Associations of chronotype and socio-demographic factors with timing of eating in finnish preschool-aged children

The participants were drawn from the Increased Health and Wellbeing in Early Education and School Children (DAGIS) survey. The survey was conducted between September 2015 and April 2016 in 66 early childhood education and care (ECEC) centers situated in southern and western Finland and is described in more detail elsewhere [36]. A total of 864 preschoolers (24% of those invited) aged 3–6 years participated in the DAGIS study, of whom 677 (78%) were included in the current study. 175 (20%) of the original participants were excluded due to having less than 3 days of food records, or because the dates from the food records did not match the dates from their sleep data, and 12 (1%) due to having less than 7 nights of sleep records. The DAGIS study was approved by the University of Helsinki, Research Ethics Committee in the Humanities and Social and Behavioral Sciences 2/2015 (#6/2015).

A combination of food records and actigraph-measured sleep data collected on the same dates was used to create the chrononutrition variables. 3-day food records containing two weekdays and one weekend day, including food quantities and timing, were completed by the ECEC staff during preschool hours and by parents of the participants at home. 40% of the participants documented their food intake on consecutive days, while the remaining participants recorded non-consecutive days. Share of weekdays and weekend days recorded did not differ between the two approaches. Portion sizes were estimated with the help of a validated Children's Food Picture Book [37]. The food record data collected were entered into AivoDiet dietary software, where energy intake was calculated for the analyses. Further details about the collection and handling of the food record data have been described by Korkalo et al. [38]. Sleep was measured with a hip-worn ActiGraph wGT3X-BT triaxial accelerometer (Pensacola, FL, USA) for 7 consecutive days, 24 h per day except during water-based activities such as taking a swim or bath. Less than 16 h of wear from a noon-to-noon window was considered invalid. Weekend nights were defined as the nights between Friday and Saturday and between Saturday and Sunday. Additional details about the actigraph data collection and processing have previously been described elsewhere [39].

The study focused on the following chrononutrition variables derived from the food and sleep records: (i) timing of the first EO after waking up in the morning, and (ii) the last EO before falling asleep at night, (iii) duration of the fasting window, (iv) number of EOs, (v) duration of morning and (vi) evening latency as well as (vii) the timing of eating and (viii) energy midpoints. An EO was defined in accordance with earlier research [14, 40] as food and beverages recorded within 15 min. EOs containing less than 210 kJ (6.7% of total entries) were excluded. The fasting window was defined as the time elapsed between the last EO at night and the first EO the next day. Morning latency was calculated by subtracting the wake-up time from the time of the first EO, and evening latency was calculated by subtracting the time of the last EO from the time the participant fell asleep. All included participants provided 3-day food records with corresponding sleep data. However, if a participant lacked complete food records for both evening latency and morning latency on the following day, the corresponding variable was classified as missing (NA) for that instance. Eating midpoint was determined as the mid clock time between the first and last EOs within a single day, and energy midpoint was defined as the time of day when half of the total energy intake was reached. For all variables, separate counts were made for all 3 days as an average as well as for weekdays and weekends separately. For all chrononutrition variables, except the number of EOs, weekends were defined as spanning from 18:00 on Fridays to 17:59 on Sundays. This definition aligns with the transition from weekday routines to weekend behaviors, capturing changes in timing patterns that may occur as individuals shift away from a structured weekday schedule. For the analysis of the number of EOs, Saturday and Sunday were classified as weekend days, while Monday through Friday were considered weekdays, reflecting the structured routine associated with preschool attendance during the week.

Sociodemographic data, sex (boy/girl), and date of birth were obtained through a questionnaire filled out by parents. Age was calculated by subtracting the birthday from the research date, and the children were based on midpoint divided into two groups, 3–4 yrs and 5–6 yrs old. The highest parental educational level in the family was used as an indicator for SES. The level of education ranged from comprehensive school to a doctoral level degree. Responses were categorized for analyses into three groups: (1) low SES (high school degree, vocational diploma, or less), (2) medium SES (associate or bachelor's degree), and (3) high SES (master's, licentiate, or doctorate degree). Parents were asked to provide information about their work hours, resulting in the following three categories: (1) Regular working hours (regular daytime hours (6–18)), (2) shift work (regular night work, fixed and rotating two- and three-shift patterns), and (3) do not work (not employed). Participants who checked alternatives that did not conform to the classification schemes (seasonal work, work only on weekends, other working arrangements, part-time jobs, and entrepreneur) or had missing information were excluded from that particular analysis. The formula by Roenneberg et al. [17] was applied to calculate continuous chronotype tendency as follows: mean midpoint of weekend sleep—(mean weekend sleep duration—mean weekly sleep duration) / 2. If sleep duration was longer on weekends, the midpoint of sleep was adjusted for potential sleep debt accumulated during the weeknights. In cases where sleep duration was not longer on weekends, the weekend midpoint value was used directly. None of the children attended preschool on weekends; therefore, it was assumed that weekend sleep (Friday-Saturday, Saturday-Sunday) was free from a set schedule. Categorical chronotype tendency was determined using a conservative method based on prior studies in similarly aged children reporting the prevalence of children with evening chronotype being 10–26% [41 –44]. Children in the lowest 10th percentile were classified as having a morning chronotype tendency, those in the highest 10th percentile as having an evening chronotype tendency, and those between the 10 and 90th percentiles as having an intermediate chronotype tendency.

