Variability in Sleep Timing and Dietary Intake: A Scoping Review of the Literature

The objective of this scoping review was to summarize previous studies which examined the effect of day-to-day variability in sleep timing and social jetlag (SJL) on dietary intake. A systematic literature search was conducted in PubMed, Embase, and Clarivate Analytics Web of Science and we identified 22 records. No difference in caloric and macronutrient intake between SJL groups was observed in studies that enrolled healthy young adults. However, studies that enrolled participants with obesity and obesity-related chronic conditions reported a higher caloric intake and a higher intake of carbohydrates, total fat, saturated fats, and cholesterol in participants with SJL than in those without. Most studies reported a lower quality of diet, a delayed mealtime, and eating jetlag in participants with SJL vs. those without SJL. No correlation of day-to-day variability in sleep timing with average caloric intake was observed, but bed-time variability was negatively associated with diet quality. Methodological issues have been identified in sources assessed including study design, power calculation, population enrolled, and tools/metrics used for sleep timing variability assessment. Future well powered longitudinal studies, with clear protocols, standardized metrics, including all age groups from general population are needed to clarify the dietary intake consequences of variability in sleep timing.


Introduction
Sleep health is an essential element of cardiovascular (CV) and general health. The construct of Life's Simple 7, which included seven health behaviors and health factors-diet, physical activity, nicotine exposure, body mass index, blood lipids, blood glucose, and blood pressure-associated with CV disease (CVD)-free survival, total longevity and quality of life, was recently updated by the American Heart Association to Life's Essential 8, including sleep as the new health behavior in the construct [1]. Sleep is one of the foundational elements in human biology and affects numerous physiological functions. Sleep health is defined by several characteristics including sleep duration, timing, regularity, efficiency, satisfaction, and impact on daytime alertness [2].
Much of the existing research is focused on the relationship between sleep duration and CVD risk and all-cause and CV mortality [3], as well as the impact of sleep duration on various health risk factors including obesity [4]. Variability in sleep timing has been recently emphasized as a link between sleep health indicators and risk of illnesses. Social jetlag (SJL), defined as variation of wake up and sleep onset time between weekdays and weekends [5], is a frequent form of circadian rhythm disruption with 70% of students and workers experiencing ≥1 h of SJL and almost half of them experiencing ≥2 h [6]. SJL has been associated with several related risks for human health such as impaired sleep and cognitive performance, obesity, diabetes, negative CV outcomes, and psychiatric

Information Sources and Literature Search
To identify studies of interest, the literature search was conducted in 3 databases (PubMed, Embase, and Clarivate Analytics Web of Science) for articles published from the inception dates until the 31st of December 2021.
The search queries used are provided in Appendix A. Abstracts and proceedings, considered as gray literature, were identified during Embase search. References of reviews [11][12][13] were screened for relevant studies not identified during database searches.

Selection of Sources of Evidence
Investigators working in pairs screened the identified records. The search results were exported from databases, duplicate reports were removed, and the titles and abstracts of the remaining reports were screened to exclude studies that did not meet the eligibility criteria. Copies of all remaining studies were retrieved and reviewed in full to identify studies assessing dietary intake according to sleep variability and fulfilling the inclusion and exclusion criteria presented below. During the study selection step, any disagreement between the investigators was resolved by discussions and consulting a review author not actively involved in the study selection.

Data Charting Process, and Data Items
After an agreement was reached on all articles, data were extracted using an Excelbased data extraction form. The following information was extracted whenever available: author, year of publication, study design, type of variability in sleep timing reported (dayto-day, SJL), method of assessment of sleep variability, sample size, age, gender, weight status, other metabolic diseases, caloric intake, intake of macronutrients (carbohydrates, proteins, lipids, cholesterol, saturated fats), intake of food groups or dietary patterns, food preferences/appetite for certain foods, and meal timing.

Critical Appraisal of Individual Sources of Evidence
All studies included in this review were observational ones. Thus, their quality was assessed with the National Heart, Lung, and Blood Institute's Quality Assessment Tool [16]. Due to the limited number of available sources and the lack of a previous systematic or scoping review on this subject we chose to also explore their quality.

