolescents. This was a cross‐sectional observational study of 10,619, 13–15 years olds. Adjusted linear and logistic regressions were used to investigate associations of self‐reported sleep characteristics with adiposity markers (body mass index z‐score and percent body fat) and weight status. Compared with a sleep onset before 10pm, later sleep timing was associated with higher adiposity and higher likelihood of overweight and obesity in boys (after midnight, odds ratio [95% confidence interval]: 1.76 [1.19–2.60]) and girls (between 11pm and 11:59pm: 1.36 [1.17–1.65]). Sleeping ≤ 8 hr, compared with > 9–10 hr, was associated with higher odds of overweight and obesity in both sexes (boys: 1.80 [1.38–2.35]; girls: 1.38 [1.06–1.79]), and so too was sleeping > 10 hr in girls (1.31 [1.06–1.62]), indicating evidence for a U‐shaped association. Also in girls, compared to a sleep latency of 16‐30 min, sleep latencies ≥ 46 min were associated with higher adiposity (46–60 min, beta coefficient [95% confidence interval], percent body fat: 1.47 [0.57–2.36]) and higher likelihood of overweight and obesity (46–60 min: 1.39 [1.05–1.83]), and often as opposed to never waking in the night was associated with higher adiposity (body mass index z‐score: 0.24 [0.08–0.41]; percent body fat: 1.44 [0.44–2.44]). Sleep duration and timing in both sexes, and sleep quality in girls, appear to be independently associated with adiposity and weight status in adolescence, and may be important targets for obesity prevention.


| INTRODUC TI ON
Developmental shifts in sleep physiology predispose adolescents to biologically driven evening alertness. Around the same time parental enforcement of bedtimes typically fades, academic pressures increase, social networks expand, and sedentary screen-based activities, that are stimulating or produce light stimuli, increasingly take place late into evenings; these factors reinforce evening alertness and perpetuate a cycle of late sleep onset (Crowley et al., 2018).
When accompanied by imposed early mornings for school, insufficient sleep and tiredness result; both are prevalent in adolescents worldwide (Gariepy et al., 2020;Kronholm et al., 2015). This has farreaching ramifications for health and well-being, including repercussions for adiposity levels and obesity risk (Matricciani et al., 2019). A plethora of studies have investigated short sleep relative to weightfor-height indices or weight status categories. Meta-analysed results indicate that short sleep elevates obesity risk in youth (Sun et al., 2020).
Sleep is a multidimensional construct of partly overlapping dimensions, and it is increasingly recognised that parameters such as sleep timing and quality may also be influential with regards to health (Miller et al., 2015). A small number of studies have found that later bed and sleep onset times are associated with higher body mass index (BMI) z-score (Jarrin et al., 2013;Olds et al., 2011), waist circumference and percent body fat (%BF; Jarrin et al., 2013), overweight and obesity (Olds et al., 2011;Skjåkødegård et al., 2021), and obesity severity in adolescents (Hayes et al., 2018). Sleep quality encompasses an array of continuity and architecture features (Ohayon et al., 2017). A meta-analysis found that poor sleep quality, which was broadly defined as longer sleep latency, more sleep disturbances, recurrent night waking or lower sleep efficiency, predicted 46% (95% confidence interval [CI]: 1.24-1.72) higher odds of overweight and obesity in young subjects (Fatima et al., 2016).
It has consistently been shown that associations between short sleep duration with higher adiposity are stronger in boys than girls (Matricciani et al., 2019;Sun et al., 2020). Whether or not this may partly be due to sex differences in other sleep characteristics remains unknown, but adolescent girls experience poorer sleep quality than boys (Galland et al., 2017). Whereas duration may be more closely related to adiposity levels in boys, sleep quality could be a stronger determinant of adiposity in girls. It remains a limitation that the few studies conducted to date on sleep timing and quality, and their associations with adiposity, have been conducted in small and select samples, and thus effect modification by sex has not been investigated. Information of this type could help to optimise obesity prevention policies and strategies, by highlighting that sleep components should be tailored according to sex-specific requirements.
It is also important to decipher the possible pathways by which sleep parameters may relate to adiposity. The best available evidence remains of low quality, but suggests that biological and behavioural factors likely underpin the associations of sleep characteristics with adiposity. Inadequate sleep may stimulate hormonemediated preferences for nutritionally poor calorific foodstuffs, which could be compounded by poor sleep enabling more time to eat, shifting meal times to later in the day, creating more opportunities for screen-based sedentary time, and reducing daily physical activity due to fatigue (Duraccio et al., 2019;Krietsch et al., 2019).
Providing additional evidence to support one or more of the hypothesised pathways could help to ameliorate the consequences of inadequate sleep, by identifying modifiable intermediate behaviours that can be targeted for mitigation. A research focus on adolescence is important, it is a key transitional stage of growth and maturation, characterised by marked behavioural changes in sleep, diet, sedentary behaviour and physical activity. It is a critical period for the development of persistent obesity (Reinehr, 2018).
This study examined the associations of multiple sleep dimensions, including sleep timing, duration and quality (sleep latency and night waking frequency), with BMI z-score, %BF, and overweight and obesity in a large and nationally representative sample of UK adolescents. Sleep dimensions were mutually adjusted for one another, and effect modification by sex was examined. To shed light on pathways of action, associations were scrutinised when adjusted for potential mediating factors, including diet, screen time and habitual physical activity.

