Annual, seasonal, cultural and vacation patterns in sleep, sedentary behaviour and physical activity: a systematic review and meta-analysis

This review follows PRISMA guidelines [20] and was registered on PROSPERO (ID CRD42020184012) [21].

Studies were included if: they were cohort studies reporting objectively or subjectively measured minutes of sleep, sedentary behaviour, and/or physical activity (light and moderate-to-vigorous); participants were healthy adults aged 18 to 65 years (reflecting the typical age range of adults in health behaviour guidelines); the study explored a temporal pattern (e.g. seasonal effects, pre/post/during vacation or cultural festivals); and the study included at least two assessment time points within a 12-month period using a within-subject design (the most robust design for addressing our research question). Intervention and cross-sectional study designs were excluded along with studies exploring non-general population samples and studies conducted in controlled settings. Studies comparing weekdays to weekends were excluded if only one time point was measured, however longitudinal studies using this comparison were included and weekly means calculated. Studies with populations that spread outside the eligible age range were included if data were able to be extracted (or obtained from authors) for only those participants aged 18–65-years.

The search strategy was developed in consultation with an academic librarian. Six databases were searched in 2020; Embase, Medline, Ovid Emcare, Scopus, Sport Discus, and Web of Science. Temporal patterns included terms relating to seasons, holidays, vacations, and festivals. Specific terms relating to religious events were included for the three most commonly observed religions: Christianity, Islam and Hinduism [22]. Where possible, searches were limited to English language, humans, and adults 18–65 years of age. Athlete* or player* were excluded terms and no limitation was placed on date range. Figure 1 shows the search strategy used for Medline.

All studies were imported into EndNote (EndNote × 9, Clarivate, Philadelphia, USA) for duplicate removal. Remaining studies were exported to Covidence (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia) for screening. Two independent reviewers completed title and abstract and full-text screening, with a third reviewer consulted for disagreements.

Where necessary, authors of potentially eligible studies were emailed to request additional data. Reference lists of included studies were searched to identify additional studies.

Two independent reviewers used a custom data extraction template, including: study design, exposure (e.g. season, vacation, cultural festival), participant characteristics (sample size, setting, age, sex), outcomes measured (sleep, sedentary behaviour and physical activity), tool used (e.g. accelerometry, questionnaire), assessment time points (duration and frequency), and results (means and standard deviations for daily minutes of each outcome).

Each study was appraised by two independent reviewers using the Joanna Briggs Institute (JBI) critical appraisal checklist for cohort studies [23]. Conflicts were resolved by a third reviewer. Three items (1, 2 and 6) were irrelevant to single-group study designs and were removed. Each item was rated “yes”, “no”, “unclear”, or “not applicable”. Studies were not excluded based on risk of bias results.

The primary outcome measure was mean daily minutes spent in sleep, sedentary behaviour, light physical activity and moderate-to-vigorous physical activity. Where moderate and vigorous physical activity were presented separately, the values were added. Where required, data was converted to be expressed as mean daily minutes. This involved combining groups where appropriate (e.g. combining males and females where reported separately), converting monthly data to seasonal data, and converting weekly data to daily data. Standard errors were converted to standard deviations. Where more than 10 studies are identified for an outcome, funnel plots will be used to examine publication bias [24].

Meta-analysis was conducted for outcome and exposures reported in multiple studies. For studies presenting seasonal data, an overall yearly mean (minutes per day) for each outcome was calculated to allow for comparisons with seasonal means. For studies exploring Ramadan, during-Ramadan data were compared with pre/post-Ramadan data. Where multiple data points were reported during or pre/post Ramadan, overall means were calculated by combining relevant data for each condition. A random-effects model was used to calculate standardised mean difference (SMD), pooled SMD, and associated confidence intervals (CI), along with forest plots generated using Review Manager (version 5.4, The Cochrane Collaboration). For seasonal data, estimates of the weighted mean difference (WMD) were calculated as the season mean minus the yearly mean expressed as minutes per day. Each calculation was performed by multiplying the overall yearly standard deviation by the SMD for the season of interest. Estimates of the WMD for during Ramadan compared with outside Ramadan was calculated by multiplying the overall standard deviation for outside-Ramadan by the SMD for during-Ramadan and expressed as minutes per day. Heterogeneity was assessed using the i² statistics, calculated using Review Manager. Analyses were performed in 2020.