Synthesis of the Results
A narrative synthesis of the results was performed. Characteristics of studies included were categorized and summarized according to the type of variability in sleep timing (day-to-day or SJL), including populations, study designs, measures used for assessing sleep patterns variability and dietary intake and the main findings of interest.

Selection of Sources of Evidence
Our search strategy identified 3100 database records. After removing 1178 duplicates, the titles and abstracts of 1922 records were screened for eligibility criteria and outcomes of interest. Of these, 1844 were removed due to the lack of assessment of any circadian rhythm or sleep parameters, lack of assessment of any dietary intake or mealtimes, the inclusion of children or adolescents, shift workers, animal studies, records in other languages than English, or record type review, case presentation, comment, or book chapter. The full text of 78 records was retrieved for further assessment. Of these, 1 could not be retrieved, 1 presented results of the same population as an article already included, and 53 did not fulfill eligibility criteria (the majority did not assess any circadian rhythm or sleep parameters, or any dietary intake or mealtimes). Thus, 22 records (20 full-text articles and 2 conference abstracts) were included in the review (Figure 1). be retrieved, 1 presented results of the same population as an article already included, and 53 did not fulfill eligibility criteria (the majority did not assess any circadian rhythm or sleep parameters, or any dietary intake or mealtimes). Thus, 22 records (20 full-text articles and 2 conference abstracts) were included in the review (Figure 1).
Daily sleep timing variability was assessed using StDev in two studies [18,37], with he weighted average of weekly bedtimes variability dichotomized between lower and higher variability according to scores obtained with the Sleep Timing Questionnaire in one study [19] and with a self-assessment of sleep onset/offset variability in one study [36]. In one study the metrics used and the methods for sleep onset/offset calculations were not disclosed [34].

SJL and Meal Timing
No difference in caloric and macronutrient intake between groups with or without SJL was observed in cross-sectional studies that enrolled a general population sample, healthy young adults, or patients with type 2 diabetes and in which the estimation of caloric intake was performed using only one 24 h food recall [24,25,28]. However, studies that enrolled participants with obesity and obesity-related chronic conditions reported a higher caloric intake and a higher intake of carbohydrates and total fat, mono-and polyunsaturated fats, saturated fats, and cholesterol in participants with SJL as compared to those without [25,27,38]. One study examined changes in caloric intake between workdays and free days and reported a significantly higher caloric intake and a higher intake of total fat, saturated fat, and MUFA as well as a lower fiber intake during free days than during workdays in participants with SJL but not in those without SJL [20].  Participants with a morning chronotype and a SJLsc ≥ 2 h had lower adherence to a healthy Nordic diet than those with a SJLsc < 1 h (BSD scale for adherence 10.5 points vs. 11.8 points, p = 0.006), independent of age, sex, educational years, smoking, leisure-time physical activity, and energy intake. Participants with a morning chronotype and a SJLsc ≥ 2 h consumed less fruits, berries, and cereals, but more alcohol, than those with a SJLsc < 1 h (p < 0.05) independent of age, sex, educational years, smoking, leisure-time physical activity, and energy intake. Total energy intake did not differ between SJLsc groups according to diurnal preference (p = 1.00). A larger SJL was associated with lower total energy intake (β = −0.094, p = 0.013), lower grains consumption (β = −0.143, p < 0.001) and greater consumption of sugar and confectioneries (β = 0.231, p = 0.010) independent of age, sex, BMI, residential status, and WD and WE total sleep duration.
Young, healthy, and highly educated participants. No power calculation.   Most studies reported a lower quality of diet in participants with SJL vs. those without SJL [20,35] and lower adherence to the Mediterranean diet [32] or a healthy Nordic diet [30]. In the study of Almoosawi et al. [35] an inverse u-shaped association of SJL with diet quality was observed with the inflexion point at 1 h and 45 min; below this point the diet quality increased in parallel with SJL duration, while above this point the diet quality decreased. SJL was also associated with a lower consumption of total fruits, berries, vegetables, whole grains, beans, and milk [29][30][31][32] but more alcohol, sugar, or sugar-sweetened beverages, meat, and eggs [20,27,30,31]. Of note, two studies reported no difference in diet quality between SJL and no SJL [17]; the largest one enrolled only young, highly educated women [17], and the other study enrolled a very small sample and the full list of inclusion/exclusion criteria could not be assessed [25].
Prospective appetite was assessed using VAS in two small sample-size studies [24,28]. A higher perceived appetite was reported in fasting conditions for vegetables, pork, poultry, fish, eggs, milk, and dairy products in participants with SJL as compared to those without SJL [28]. Furthermore, higher ratings of hunger and prospective food intake after the meal was reported in participants with SJL > 1 h than in the SJL ≤ 1 h group despite similar caloric intake [24]. Post-meal satiety quotient (mean value of the satiety quotients for hunger, prospective food intake, satiety, and fullness) was significantly lower in participants in the 1 < SJL ≤ 2 h and SJL > 2 h groups (1.3 and 1.7 times, respectively) compared to those in the SJL ≤ 1 h group (p < 0.010) [24].
The relationship between SJL and meal timing was reported in six cross-sectional studies [20,21,[27][28][29]33] (Table 2). A delayed mealtime (breakfast, lunch and/or dinner) and eating jetlag was reported in those with SJL, in all studies that assessed these parameters [21,[27][28][29]33]. Also, a higher frequency of snacking after dinner and a later time of this snack as well as a higher caloric intake after 9 p.m. and a higher proportion of calories consumed after 9 p.m. were observed in persons with SJL than in those without [27,28].
In a study that assessed the change in meal timing in free days compared to workdays, Bodur et al. [20] reported the delayed timing of all meals irrespective of SJL presence with a significantly longer delay in breakfast timing in those with SJL than in those without.
With regards to eating window, it was assessed in two studies. In the first one, Mota et al. [27] reported overall a longer eating window in patients with obesity-related chronic diseases and SJL. In the second study, Bodur et al. [20] found a shorter eating window during free days in those with SJL as compared to those without SJL (8:42 vs. 9:00).

