CLINICAL REVIEW Associations of screen time, sedentary time and physical activity with sleep in under 5s: A systematic review and meta-analysis

Sleep is crucial to children's health and development. Reduced physical activity and increased screen time adversely impact older children's sleep, but little is known about these associations in children under 5 y. This systematic review examined the association between screen time/movement behaviors (sedentary behavior, physical activity) and sleep outcomes in infants (0 e 1 y); toddlers (1 e 2 y); and preschoolers (3 e 4 y). Evidence was selected according to Preferred Reporting Items for Systematic Reviews and Meta- Analyses guidelines and synthesized using vote counting based on the direction of association. Quality assessment and a Grading of Recommendations, Assessment, Development and Evaluationwas performed, strati ﬁ ed according to child age, exposure and outcome measure. Thirty-one papers were included. Results indicate that screen time is associated with poorer sleep outcomes in infants, toddlers and preschoolers. Meta-analysis con ﬁ rmed these unfavorable associations in infants and toddlers but not preschoolers. For movement behaviors results were mixed, though physical activity and outdoor play in particular were favorably associated with most sleep outcomes in toddlers and preschoolers. Overall, quality of evidence was very low, with strongest evidence for daily/evening screen time use in toddlers and preschoolers. Although high-quality experimental evidence is required, our ﬁ ndings should prompt parents, clinicians and educators to encourage sleep-promoting behaviors (e.g., less evening screen time) in the under 5s. © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Adequate sleep plays a critical role in children's health and development, particularly in the early years. Short sleep duration in preschool children is linked to obesity in later childhood [1]. Furthermore, sleep problems beyond age two are associated with reduced grey matter volume at seven years, indicating a role of sleep in early brain development [2].
International guidelines recommend that infants (0e1 y) sleep for up to 17 h/d, while toddlers (1e3 y) and preschoolers (3e5 y) should sleep between 10 and 14 h/d [3]. However, today's children sleep less than they did a century ago [4] and 20e30% of parents report that their child has difficulties falling or staying asleep [5,6]. The causes for this apparent epidemic of sleep problems are likely multi-factorial but lifestyle changes in an increasingly digitized world are a cause for concern [7].
Australia, Canada, South Africa, New Zealand and WHO have issued 24-h movement guidelines for under 5s, recommending an 'optimal day' in terms of children's sleep, physical activity and sedentary behaviors (including screen time) [8e11]. This 'whole day matters' approach places each behavior along a continuum, where declines in one behavior results in an increase in another. Studies in older children and adults have shown that daytime physical activity and screen time both influence sleep [12e14], but less is known about these relationships in children under 5 y of age. The early years are also a critical period in life for establishing healthy behaviors as screen time and physical activity appear to track from early into later childhood and adolescence and consequently may influence sleep later in life [15].
No reviews to date have synthesized and evaluated the quality of international research evidence in the under 5s. This review therefore sought to determine how screen time, sedentary time and physical activity are associated with eight sleep outcomes (i.e., total sleep duration; night awakenings; sleep onset latency; bed time; daytime napping; sleep efficiency; sleep stability; and sleep quality) in children aged 0e4 y.

Data sources and search strategy
This systematic review was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [16]. A systematic literature search was undertaken in April 2018 and updated in March 2019, using search terms related to: population; study design; outcome; exposure; and exclusion of clinical populations (Appendix 1). The search was conducted in 17 electronic databases: EBSCO (CINAHL); Cochrane Library (CENTRAL); OVID (EMBASE, MEDLINE, PsycINFO) and Web of Science (all databases). Citations were downloaded into Endnote citation management software (Thomson Reuters, Philadelphia, PA, USA) and de-duplicated. Included papers were searched for additional relevant publications, as were relevant reviews. No language or publication date restrictions were placed on the search.

Study selection
Studies were included if they: 1) reported results from a cross-sectional, longitudinal or experimental study and 2) assessed the relationship between screen time (total daily screen time; evening screen time) or any movement behavior (i.e., sedentary time; total, light, moderate-to-vigorous physical activity; floor-based play (infants); outdoor play/time; sports participation) and any sleep outcomes reported. Studies assessed healthy children (i.e., general populations, including those with overweight/obesity) aged birth to 59 mo at baseline; objective or subjective measures of exposures and outcomes were considered. Exclusion criteria included: 1) clinical populations (e.g., children with chronic health conditions e.g., asthma, or developmental disorders e.g., cerebral palsy, autism) 2) qualitative studies; 3) studies assessing screen-based content; and 4) those assessing electromagnetic radiation.