The search identified 7009 results, including 4038 different studies, of which 3967 were deemed ineligible based on titles and abstracts. Of the 71 studies reviewed in full text, a further 54 studies were ineligible for several reasons, including 25 had an ineligible study design (e.g. experimental study or cross-sectional), 21 did not report an outcome of interest (minutes of activity), seven had an ineligible population (e.g. children, over 65 yrs., disease populations) and one full-text version of the study was unavailable after extensive searching. Seventeen studies, published between 2003 and 2019, were included (see Fig. 2). Two of the included studies became eligible as a result of contact with authors who provided additional data of interest [25, 26].

Table 1 presents a summary of the included studies. Of the 17 included studies, four temporal exposures were present: seasonal patterns (n = 11) [26, 32‐41], Ramadan (n = 4) [28‐31] vacation (n = 1) [25], and daylight savings time transitions (n = 1) [27]. Studies were conducted across four continents: Europe, Africa, Asia, and North America. Most recruited participants from a single country, however, three studies included two or more countries [26, 38, 39]. Multiple studies were conducted in the USA (n = 4) [25, 32, 33, 40], Saudi Arabia (n = 3) [28‐30], Sweden (n = 2) [37, 38] and England (n = 2) [34, 36], with single studies from Ghana [39], Norway [39], Denmark [38], Germany [35], Czech Republic [26], Slovakia [26], Poland [26], Japan [41], Italy [27], and Libya [31].

The studies involved a total of 1951 participants, with sample sizes ranging from 8 [30] to 960 [41], 14 had sample sizes less than n = 100. Mean participant age ranged from 18.1 years (SD = 0.3) [40] to 61.2 years (SD = 2.4) [26]. Most studies (n = 12) included mixed sex cohorts (range = 39 to 82% female), two had only female participants [32, 33], and three had only males [27, 30, 31].

Sleep was the most common activity examined (n = 13) [27‐31, 34‐41], with sedentary behaviour and light physical activity both measured in four studies [25, 26, 32, 34], and moderate physical activity and vigorous physical activity measured in five studies, either separately or combined as moderate-to-vigorous physical activity [25, 26, 32‐34]. Ten studies used subjectively measured activity (i.e. survey, questionnaire, sleep diary) [25, 28, 29, 31, 33, 36, 37, 39‐41], four studies used objective measures (accelerometry) [26, 27, 30, 32], and three studies used a combination (accelerometer and sleep diary) [34, 35, 38]. One study used a 24-h accelerometry protocol, allowing changes in sleep, sedentary behaviour, light physical activity and moderate-to-vigorous physical activity to be examined simultaneously [34]. Analyses were not performed to detect potential publication bias due insufficient studies for each outcome and subgroup analyses were not performed due to the limited data available.

Results from the JBI critical appraisal are presented in Table 2. Three studies received ‘yes’ responses for all items [26, 32, 34]. All studies measured the exposure in a valid and reliable way. Almost half of the studies did not identify confounding factors [25, 28‐30, 33, 35, 37, 39], nine studies did not state strategies that were used to deal with confounding factors [25, 28‐30, 33, 35, 37, 39, 40], whilst it was unclear for one study if strategies were used [31]. It was unclear whether Suzuki et al. [41] and Waterhouse et al. [31] measured outcomes in a valid and reliable way. More than half of the studies had incomplete follow-up [25, 28, 33, 36‐41] whilst completeness of follow-up was unclear for one study [31]. Where follow-up was incomplete or unclear, most reported strategies used to address this, however it was unclear for the studies by Garde et al. [38] and Waterhouse et al. [31]. All studies were deemed to have used appropriate statistical analyses.