Day-to-Day Variability in Sleep Timing, Dietary Intake, and Meal Timing
Four cross-sectional studies assessed the day-to-day variability in sleep timing and dietary intake [18,19,34,36] (Table 3). In a study performed in young and middle-aged adults from US general population, Hooker et al. [18] found no significant correlation between the variability in sleep timing, as assessed by StDev, with average caloric intake or with variability in caloric intake. In another study using 7 days of wrist actigraphy for the assessment of the variability in sleep timing and food diaries for 7 days, Chan et al. [34] reported that young adults with bedtime variability over 90 min consumed high palatability foods at breakfast less frequently than those with bedtime variability < 90 min; no difference was observed for lunch and dinner. However, in this later study the metric used for sleep time variability was not disclosed.  A weak negative correlation between SJL and breakfast time (r = −0.23, p < 0.01) was observed, and this correlation was independent of age.
No power calculation. Young, healthy, and highly educated participants.   The largest studies were performed in Australian and Japanese populations. In the study performed in Australian middle-aged adults and elderly, Duncan et al. [19] reported that bed-time variability was negatively associated with diet quality, independent of waking time variability, usual bedtime and waking time, age, gender, smoking status, number of health conditions, work schedule, body mass index (BMI) and days of insufficient sleep. In this study the sleep variability was assessed using a weighted average of variability of sleep timing (sleep timing questionnaire) [19]. In the other study performed in Japanese adults, sleep pattern variability was self-declared for the previous year and Yamaguchi et al. [36] showed that poor sleep regularity was associated with low protein intake, high carbohydrate intake and a variability of staple foods consumption between breakfast, lunch, and dinner.
The association of day-to-day variability in sleep timing with meal timing was assessed and confirmed in only one study that enrolled patients with eating disorders and for which sleep variability was measured as StDev of center of daily inactivity [37].