Study screening, data extraction and quality assessment
Identified titles and abstracts were screened for relevance (KH, GK) and included titles were separated by exposure type (sedentary time, physical activity or screen time; KH). Full texts of identified articles were retrieved and read in full to assess eligibility for inclusion (physical activity: KH, RK; sedentary behaviors: XJ, AM). Reviewers independently extracted and crosschecked relevant data using a pre-piloted data extraction form (physical activity: AM, KH; sedentary behaviors: AH, CH, XJ). Data were extracted per age group; infants (0 to <1 y), toddlers (1 to <3 y), preschoolers (3 to <5 y) and for each exposure-outcome association. The split between age groups was chosen for two reasons. First, major developmental differences exist during the early years in both physical and cognitive development. Therefore, we hypothesize that the investigated associations may be different for each of these age groups. Second, the chosen split in age groups is consistent with the international 24-h movement guidelines.
Investigated exposures were: 1) daily and evening screen time including parent report of child time spent on TV, tablet, phone, playing computer games, using the internet; 2) accelerometry measured physical activity including total sedentary behavior, light physical activity and moderate-to-vigorous physical activity; 3) parent reported floor based play, organized sport and outdoor play. Total sedentary behavior and screen time were treated as two different exposures to provide more detailed evidence about whether screen time or all sedentary behaviors influence sleep.
For longitudinal studies, all time points up to age five were included. The latest time point included was before, or as soon after, the children were five years old (>5 y if no follow-up data on <5 y was provided). If two or more papers reported on the same study sample, both were treated as separate studies if they reported different exposure-outcome relationships (n ¼ 4) [17e20]. Several papers examined multiple exposure-outcome associations (e.g., how total screen time and TV time were associated with sleep) and reported findings for different groups (e.g., examined differences across age groups, by time of the week, or by sex). Each exposureoutcome was therefore counted as an individual association, e.g., a paper examining the association between screen time and total sleep duration, but reporting results for weekdays and weekend days separately, was counted as one study but two associations. For experimental studies, differences in outcomes between control and intervention groups over time were used to assess influence of exposures. Where possible, results from adjusted multivariable models were reported. Reviewers who extracted the data also assessed the methodological quality of primary studies and any discrepancies were resolved by consensus. Risk of bias assessment was completed as part of the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) of evidence quality. Six domains (selection, performance, detection, attrition, reporting, and other sources of bias) specific to study design (observational or experimental) were assessed. Each domain was determined to have a low, unclear or high risk of bias [21].

Data synthesis
Due to the heterogeneous nature of included studies, and the range of exposures and outcomes assessed, meta-analysis was only appropriate for one exposure-outcome association total screen time and sleep duration in infants, toddlers and preschoolers. Where available, correlation coefficients were recorded for each study. If studies did not report correlations coefficients but provided beta coefficients these were converted in to correlation coefficients using the method described by Peterson and Brown (2005). Only studies reporting cross-sectional associations were included in the main analysis. A sensitivity analysis including longitudinal outcomes was conducted but no significant differences were found between the two models. Data were pooled in a random-effect meta-analyses using Comprehensive Meta-analysis, version 3.3.07. Heterogeneity across studies was assessed using I 2 statistics (I 2 of 0e40% represents low heterogeneity and 75e100% considerable heterogeneity) [23].
For the remaining associations, as recommended by the Cochrane handbook for systematic reviews of interventions, vote counting based on the direction of association was conducted [24], comparing the number of favorable to unfavorable associations. Favorable associations were categorized as those where the exposure measure resulted in a positive association with sleep outcomes (e.g., less screen time associated with longer total sleep duration). Associations were unfavorable if the exposure measure resulted in a negative association with sleep (e.g., more screen time associated with shorter total sleep duration). Summary results per exposureoutcome association were presented as number of unfavorable (for screen time and sedentary behavior) and favorable (for physical activity, outdoor play and sport club attendance) associations divided by the total number of studies included. A binomial probability test was conducted. The p-value from this test indicates the probability of observing the summary results if the exposureoutcome associations were in the opposite direction, thus a small p-value indicates a higher probability the results are valid [24]. This method does not rely on p-values reported by the authors of the primary studies.
GRADE was performed on all findings, stratified according to child age (infants; toddlers; preschoolers), exposure, and outcome measure, with possible range from very low to high [25,26]. GRADE assigns an initial rating to each study design (i.e., high for randomized controlled trials, low for observational studies -both longitudinal and cross-sectional). This was then upgraded or downgraded according to the risk of bias, inconsistency, indirectness, imprecision, publication bias, doseeresponse relationship, residual confounding or the size of the magnitude.