This systematic review has brought together current evidence regarding annual rhythms in adults’ daily movement behaviours and examined key temporal factors which may influence these rhythms. Results reveal only a handful of studies identified per temporal factor, mostly involving small sample sizes. Regardless, several trends emerged. Sleep appeared to be higher during autumn and winter, outside of Ramadan and after the transition into daylight savings time. Sedentary behaviour also appeared higher during autumn and winter, and during non-vacation periods. Light physical activity was higher during spring and summer and whilst on vacation, and moderate-to-vigorous physical activity appeared to be higher during spring and summer, and during non-vacation periods.

Seasonal effects on movement behaviours (particularly sleep) was the most commonly explored temporal factor. The meta-analyses were generally non-significant (except for light physical activity), however methodological factors may have contributed to this. Many studies used self-reported tools, which are known to have modest validity [42] and may in turn have limited the ability to accurately detect change over time. Secondly, study populations were small (especially for the sedentary behaviour, light physical activity and MVPA meta-analyses). However, the direction for change in movement behaviours across seasons were generally consistent across studies, and varied in the expected directions, suggesting that the non-significant trends may be real.

Results suggested sleep duration was approximately half-an-hour less in summer compared with autumn. This may influence health outcomes, particularly if sleep totals fall outside the optimal range for positive health outcomes of 7-8 h [43]. For example a one-hour decrease in sleep below 7 h is associated with a 6% increase in all-cause mortality, an 11% increase in cardiovascular disease risk [43], and a 9% increase in obesity risk [44]. Results suggested that sedentary behaviour is around 30 min higher in winter, relative to spring. It is questionable whether this would impact health. However, it was coupled with lower light physical activity (45 min less, relative to summer) and less moderate-to-vigorous physical activity (5 min less, relative to summer), which is likely to be of sufficient magnitude to have health consequences. For example, a 1 min per day increase in moderate-to-vigorous physical activity has been associated with a 2% decrease in all-cause mortality [45]. It has been suggested that as little as 15 min extra light physical activity, or 5 min extra moderate-to-vigorous physical activity, has a positive effect on all-cause mortality and cardiovascular disease [46].

It is important to consider that season interacts with climate and daylight hours differently in different geographic regions. Almost all seasonal studies included in this review took place in temperate climate locales in the mid-latitudes (between 23^o and 66^o) of the northern hemisphere. Temperate climates are characterised by mild-to-warm summers and cool-to-cold winters with moderate yearly rainfall [47]. Mid-latitudes have distinct differences in daylight hours between seasons of approximately 4-8 h between winter and summer. Only a single study sample originated in a tropical climate and latitude (Ghana) [39]. Given the small number of studies in this review, it is not possible to disentangle the effects of season, climate and daylight hours. Temperate climates dominate the evidence presented here, and further research in tropical and subpolar regions is warranted.

Whilst this review sought studies examining various cultural, religious and festive periods’, all eligible studies of this type related to Ramadan only, and its impacts on sleep. Ramadan is an example of a periodic change to typical lifestyle patterns. Specifically, Ramadan is a spiritual disciplinary practice, where people observing Ramadan do not eat or drink during daylight hours, changing usual mealtimes, along with usual bedtimes and wake times, and therefore changes in movement behaviours could be expected. Additionally, during this month, people are reportedly less active during daylight hours [48] however none of the included studies examined physical activity during Ramadan. Across studies, participants observing Ramadan generally went to bed and woke later, with the delay in bedtime more pronounced than wake-time. Whilst there was a trend towards less sleep during Ramadan, the weighted mean sleep duration of 7 h and 25 min falls within recommended levels of 7 or more hours per night [13]. It should be noted that Ramadan is the ninth month of the Islamic calendar, which is 10–12 days shorter than the Gregorian calendar, therefore over several years, Ramadan will occur across all seasons. It is possible the trends observed in sleep time across seasons may interact with Ramadan differently depending on the season, with shorter winter days allowing for an extended window to consume meals, impacting less on usual sleep times. Of the studies exploring Ramadan, three were conducted in Saudi Arabia, and one in Libya. Both countries are located at a similar latitude (approximately 16-33^o), experiencing similar daylight hours across the year. Across the four included studies, Ramadan occurred between late Summer and mid-winter. While changes to sleep routine during Ramadan may be unaffected by seasonal changes at latitudes with small fluctuations in daylight hours across the year, future research in different locations is warranted.