Critical Appraisal of the Individual Sources of Evidence
By the National Heart, Lung and Blood Institute's Quality Assessment Tool, one study was rated as good [38], three studies as poor [23,25,36] and eighteen as of fair quality [17][18][19][20][21][22]24,[26][27][28][29][30][31][32][33][34][35]37] (Figure 2). The largest studies were performed in Australian and Japanese populations. In the study performed in Australian middle-aged adults and elderly, Duncan et al. [19] reported that bed-time variability was negatively associated with diet quality, independent of waking time variability, usual bedtime and waking time, age, gender, smoking status, number of health conditions, work schedule, body mass index (BMI) and days of insufficient sleep. In this study the sleep variability was assessed using a weighted average of variability of sleep timing (sleep timing questionnaire) [19]. In the other study performed in Japanese adults, sleep pattern variability was self-declared for the previous year and Yamaguchi et al. [36] showed that poor sleep regularity was associated with low protein intake, high carbohydrate intake and a variability of staple foods consumption between breakfast, lunch, and dinner.
The association of day-to-day variability in sleep timing with meal timing was assessed and confirmed in only one study that enrolled patients with eating disorders and for which sleep variability was measured as StDev of center of daily inactivity [37].

Main Findings Related to Research Question and Discussion of Specific Findings
This scoping review is the first one in the literature to comprehensively assess the available evidence on the effect of weekly and day-to-day variability in sleep timing on dietary intake, appetite, and mealtimes. Limited evidence from the studies discussed in this scoping review suggests that variability in sleep timing (both SJL and day-to-day variability) are associated with a less healthy diet [19,20,30,32,35], characterized by a lower consumption of fruits, vegetables, whole grains, beans, and milk [29][30][31][32] and a higher intake of sugar, or sugar-sweetened beverages, and meat compared with persons without SJL [19,27,30,31]. Also, very limited evidence available suggests that persons with SJL have a higher perceived appetite for energy-dense foods, higher ratings of hunger, and prospective food intake after the meal and lower post-meal satiety quotient [24,28].
These results are in line with previous exploratory studies that showed that acute circadian dysregulation induced by alterations in the sleep-wake schedule is associated with increased ratings of hunger, prospective food consumption, and a desire to eat savory foods [39,40]. Interestingly, chronic circadian dysregulation in exploratory conditions was associated with decreased energy expenditure and decreased hunger and appetite for various foods [41]. The mechanism regulating unhealthy food preferences in SJL and day-today variability in sleep timing is unknown. Preliminary observational data from functional magnetic resonance scans suggest an increased activation of brain regions associated with reward in persons with SJL independent of sleep duration [42]. Food intake is regulated by (1) the homeostatic system, controlled by structures located in the hypothalamus and the brainstem which regulates the need for food intake [43][44][45], and (2) the reward system, consisting of structures from the mesolimbic pathway and having dopamine as the main neurotransmitter [46,47], which regulates the hedonic aspects of feeding and the pleasure to eat. In addition to the homeostatic and the hedonic control of eating, the circadian system regulates the time of eating [48] and circadian dysregulation may result in changes of eating behavior [49]. It has been shown that caloric intake and food preferences have circadian rhythms [50][51][52] and are controlled by the circadian system through projections from the suprachiasmatic nucleus master clock to hypothalamic nuclei regulating homeostatic feeding behavior [53][54][55] and to the striatum regulating hedonic food intake [49]. Thus, the circadian system plays a key role along with the homeostatic and the hedonic systems in the regulation of food intake. Other mechanisms potentially associated with the unhealthy food choices observed in those with SJL and day-to-day variability in sleep timing are the use of certain foods, such as sugar-sweetened beverages or sugar-containing foods, being stimulants allowing subjects prone to sleep loss to stay alert [22,56]. Additionally, the preference for certain foods may be influenced by social factors and limited time available for sourcing and eating certain foods due to work schedules and daily habits, as reported by shift workers [56].
In addition to an unhealthy diet, studies that enrolled participants with obesity and obesity-related chronic conditions included in this review reported a higher caloric intake and a higher intake of carbohydrates and total fat, mono-and polyunsaturated fats, saturated fats, and cholesterol in participants with SJL as compared to those without [25,27,38]. Despite the poorer quality of diet and an increased self-rated appetite for unhealthy foods, no difference in caloric and macronutrient intake between groups with or without SJL was observed in studies that enrolled the general population and healthy young adults [24,26,28]. This could be explained by the use of 24 h dietary recall only for the previous day which may not adequately reflect the usual caloric and macronutrient intake. In addition, many studies that showed no difference in caloric intake included participants with higher education and good health literacy on the link between diet and health which might mitigate SJL-related food behavior. This latter hypothesis is supported by our previous observation on higher cognitive restraint scores in highly educated subjects with SJL. In these subjects, caloric or macronutrient intake was similar to those of subjects without SJL despite an increased self-rated appetite for energy-dense foods [28]. Additionally, obesity is commonly associated with an increased caloric intake; thus, it remains to be established whether the higher caloric intake in persons with obesity and SJL is a consequence of SJL or just coexisting with SJL in the context of obesity.
Other observations, although arising from limited evidence, point toward a delayed mealtime and eating jetlag [20,21,[27][28][29]33] as well as a higher frequency of snacking after dinner [27,28], a higher caloric intake after 09:00 p.m. and a higher proportion of calories consumed after 09:00 p.m. [27,28] in persons with SJL than in those without. Meal timing variability was also reported in persons with day-to-day variability in sleep timing, although the evidence was in only one study available in the literature, performed in patients with eating disorders [37] and thus limiting the generalizability of the results. Caloric intake during inappropriate circadian phases (e.g., during typical sleep time or biological night) has been associated with metabolic changes thought to be responsible for the deleterious effect on health in both animal models and humans-obesity, cardiovascular diseases, and diabetes [7,[57][58][59]. In animal models it has been shown that at the same caloric intake and locomotor activity, the weight gain is higher if the food intake occurs during the biological night as compared to the biological day [58,59]. Furthermore, feeding during the biological night resulted in increased weight gain, while restriction of feeding to the biological day prevented this effect [59]. In humans studied in exploratory conditions under a circadian misalignment protocol mimicking the night shift, McHill et al. [57] showed that the thermic effect of food (TEF) and carbohydrate and protein utilization decreased in response to a late-night dinner. Similarly, TEF was significantly lower after a meal consumed during the night (at 01:00 a.m.) as compared to meals consumed during the morning (09:00 a.m.) and afternoon (05:00 p.m.) [60]. TEF accounts for up to 10% of daily energy expenditure; thus, a reduced TFE may promote weight gain if the caloric intake is maintained at the same level [57]. In real-life settings, Baron et al. [61] showed that caloric intake after 08:00 p.m. predicted the BMI even after controlling for sleep timing and duration. Confirming these findings, Gill et al. [62] showed that restricting the food intake to daytime (06:00 a.m. to 06:00 p.m.) resulted in lower caloric intake and weight loss.