Evening screen time
The relationship between evening screen time and sleep was examined in eight studies (infants n ¼ 3; toddlers n ¼ 4; preschoolaged children n ¼ 4; Table 3) [30,22,34,35,39,43,46,47]. In infants, higher levels of evening screen time were associated with shorter nighttime sleep duration (2/2) and later bedtime (1/1). Toddlers' and preschoolers' evening screen time was unfavorably associated with sleep in 8/9 associations. In toddlers, higher levels of evening  Intention-to-treat basis. The core model assumed a linear change of the outcomes with time and included two normally distributed random effects (one at the preschool level and one at the child level) to adjust for clustering in the data due to the hierarchic sampling scheme. Further, all models included the variables age, gender, rural versus urban community of preschools, and season as covariates to adjust for a potential confounding effect of these variables.  (continued on next page)  screen time were associated with shorter total sleep duration (2/2) and later bedtime (1/1). In preschoolers, higher levels of evening screen time were associated with shorter total sleep duration (3/4), later bedtime and lower sleep quality (1/1). Importantly, only one study reported a favorable association between evening screen time and any sleep outcome.

Total sedentary time
The association between total sedentary time and sleep was examined in five studies (infants n ¼ 2; toddlers n ¼ 0; preschoolers n ¼ 3; Table 4) [20,27,31,36,49]. In infants, higher levels of total sedentary time were associated with shorter sleep time duration (2/2), more night awakenings (1/1), less daytime napping (1/2) and better sleep efficiency (1/1). In preschoolers, higher levels of total sedentary time were associated with shorter sleep time duration (1/2) associations and later bedtime (1/1). More sedentary time was associated with fewer night awakenings (1/1). A decrease in sedentary time showed an association with improved sleep quality (1/1). No evidence was available for toddlers.
Seven studies assessed the relationship between physical activity intensity and sleep [19,20,31,33,36,45,52]. In one study, conducted in preschoolers, light physical activity was associated with later bedtime (1/1 association). In toddlers, higher levels of moderate-to-vigorous physical activity were associated with better sleep quality (1/2), and better sleep stability (1/1), and shorter total sleep duration (1/1). In preschoolers, higher levels of moderate-tovigorous physical activity were associated with shorter total sleep duration (1/2) and later bedtime (1/1), better sleep quality and better sleep stability (1/1). No evidence was available for infants.
The relationship between floor-based play was examined in one study for infants [49]: floor-based play was associated with shorter total sleep duration, less daytime napping (1/1) and better sleep efficiency (1/1). The relationship between time spent playing outdoors and sleep was examined in two studies [40,51]. Toddlers' outdoor play was associated with shorter total sleep duration (2/2), shorter sleep onset latency, fewer night awakenings and earlier bedtime (2/2). Preschoolers' outdoor play was associated with longer total sleep duration (1/2) and fewer night awakenings, earlier bedtime and shorter sleep onset latency (1/1). Preschoolers' attendance at sports clubs was associated with earlier bedtime and better sleep efficiency (i.e., higher fraction of total sleep spent asleep after sleep onset; 1/1) [44].

Quality of evidence
The quality of evidence ranged from very low to moderate for moderate-to-vigorous physical activity; and very low for all other exposure-outcome associations across age groups (Tables 2e4). Most studies were downgraded due to a serious risk of bias (commonly due to use of exposure or outcome measures with unknown psychometric properties; Table S1).