| Study population
The Millennium Cohort Study (MCS) is a nationally representative birth cohort study of children who were born in the UK between 2000-2002. A total of 18,818 children were recruited (Connelly & Platt, 2014). A sixth sweep of data collection was conducted in 2015-2016, when children were followed-up in adolescence, and for the first time provided a detailed description of their sleeping habits (Joshi & Fitzsimons, 2016). This is a complete case analysis that included only the first child surveyed in any household (thereby excluding a small number of twin and triplet siblings [n = 121]), aged 13-15 years, who provided full data for sleep, adiposity and potentially important covariates. Ethical approval for all MCS surveys was obtained via the National Health Service Research Ethics Committee system, and participants provided consent to participation.

| Adiposity and weight status
Trained researchers followed standard protocols to measure weight (Tanita BF-522W) and height (Leicester height measure). BMI was calculated and converted to z-scores and weight status categories (Cole et al., 1995); underweight and normal weight categories were combined. %BF was predicted from height, weight and demographic factors using a validated prediction model (Hudda et al., 2020). %BF was further estimated by bioelectrical impedance analysis (Tanita BF-522W), but as this method can be influenced by numerous environmental and physical factors, and because some participants (n = 135) opted out of the assessment, bioimpedance data were used only in a sensitivity analysis (Dehghan & Merchant, 2008).

| Covariables
Potential covariables were selected based on evidence linking them with sleep (Bartel et al., 2015) and observations in the MCS cohort that they predicted adolescent weight status (Pongiglione & Fitzsimons, 2019). Age was calculated as the elapsed time between the child's date of birth and the date of the sixth sweep assessment; which was also used to infer season of measurement. Adolescents self-reported their ethnic group and parents reported their incomes, which were equivalised to indicate net disposable household income per week (Mireku & Rodriguez, 2020

]). Adolescents completed a 13-item Mood and
Feelings Questionnaire from which a summed score ranging from 0 to 26 was created; higher scores indicate more severe depressive symptoms (Messer et al., 1995).

| Statistics
Linear regression was used to investigate associations of sleep characteristics with adiposity and logistic regression-quantified as-

| Descriptive characteristics
Characteristics of study participants (n = 10,619; 50.3% boys) are shown in Table S1, and weighted sample characteristics are provided in