Gaps Identified in the Sources Included, Implications for Future Studies
Although data on SJL and its association with dietary intake are accumulating, evidence on the day-to-day sleep timing variability and its relationship with dietary intake is scarce. Only four studies are available on its association with diet quality, and caloric and macronutrient intake [18,19,34,36], and only one study is available that evaluates its correlation with mealtimes [37].
Furthermore, some methodological issues with existing studies must be noted. All studies included  but one [38] were cross-sectional thus making difficult the inference of a causal relationship between sleep timing variability and dietary intake and their effect on health outcomes. For most of the studies assessed, a sample size calculation was not performed [18][19][20][21][22][23][24][25][27][28][29][30][32][33][34]37] or did not assess sleep variability [17,26,35,36] and, thus they may be underpowered to assess the outcomes. Furthermore, in several studies the association of SJL with dietary intake or meal timing was an incidental finding and its was not included among published study objectives [17,21,26,33].
If a consistency was observed in measuring SJL (Munich Chronotype Questionnaire or several items from this questionnaire), for studies assessing the day-to-day variability in sleep timing, no systematic approach has been used and the metrics varied from StDev to self-declared variability and this may limit the accuracy of the results [18,19,34,36,37]. Although StDev is an accepted metric that estimates sleep variability across multiple days it does not consider naps and fragmented sleep and tends to underestimate sleep variability if calculations are based on data collected on ≤7 days [10]. In one of the two studies that assessed sleep variability using StDev, data were collected for 7 days of sleep actigraphy [18]. Thus, the lack of any correlation between sleep variability and caloric intake in this study may be also due to the metric used.
Although SJL was relatively consistently assessed, calculations were performed using time in bed instead of mid-sleep time in 2two studies [30,32], thus ignoring sleep latency. Also, although in the original formula for SJL calculation developed by Wittmann et al. [5] SJL is calculated as the absolute difference between the mid-point of sleep time on free days and workdays, more recently, Jankowski [63] proposed a formula corrected for sleep debt. Debate exists regarding the use of these formulas and there is a need to clarify which metric reflects better the circadian misalignment and is associated with adverse health outcomes [64].
Chronotype influence on the relationship between sleep timing variability and dietary intake was not considered in all but one study, which showed the dietary influence of SJL only in participants with a morning chronotype [30]. SJL is more frequent and greater in evening chronotype individuals who prefer a later bedtime and wake up time [64]. Previous research showed changes in clock gene expression patterns in both evening chronotype and SJL [65] and similar disease risk in humans [64]. Both are age-dependent and vary in parallel-chronotype delays up to the end of adolescence and then advances with age, while SJL is higher during adolescence and then decreases through adulthood until retirement [64]. From these observations arises the question on whether the evening chronotype can be considered a form of SJL. And the answer is most probably 'no' as although chronotype is a state of phase of entrainment reflecting the continuous adaptations of circadian system to zeitgeber changes [66,67], it also has a genetic basis [68]. SJL is a metric describing the chronic misalignment between the circadian system and social schedules, results from sleep timing variability between weekdays and weekends and includes a component of sleep debt that may influence associated health risks [64]. Furthermore, SJL is not solely seen in evening chronotypes; due to social pressure, it is observed in morning chronotypes [64]. Thus, studies should explore the chronotype-related differences in sleep and meal regularity.
Given the identified gaps detailed above, well-powered studies are needed to assess whether sleep variability is associated with dietary intake. Even more so, longitudinal studies may help clarify a potential causal relationship between sleep variability and dietary intake and other health issues reported in the literature. Additionally, methodological limitations arising due to the use of unvalidated metrics for the evaluation of day-to-day sleep variability could be overcome by clear protocols, clear definitions of outcomes and metrics used, and choosing validated metrics adapted for the sample size, study length, and pertinent for research question. The inclusion of all age groups and general population, irrespective of comorbidities, without limitation of participants to young, healthy, and highly educated ones is also recommended to avoid negative results biased by health literacy that may influence food choices and food intake. Future research should also clarify whether nutritional changes associated with variability in sleep timing, are associated with poor health outcomes.