Discussion
To our knowledge, this is the first systematic review to explore how screen time and movement behaviors are associated with sleep in children under 5 y. This review highlighted a trend for an unfavorable association between higher levels of total daily and evening screen time and sleep outcomes in infants, toddlers and preschoolers; very few studies showed favorable screen-sleep associations. Meta-analysis conducted in a sub-sample of studies to examine the association between daily screen time and sleep duration confirmed these unfavorable associations in infants and toddlers. In preschoolers, the meta-analysis did not show a significant association, but this may be due to the heterogeneity of the included studies. Evidence for associations between total daily sedentary time/physical activity and sleep was less conclusive: there was an indication that more outdoor play and higher levels of moderate-to-vigorous physical activity were favorably associated with sleep outcomes in toddlers and preschoolers. Most evidence was from observational studies (both cross-sectional and *n ¼ number of associations showing unfavorable association, N ¼ total number of associations for the specific exposure-outcome relationship reported; # two-sided p-value from the binomial probability test. Small p-value indicates higher probability that the results are valid. a Quality of evidence was downgraded due to serious risk of bias. Quality rating of individual studies can be found in Table S1. b Indicates Longitudinal study. longitudinal), did not show significant associations, and did not report on doseeresponse relationships leading to evidence frequently being classified as low quality. In addition, no clear differences were found between studies including large (>500) or small (<500) sample sizes.
Established early in childhood [41], sleep patterns are governed by a complex interplay of physiological, genetic, psychological and social/environmental factors. A range of behaviors, including physical activity, sedentary and screen time, may delay or displace sleep -'5 more minutes please!' Moreover, socio-environmental influences such as parenting style, the home environment, and socioeconomic status are likely to influence young children's sleep, screen and activity behaviors [54e56].
In line with a recent systematic review in older children (5e20 y) [13], our review and meta-analysis highlights that screen time appears to be unfavorably associated with young children's sleep. Short wavelength (blue/green) light emitted from screens suppresses pineal melatonin secretion, influencing both circadian entrainment (via supra-chiasmatic nucleus signaling) and sleep onset (via the hypothalamic ventrolateral pre-optic nucleus) [57,58]. Although evidence is limited in very young children, differential diurnal rate of melatonin secretion appears to emerge early in development at around 27e41 d of age [59]. Theoretically therefore, evening screen exposure in very young children may not only delay sleep onset on exposure [60] but also potentially cause longer term disturbance to sleep stability [61]. A dim light environment prior to bedtime is likely to be conducive to melatonin secretion, simultaneously promoting earlier sleep onset whilst helping to establish and maintain an optimal circadian rhythm [62], with less night waking [63]. In addition to the light emitted from screens, the content, its interactivity, and subsequent level of arousal, may also adversely affect sleep.
Despite a wealth of evidence for a positive association between physical activity and sleep in older age groups [64], very few studies have examined the association between physical activity and total sedentary time on sleep in children 0e4 y. Our review identified evidence suggesting that more outdoor play and time spent engaged in moderate-to-vigorous physical activity may be associated with better sleep outcomes in toddlers and preschoolers. Although experimental research is largely lacking in children and young people, it has been noted in infants that natural light exposure, particularly during the afternoon, may improve nighttime sleep [65]. Such exposure, as part of children's outdoor play, may help to regulate melatonin secretion and circadian rhythm, encouraging regular sleep onset. Several other physiological mechanisms have also been proposed to explain how higher intensity physical activity may positively influence sleep (albeit in the context of adult sleep). These include: 1) activity triggering an increase in body temperature and subsequent cooling with rest to promote sleep onset, and 2) activity reducing negative arousal states which may otherwise lead to sleep problems [63]. Future experimental studies should determine why and how screen time and movement-behaviors impair and promote sleep respectively. This is particularly important given a number of 24-h movement behavior guidelines have recently been published worldwide, which outline an 'optimal day' for children's sleep, physical activity and sedentary behaviors (including screen time) [8e11]. Where the 'whole day matters' and each behavior is placed along a continuum, declines in one behavior may feasibly result in an increase in another.
This review highlights important gaps in the evidence base around screen-based and movement behaviors, and sleep outcomes in young children. The quality of evidence summarized in this review was low and in some instances inconclusive. The variation in results may be due to the wide range of exposure and outcome measures used across studies. Moreover, study quality tended to be downgraded due to use of measurement tools with untested psychometric properties, with 20 out of the 28 articles reporting exposures measured using an unpublished questionnaire/failing to report the questionnaire's psychometric properties (Table S1). Studies included in this review frequently focused on television-based screen time, and did not examine the use of more contemporary screens (e.g., tablets, phones) and/or the type of activities children engaged in while using screens (e.g., watching a movie or talking to grandparents on tablet/phone). With the advances in technology over the last decade, it is important that studies now consider the influence of alternative electronic media and screen-based activities (such as e-readers and tablets) on children's sleep. In addition, it is important studies examine the influence different media content may have on children's sleep (e.g., education v. recreational content) [66] While screen-based technology can positively support learning [67], neglecting its influence on sleep may paradoxically constrain neurodevelopment in the under 5s.
While objective measures of sleep duration (e.g., accelerometry) and valid and reliable sleep questionnaires are available, very few studies used either to assess sleep outcomes here (n ¼ 10). Accelerometry is known to poorly differentiate between prolonged sedentary behavior and sleep [68]; included studies using accelerometry all used different methods to estimate sleep and wake periods [18,19,31,36,45] which could have led to the discrepancy in results [69]. Standardized measurement and analysis procedures of exposure and outcomes would allow consistency and validity across studies. There was also a lack of experimental or intervention studies aiming to improve sleep practices in the early years. Last, as the majority of studies included in this review were cross-