The single study exploring daylight savings transitions reported a significant change in sleep duration of approximately half-an-hour after each transition, with less in autumn and more in spring. It is worth noting duration at all time points fell below recommended levels of sleep per night of 7 or more hours [13]. The immediate one-hour shift in clock time requires an adjustment period for our internal biological clocks to reset to the new timing of daylight hours, accounting for the immediate disruptions to sleep duration [49]. It is likely the observed changes in sleep duration return to pre-transition levels in following weeks, however further research for longer periods is needed to examine this. Further studies are needed before conclusions can be made about the impact of daylight savings transitions.

The final periodic temporal factor explored was vacation periods where again, only a single study was found. This study suggested participants on vacation got less moderate-to-vigorous physical activity and more light physical activity and sedentary time. Further research is needed to allow for the impact of vacations on movement behaviours to be understood. Additionally, future studies should explore movement behaviour intentions for vacation periods as it may be a risk factor for poor movement behaviour during vacation. The location or time of year when participants went on vacation was not reported, and thus it is important to consider that season may be a confounding factor.

This review identified trends towards negative changes in daily movement behaviours during several periods. Results suggest physical activity interventions may be needed during winter and autumn months where activity levels decline. Efforts could focus on alternate exercise options for these periods, such as indoor or home-based exercise routines, along with the provision of indoor or sheltered environments within the community to combat the environmental barriers during these colder, wetter and darker periods. Interventions promoting regular movement during winter and autumn could also target sedentary behaviour. Periodic temporal factors such as observing religious practices have potential to provide spiritual, physical and/or mental health benefits which would suggest these periods may not be the optimal time for movement behaviour intervention. Further research is needed to understand whether there are significant and lasting effects on usual movement behaviours resulting from periodic temporal factors. Opportunities may exist to promote pre-emptive behaviour changes, to counteract the effects of these periodic temporal factors if those effects are found to be impactful on health. For example, increased physical activity could be encouraged in the lead up to vacation or moving bedtime earlier before the autumn daylight savings transition.

This is the first systematic review and meta-analysis of its kind. Several strengths to this systematic review include:, the search strategy was rigorous and designed in consultation with an academic librarian, and followed PRISMA guidelines [20]; study screening, data extraction and risk of bias were all performed in duplicate, increasing accuracy and trustworthiness of results.

A key limitation of the review was the relatively small number of studies exploring each of the four included temporal factors, ranging from one study each for daylight savings time and vacation, to 11 studies for season. Most studies had small sample sizes (n = 60 or less in 14 studies). The sleep meta-analyses were dominated by a single study with a comparatively large sample (n = 960). Only one of the included studies reported the 24-h day, with a large portion of studies measuring sleep only. Limiting to English language studies may have reduced the variety of temporal patterns explored, specifically cultural or religious events in non-English speaking countries. Only a limited number of countries were included, each with their own seasonal and daylight hour patterns, therefore findings may not be representative of countries not explored in this review.

This systematic review highlights the limited number of cohort studies exploring annual patterns of activity. Future research with larger sample sizes, in more diverse geographic areas, and using high quality movement behaviour measurement methods, are needed. Very little research has examined changes in daily movement behaviours during festive and cultural periods. Additionally, no studies have explored the interaction between temporal factors. Changes in a particular behaviour must be necessarily compensated by a change in another behaviour, given that there are 24 h in a day. Methodologies which measure multiple movement behaviours simultaneously, ideally using a 24-h measurement approach, are needed.