Limitations of the Scoping Review
Although this is the first scoping review assessing the published literature on the association between variability in sleep timing and dietary intake, this review has limitations. The narrative synthesis did not allow the weighting of the studies analyzed. The results may be biased by the methodological quality of the studies; most of the studies were observational and the risk of publication bias is high. The results of this review may also be biased by not including studies on this topic indexed in other databases or published in languages other than English. The methodology used varied across the included studies, in terms of design, data collection, or outcomes assessed, thus the development of a meta-analysis was not possible, and a narrative review was carried out.

Conclusions
The variability in sleep timing, either as SJL or day-to-day variability, promotes an unhealthy diet characterized by lower consumption of fruits, vegetables, whole grains, beans, and a higher intake of sugar and meat. Although limited evidence exists, it seems that persons with SJL may have a higher perceived appetite for energy-dense foods, delayed mealtime, and eating jetlag. Whether these are associated with increased caloric intake, a change in macronutrient intake, and long-term consequences on human health remains unknown. In addition to the limited body of evidence currently available, conclusive statements on the findings of this review are also limited by the methodological variations observed across studies. Future well powered longitudinal studies, with clear protocols, standardized metrics, including all age groups from general population are needed to clarify the dietary intake consequences of variability in sleep timing. Additionally, future research should clarify whether these changes in dietary intake, if they are to be confirmed, are associated with poor health outcomes and whether therapeutic nutritional interventions in persons with circadian and sleep disruption could reverse some of their harmful effects.

Appendix A
Search queries used.  social jetlag (All Fields) AND food intake (All Fields) -social jetlag (All Fields) AND dietary intake (All Fields) -sleep (All Fields) AND variability (All Fields) AND food intake (All Fields) -sleep (All Fields) AND variability (All Fields) AND dietary intake (All Fields) -sleep (All Fields) and regularity (All Fields) and food intake (All Fields) -sleep (All Fields) and regularity (All Fields) and dietary intake (All Fields)