Strengths and limitations
We applied rigorous review methods, including duplicate assessment at every stage. Given that this review was restricted to published studies, publication bias cannot however be ruled out. All included studies were conducted in high and middle-income countries. Almost half included small sample sizes (15 out of 31 studies had fewer than 500 participants), which may have limited their statistical power to detect significant associations. By using vote counting based on the direction of the effect, we limited the impact underpowered studies may have on the summarized results [24]. Nine exposures and nine outcome measures were used here, thus limiting the use of meta-analysis: where common exposureoutcome associations existed (i.e., for screen time and total sleep duration) meta-analysis was conducted. We defined the absence of daytime napping as an unfavorable outcome here as the majority of studies examining this outcome were in infants (n ¼ 3 out of 4). However, daytime napping has been linked to adverse sleep outcomes such as irregular sleep habits in preschoolers [48]. It is therefore difficult to interpret whether napping is an un/favorable behavior when assessed as an isolated outcome. Due to the codependence of movement and sleep behaviors, an increase in one behavior would be expected to result in a decline in another, e.g., if physical activity leads to an increase in sleep duration, sedentary behavior is more than likely to decrease. Future studies would benefit from assessing co-dependent behaviors across a 24-h period. Last, most studies included in this review controlled for common confounders (e.g., age, socio-economic status, sex) but few controlled for characteristics in the home and wider environment which may impact sleep (e.g., chaotic home life, shared bedrooms, noise). Future research should consider a wider range of relevant confounders in order to fully elucidate the relationship between screen time, movement behaviors and young children's sleep.

Conclusions
Screen time is unfavorably associated with multiple sleep outcomes in infants, toddlers and preschoolers. Conversely, in toddlers and preschoolers more time spent in outdoor play, and in higher intensity physical activity, was associated with better sleep outcomes. There is a pressing need for future research to establish how contemporary screen time (e.g., tablets and e-readers) influences the 24-h equipoise of activity and sleep in young children. Public health initiatives and policies are needed to help parents and educators encourage balanced use of screen-based technologies and positive movement behaviors to promote healthy lifestyles and development in the under 5s.

Conflicts of interest
The authors do not have any conflicts of interest to disclose. Drafting of the manuscript: Janssen and Hesketh. Critical revision of the manuscript for important intellectual content: All authors.
Final approval of the submitted article: All authors. AM was supported by the UK Medical Research Council (grant number MC_UU_12017/14) and the Scottish Government Chief Scientist Office (grant number SPHSU14). KRH is supported by the Wellcome Trust (107337/Z/15/Z). The work of KRH was supported, in part, by the Centre for Diet and Activity Research (CEDAR), a UKCRC Public Health Research Centre of Excellence (RES-590-28-0002). Funding from the British Heart Foundation, Department of Health, Economic and Social Research Council, Medical Research Council, and the Wellcome Trust, under the auspices of the UK Clinical Research Collaboration, is gratefully acknowledged. We thank Dr Ruth Kipping for her assistance during the initial screening phase. Dr Kipping did not receive any compensation for her work.