Before conducting the analyses, the assumptions underlying each statistical method were systematically checked. For one-way analysis of variance (ANOVA), normality was visually assessed using Q-Q plots, and the homogeneity of variance was evaluated using the Breusch-Pagan test. For analyses of covariance (ANCOVA) and multivariate linear regression, assumptions regarding linearity and homoscedasticity were visually examined through residual plots, while the absence of multicollinearity among predictor variables was assessed using the Variance Inflation Factor (VIF). Additionally, potential outliers were identified and examined to ensure robust results. Addressing these assumptions strengthened the validity of the statistical findings and the conclusions drawn from the analysis.

The associations between each independent determinant (chronotype tendency, SES, parents' work hours, age, sex) and the chrononutrition variables (first and last EO, fasting window, morning and evening latency, eating and energy midpoint, number of EOs) were assessed via ANOVA. To preserve the full range of individual variability regarding continuous determinants (chronotype and age), univariate linear regression was conducted to test for associations with the chrononutrition variables. To produce results for each potential determinant's relationship with the chrononutrition variables independent of the other determinants, ANCOVAs with all independent determinants included simultaneously were applied. Tukey's post-hoc test was performed for comparisons between subgroups following significant associations from the ANCOVAs. Multivariate linear regression was carried out for continuous chronotype and age, with all independent determinants incorporated concurrently, enabling the examination of both continuous variables' independent relationships with the chrononutrition variables.

All analyses were conducted using Statistical Software R version 4.4.1. Results with p < 0.05 were considered statistically significant.

Chronotype tendency showed an association with several chrononutrition variables. Individuals with an earlier chronotype had earlier timings for their first and last EOs, as well as earlier eating and energy midpoints. This finding aligns with previous results from a cross-sectional study conducted on school-aged children in China [23]. Contrary to an earlier study conducted on 496 children aged 7–11 years in Hong Kong, which indicated that children with evening chronotype tendency are more likely to skip breakfast, leading to longer morning latency [24], our study found that during weekdays, children with morning chronotype tendencies exhibited the longest morning latency while children with evening chronotype tendencies had the shortest morning latency. One explanation could be that the children in our study were younger than in the previous research and therefore had less autonomy regarding timing. No association was found between chronotype tendency and the duration of morning latency on weekend days in the current study. This suggests that the shorter morning latency on weekdays observed among children with later chronotype tendencies may be attributed to a structured preschool schedule, which leaves them with less time between waking up and having their first EO compared to children with earlier chronotype tendencies. In accordance with results in the previous studies conducted in Asian children aged 7–12 years [23, 24], children with morning chronotype tendencies in our study had the earliest first EO, while children with evening chronotype tendencies had the latest first EO. In agreement with the earlier findings in older children [23, 24], children with evening chronotype tendencies in this study had their last EO an hour later in the day than children with morning chronotype tendencies. Although later eating times are generally assumed to shorten the fasting window [45], this was not evident in our sample, as no significant difference in the duration of the fasting window was observed between the groups on neither weekdays nor weekends. This discrepancy may be explained by the fact that children with an evening chronotype tendency experienced longer evening latency compared to their morning counterparts, averaging a difference of 31 min on weekdays and 1 h and 27 min on weekends.

Children identified as having a morning chronotype tendency tend to consume a greater proportion of their daily energy intake earlier in the day [30]. This aligns with our results, which showed that children with morning chronotype tendencies achieved both their energy and eating midpoints earlier than their evening counterparts. The differences in eating timing between chronotypes are often more pronounced on weekends, when individuals are not bound by structured schedules, such as preschool commitments, allowing families to follow their natural preferences [21]. A previous review concluded that adults with a morning chronotype typically exhibit a more consistent eating schedule, while evening chronotypes tend to shift their eating habits later on weekends compared to weekdays [2]. Similarly, the current study observed a notable shift in eating behaviour on weekends, with pronounced differences between chronotype tendencies regarding first and last EOs, energy midpoint, and evening latency. Additionally, it was found that children with evening chronotype tendencies had fewer EOs during the weekend compared to children with intermediate or morning chronotype tendencies, although this pattern did not hold during the week. This phenomenon may be attributed to children with evening chronotype tendencies waking up later on weekends [2], which may lead to later energy intake. Consequently, these individuals may not feel hungry in the morning, making them more likely to skip breakfast and lunch, thereby resulting in fewer EOs overall [8, 46]. Nonetheless, it is important to consider the role of parents, as young children may lack the autonomy to decide whether to skip EOs.