| Associations with BMI z-score and %BF
There was consistent evidence for effect modification by sex, hence all analyses were stratified by boys and girls (p ≤ 0.0038). Results from initial models are provided in Tables S11-S14. Initial and model 1 results were comparable, indicating that physical activity, TV viewing and diet did not confound or mediate associations. Patterns of association were consistent regardless of whether BMI z-score or %BF were modelled as outcomes. Table 1 shows that sleep onset times after 10pm were associated with higher adiposity; associations were dose-dependent in boys but for girls plateaued after 11pm. Sleeping fewer than > 9-10 hr was dose-dependently associated with higher adiposity in boys and girls, though adjustment for sleep latency and night waking attenuated associations somewhat in girls (Table 2).
There was some evidence in both sexes that, compared with a sleep latency of 16-30 min, a shorter latency time of 0-15 min was associated with higher adiposity; sleep latencies ≥ 46 min were associated with higher adiposity in girls (Table 3). Relative to never waking in the night and struggling to get back to sleep, sometimes waking was associated with higher adiposity in boys, and often and habitually waking were associated with higher adiposity in girls (Table 4). All results were substantively unchanged when, instead of TV viewing, models were adjusted for social media use or electronic gaming (Tables S15-S18), and when %BF estimated by bioimpedance was modelled as the outcome (Tables S19-S22). Adjustment for depressive symptoms tended to partially attenuate the associations of later sleep onset, shorter sleep duration, longer sleep latency, and more frequent night waking with higher adiposity in girls (Tables S23-S26). Model 2 results for sleep latency and night waking frequency were consistent regardless of whether they were adjusted for sleep duration or sleep onset time.

| Associations with weight status
Odds ratios for overweight and obesity are shown in Figure 1.  All results were substantively unchanged when, instead of TV viewing, models were adjusted for social media use or electronic gaming (Tables S31-S34), and when additional adjustment for depressive symptoms was made (Tables S35-S38). Model 2 results for sleep latency and night waking frequency were consistent regardless of whether they were adjusted for sleep duration or sleep onset time.

| DISCUSS ION
This investigation of > 10,000 UK adolescents has replicated the consistent cross-sectional finding that short sleep duration is associated with higher adiposity in youth. Uniquely, there was also evidence that sleeping longer than recommended was associated with higher odds of overweight and obesity in adolescent girls. In addition, later sleep timing was associated with higher adiposity in both sexes, and sleep quality markers were consistently associated with higher adiposity and higher likelihood of overweight and obesity in girls. The results show that sleep timing, duration, and quality exhibit independent associations with adolescent adiposity levels, and highlight potentially important differences in relationships between boys and girls. The associations were independent of covariates, including candidate mediators such as physical activity, screen time and diet, but there was some indication that depressive symptoms may have exerted a partial confounding or mediating effect in girls.
There is considerable evidence for associations between short sleep duration with higher expressions of BMI and obesity in youth.
There is also evidence for effect modification by sex, with associations appearing to be stronger in boys than girls (Matricciani et al., 2019;Sun et al., 2020). This may relate to the physiological changes that occur during puberty, or girls may have a lower sleep need than boys and may be more resilient to sleep debt. This study also found that associations between shorter sleep duration with higher adiposity were stronger in boys than girls. Furthermore, whilst associations for boys were consistent regardless of the level of statistical adjustment, magnitudes of associations were somewhat attenuated in girls when adjusted for sleep latency and night waking frequency.
In finally adjusted models, ≤ 8 hr of sleep was independently associated with 80% and 38% higher odds of overweight and obesity in boys and girls, respectively. These estimates are consistent with the summary odds and risk ratios produced by various meta-analyses to date, which have provided estimates ranging from 30% (1.11-1.53) to more than twofold (2.23; 2.18-2.27) higher likelihood of overweight TA B L E 2 Associations of sleep duration with adiposity Abbreviations: %BF, percent body fat; BMI, body mass index. and obesity as a function of short sleep (Sun et al., 2020). Rather than U-shaped associations, which are prevalent in adults, studies of youth report that sleep duration is linearly associated with adiposity.
It is important to consider, however, that tests of non-linearity are often limited in studies of children and adolescents by low variability in sleep and sparse data at the upper end of the sleep length distribution (Chaput et al., 2016). This large national study provides novel evidence of a U-shaped association in adolescent girls, for whom longer self-reported sleep (> 10 hr) was independently associated with 31% higher likelihood of overweight and obesity. This could be perceived to support the hypothesis that girls have a lower sleep need than boys, which if exceeded is detrimental. It might be that associations are also U-shaped in boys, but few exceed the higher threshold above which sleep is seemingly deleterious (Marshall et al., 2008). The results appear to show that associations between sleep length and adiposity begin to resemble adult-like patterns in girls before boys. Depression and underlying illness are believed to contribute to the adverse health outcomes associated with long sleep in adults (Yin et al., 2017), and displacement of health-enhancing behaviours such as physical activity may also be involved. Of note, the association reported here between longer self-reported sleep with higher likelihood of overweight and obesity in girls, was independent of lifestyle behaviours and depressive symptoms, but residual confounding or mediating effects cannot be ruled out due to the selfreported nature of the data. Future studies should aim to generate more precise exposure-response curves to interrogate the shape of associations between sleep duration with adiposity and weight status in adolescents, including the potential effects of sleeping more than recommended.
A small number of studies have reported that later bed and sleep onset times are independently associated with higher BMI z-score (Jarrin et al., 2013;Olds et al., 2011), %BF (Jarrin et al., 2013), overweight and obesity (Olds et al., 2011;Skjåkødegård et al., 2021), and obesity severity in adolescents (Hayes et al., 2018). This is the first study to report sex-specific associations in a large and nationally representative sample of adolescents. Later sleep onset was dose-dependently associated with higher adiposity in boys, for whom going to sleep after midnight predicted 76% higher odds of overweight and obesity. Magnitudes of association were somewhat attenuated in girls when adjusted for sleep latency and night waking frequency, but later sleep onset predicted higher adiposity, and going to sleep between 11 and 11:59pm predicted 36% higher likelihood of overweight and obesity. The best available evidence remains of low quality, but suggests that short sleep, later onset and poor sleep quality may elevate obesity risk via shared pathways, including: (a) increased neural activation in regions of the brain that are associated with food desirability, which drives a preference for nutritionally poor calorific foodstuffs, and is amplified by increased opportunity to eat by being awake longer; (b) decreased insulin sensitivity; (c) disrupted meal timing, for instance skipping breakfast then overcompensating by eating more energy-dense foods later in the day; and (d) increased sedentary behaviour due to being awake longer (particularly more screen time late in the evenings, which is associated with increased snacking behaviour) and tiredness dissuading physical activity (Duraccio et al., 2019;Krietsch et al., 2019).

TA B L E 3 Associations of sleep latency with adiposity
The results of this study were, nevertheless, independent of physical activity, myriad screen-based behaviours (including TV viewing, electronic gaming and social media use) and a composite healthy diet index. It is a weakness that all covariables were self-reported; random error and biased responses mean that residual confounding or mediation of the reported associations is likely. It is also unfortunate that only total screen time data were available, there was no time-specific information to differentiate between screen viewing in the evenings compared with other times of day. Furthermore, it is a limitation that information was not available for potentially key mediating variables, such as unhealthy snacking and meal times. Sleep restriction studies report increased snacking under short sleep conditions, particularly when participants would normally be asleep (Beebe et al., 2013;Nedeltcheva et al., 2009), and an investigation of young children found that an association of later sleep onset with higher adiposity was attenuated when adjusted for timing of the evening meal (Collings et al., 2021). Animal models and clinical trials of adult populations indicate that early and time-restricted feeding has numerous physiological benefits, including lower adiposity (de Cabo & Mattson, 2019;Sutton et al., 2018). Experimental studies of sleep manipulation, and well-designed tests of mediation in observational studies that have measured variables more precisely, are needed to better shed light on the conceivable pathways of action which might link sleep characteristics with obesity in youth.
Sleep quality is a heterogeneous construct, and as such there is considerable variability in how it is measured and defined (Ohayon et al., 2017). In this investigation, sleep latency and night waking frequency, two recognised indicators of sleep quality, were investi- which found that poor sleep quality (broadly defined as longer sleep latency, more sleep disturbances, recurrent night waking or lower sleep efficiency) predicted 46% (1.24-1.72) higher odds of overweight and obesity (Fatima et al., 2016). In the current study, there was also evidence that a short sleep latency of 0-15 min, which could be symptomatic of sleep deprivation, was associated with higher adiposity in both sexes. In addition, sometimes waking in the night was associated with higher adiposity in boys, and often (and to a lesser extent habitually) waking predicted higher adiposity in girls.
The results suggest that poor sleep quality is associated with higher adiposity and elevated obesity risk, particularly in adolescent girls, a higher proportion of whom reported longer sleep latencies and more frequent night waking than boys. This may be explained by girls drinking more hot caffeinated drinks in the evening (Galland et al., 2017), using their mobile phones more throughout the night (Godsell & White, 2019), and girls experiencing more interpersonal stress and more symptoms of anxiety and depression than boys (Hankin et al., 2007). Of note, there was some indication that the associations of later sleep onset, shorter sleep duration, longer sleep latency and more frequent night waking with higher adiposity in girls, were all partially attenuated by depressive symptoms. Depressive symptoms might be a prominent mediating or confounding factor of the sleepobesity nexus in adolescent girls.
Sleep recommendations for youth have historically focussed on duration as opposed to any other dimension of sleep (Matricciani et al., 2012). The results of this study highlight that contemporary sleep guidelines should further acknowledge the importance of sleep timing and quality. Based on the associations presented here, the optimal sleeping pattern for 13-15 year olds in terms of body composition appears to be a sleep onset time before 10pm, taking no longer than 45 min to fall asleep, and sleeping for > 9-10 hr with few-to-no periods of overnight waking. Satisfying each of these conditions amidst a prevailing backdrop of short sleep may be challenging. Nonetheless, it is reassuring that associations for sleep components were independent of one another, meaning that improving just one aspect of sleep could translate to meaningful benefit. This is important given that school-based sleep education programmes are generally ineffectual in changing sleep duration, but sleep quality improvements are achievable (Illingworth et al., 2020). Tailoring of sleep interventions to the specific needs of boys and girls may be warranted.