Children from high SES backgrounds exhibited later energy midpoints on both weekdays and weekends, while other chrononutrition variables showed no notable differences between the SES groups. Higher SES families may have more structured daily schedules with later dinner times, aligning with parental work hours and children's extracurricular activities. An Australian online study on children aged 6 months to 6 years, found that eating dinner later was not associated with high SES, but the parents rated the importance of family meals higher and also had a higher frequency of family meals [47]. Hence, the family mealtime environments and behaviours may impact the children's energy intake at mealtimes. Children from lower SES backgrounds tend to rely more on meals provided at preschool, with ECEC meals appearing to help balance differences in dietary quality between children of different socioeconomic groups (DAGIS, unpublished result). Most research concerning children has primarily concentrated on SES's association with nutritional intake [1, 38, 48], rather than on the timing of food consumption. Nonetheless, children with lower SES have been linked to higher likelihood of skipping meals, particularly breakfast, when compared to their higher SES counterparts [49, 50]. This was not observed in the current study. A possible explanation would be that, in addition to lunch and snacks, breakfast is provided free of charge at ECEC centers in Finland.

Parents' work hours were modestly associated with the chrononutrition variables in this study. Mothers' work hours were more influential during the week, while fathers' hours had a stronger association with the chrononutrition patterns on the weekend. Several studies have shown that fathers spend more time with their children during weekends compared to weekdays [51, 52], and therefore, might have a bigger impact on the timing of food intake on free days. A population-based cohort in Taiwan found that preschool-aged children whose parents worked nonstandard hours were less likely to eat breakfast regularly [22], which is in accordance with our findings, where children with mothers working shifts were associated with later first EOs and a longer fasting window on weekdays compared to children with mothers working regular hours. Parents' shift work tends to disrupt schedules more during weekdays because of the set routines e.g., preschool, while families have more flexibility during the weekend [53]. During the weekend days, children with nonworking fathers had later last EOs compared to those with working fathers, possibly due to more flexible routines.

Results from the present study indicated that age and sex were only associated with differences in the chrononutrition variables observed on weekend days. Children in the younger age group had a shorter fasting window, on average 12h12min compared to children in the older age group whose fasting window on average lasted 12h36min. An explanation for the longer fasting window duration in the older children is a potentially longer sleep duration on weekend nights compared to the younger age group children. By age 5, 94% of children no longer take naps [54]; thus, more participants in the younger age group may still be napping daily, leading to fewer hours of night time sleep and affecting the duration of the nightly fasting window [55].

This study found no differences between the sexes during weekdays, but on weekend days, boys had a longer morning latency and fewer EOs compared to girls. No previous research on preschool-aged children comparing the sexes regarding chrononutrition factors was found. However, a Polish study on preschool-aged children found that parents were more concerned with managing their daughters' overall eating habits than their sons' [56], perhaps explaining their shorter morning latency in the current study. An observational study in the U.S conducted on kindergarteners showed that girls tend to seek parental approval more often and are praised more frequently than boys by their mothers when eating [57]. Boys might prefer to engage in other activities after waking up, such as playing or using electronic devices, which might delay their first EO and lead to fewer EOs overall. More research is needed to draw conclusions about differences between the sexes regarding morning latency and the frequency of EOs, as skipping breakfast and a lower number of EOs are associated with overweight [12, 58], and studies show that overweight is more prevalent in boys compared to girls [59, 60].

It is important to recognize certain limitations inherent in this study. First, the families participating in this study had higher levels of education compared to national averages [61], potentially impacting the generalizability to our findings. All participants attended preschool, so the results might differ for children cared for at home. However, since the majority (78.6%) of Finnish children aged 3–5 attend preschool [62], the generalizability of the findings remains largely intact. Second, the cut-off scores for chronotype tendencies in this study are specific to our sample, as participants were grouped based on the lowest and highest 10th percentiles, a method previously used in the DAGIS study [63]. While this approach assumes a normal distribution of chronotypes similar to the broader population, the cut-off scores might not align with those of the entire population of children the same age. Furthermore, with conservative prevalence estimates set at 10%, there is a possibility that some children with morning or evening chronotype tendencies were categorized into the intermediate chronotype group. However, through this approach, we most likely captured true positives in the morning and evening chronotype tendency groups, and in addition, the analyses were also conducted with chronotype as a continuous variable. While weekend night sleep data were utilized to determine chronotype, we cannot guarantee that children were not awakened on weekend mornings for example due to scheduled activities such as hobbies. Additionally, we did not collect data on parental chronotypes, which may influence the children's schedules for timing of sleep and food intake. Third, the study's cross-sectional nature means that causality cannot be concluded from the findings, as it only captures data at a single point in time.

To our knowledge no previous research has examined the determinants of chrononutrition factors in preschool aged children. The strengths of the current study include its relatively large sample situated in both rural and urban areas of Finland. Participants' parents and preschool personnel recorded food and beverage intake in real time, while actigraphy was used to measure sleep, reducing recall bias. Food records were collected on the same dates as three of the 7 days of actigraphy data, encompassing both weekdays and weekends. This simultaneous data collection provided a more accurate representation of chrononutrition variables. Including both weekdays and weekends allowed for a comprehensive view of how the determinants of chrononutrition patterns varied between weekdays and weekends.