| Strengths and limitations
This investigation benefitted from contemporary data collected in a nationally representative and large cohort, which enabled associations to be quantified within sex strata. In addition to a weightfor-height proxy, %BF was investigated as a direct measure of adiposity, and each of the analyses was adjusted for a broad range F I G U R E 1 Odds ratios for overweight and obesity by sleep onset time (top left), sleep duration (top right), sleep latency (bottom left), and night waking frequency (bottom right). Results are adjusted for age, ethnicity, net disposal household income, birth weight, breast-feeding history, season of measurement, physical activity, TV viewing and diet index. Results for sleep onset time and sleep duration are further adjusted for sleep latency and night waking frequency; results for sleep latency are further adjusted for sleep duration and night waking frequency; results for night waking frequency are further adjusted for sleep duration and sleep latency of covariables. This included mutual adjustment of sleep parameters to tease apart the independent associations of distinct sleep dimensions. That said, sleep onset time and duration were not mutually adjusted as they were closely related; at the fringes of cross-tabulations there were small and empty cells (Table S5), and including both in the same statistical model produced signs of multicollinearity including reversals of direction of coefficients. As with any observational study, residual confounding or mediating effects by imperfectly measured covariables is possible, as is confounding by unspecified factors. This study did not control for maturity status as most participants would have been in the advanced stages of puberty (> 90% of boys and girls reported their voice had started getting deeper and that they had experienced menarche, respectively), and there were missing data. Sleep characteristics were self-reported, which is a limitation, but adolescents tend to have a set routine on school nights, and self-reports enabled sleep quality in the last 4 weeks to be documented. Objective sensor-based characterisation typically covers a much narrower timeframe and has low specificity for identifying periods of night waking (Meltzer et al., 2012). Owing to the cross-sectional nature of the analysis, it is not possible to establish causality of the observed associations.

| CON CLUS IONS
Sleep timing and duration in both sexes, and sleep quality in girls, appear to be independently associated with adiposity and weight status in adolescents. In addition to duration, sleep timing and quality components may be important modifiable determinants of youth overweight and obesity.

ACK N OWLED G EM ENTS
The author thanks the Millennium Cohort Study families for their time and cooperation, and is grateful to the Centre for Longitudinal Studies, University College London Institute of Education, for the use of the Millennium Cohort Study data, and to the UK Data Archive and UK Data Service for making them available. However, they bear