Associations Between Sleep and Dietary Patterns Among Low-Income Preschool Children

Abstract

Background:

Sleep disturbances and low-quality diets are prevalent among children in low-income settings, yet the nature of their relationship remains unclear. In particular, whether aspects other than sleep duration- including timing and quality- are associated with dietary patterns has rarely been examined, especially among preschool-aged children.

Objective:

To evaluate whether nightly and total sleep duration, sleep timing, differences in timing and duration from weekdays to weekends, and sleep quality were related to dietary patterns.

Design:

A cross-sectional analysis of preschoolers. Parents completed questionnaires about children’s sleep habits as well as a semi-quantitative food frequency questionnaire (FFQ).

Participants/setting:

354 English-speaking children (49.9 % male) with no serious medical conditions aged 3-5 years enrolled in Head Start in Michigan (2009-2011) with complete information on sleep and diet.

Main outcome measures:

Dietary pattern scores derived from food frequency questionnaire

Statistical analyses performed:

Principal component analysis was used to identify dietary patterns. Separate linear regression models with dietary pattern scores as the dependent variable and continuous sleep measures as independent variables were used to evaluate associations between sleep and diet, adjusting for sex, age, parental education, and sleep hygiene.

Results:

Three dietary patterns were identified- Vegetables, Healthy Proteins & Sides; Breads & Spreads; and Processed & Fried. Longer average weekend sleep duration and a greater difference in weekend-to-weekday sleep duration was related to lower Vegetables, Healthy Proteins & Sides pattern scores. Later sleep midpoint during the weekday was related to lower Vegetables, Healthy Proteins & Sides pattern scores, while later sleep midpoint on the weekend was associated with higher Processed & Fried pattern scores. Similarly, a larger weekend-weekday midpoint difference was associated with higher Processed & Fried pattern scores.

Conclusions:

Later sleep timing and differences in sleep duration and timing from weekends to weekdays were related to less-optimal dietary pattern scores in young children.

INTRODUCTION

Children from lower-socioeconomic environments have a higher likelihood of disordered sleep,¹ poorer quality diets,² and high rates of childhood obesity.³ Sleep disturbances and low diet quality are both predictive of obesity risk in children;⁴ thus, examining the associations between sleep and dietary patterns is an important step toward understanding the pathways to obesity risk, particularly in vulnerable populations.^5,6 Further, sleep and dietary habits are established at a young age,^7,8 and habits developed even as early as preschool age may have implications for later risk.^9,10

In populations of older children, there is evidence of associations between sleep and diet. For example, cross-sectional studies of school-age children and adolescents in the US showed that short sleep duration was related to greater energy density, sugar-sweetened beverage, and added sugar intake.^11,12 Similarly, studies among children in Portugal¹³ and Finland¹⁴ found that shorter sleep duration was related to energy-dense dietary patterns. Other aspects of sleep, including timing (i.e., late bedtimes and wake times; late midpoint of sleep), have also been related to lower quality diets.^12,15 One comprehensive 12-country study among 9- to 11-year-olds¹⁶ found that shorter sleep duration, lower sleep efficiency, and later bedtimes were each associated with unhealthy dietary patterns high in added fats and sugars (e.g., fast food, hamburgers, soft drinks, sweets), while earlier bedtimes were related to healthy dietary patterns (vegetables, fruit, whole grains). A similarly conducted study of 9- to 11-year-olds from New Zealand showed that children in the late sleep/late wake category had a lower “Fruit and Vegetables” pattern score.¹⁷

In contrast to the growing number of studies on sleep and diet in school-aged children and adolescents, there are fewer studies in preschool-aged children. One potential reason for the lack of studies is that parents are assumed to have considerable control over their young children’s habits. Nonetheless, child eating behaviors (e.g. picky eating or temper tantrums) and temperament (e.g. negative emotionality)¹⁸ may influence parental feeding practices.^19,20 Further, there is recent experimental evidence that short sleep alters dietary preference and intake even among young children.²¹ In this study, researchers found that US preschool children whose sleep was restricted by ~3 hours (bedtime delay and no nap) had a 20% increase in total calories the next day due to higher intake of sugar, carbohydrates and fat when compared to baseline. Whether these associations persist in free-living environments is less clear, although one observational study showed that longer sleep duration in 2 to 4-year-olds was associated with greater fat but lower carbohydrate intake, and that higher shifts in sleep duration from weekdays to weekends was related to higher calorie consumption.^22,23 In contrast, a different study in low-income preschoolers did not report associations between nighttime sleep duration and diet after accounting for confounding factors.²⁴

Thus, the overarching aim of our study was to evaluate how multiple indicators of sleep health related to diet in preschoolers from low-income households. We chose to focus on dietary patterns because foods naturally cluster in the diet (e.g. milk and cereal); thus, the analysis of dietary patterns may provide interpretations that are more intuitive and relevant for public health than examination of single foods.²⁵ We hypothesized that multiple distinct dietary patterns would be identified within this dataset, and that longer sleep duration, earlier sleep timing, less difference in weekday to weekend sleep duration and timing, and higher sleep quality would each be associated with higher healthy dietary pattern scores and lower unhealthy dietary pattern scores.

METHODS

Study Population

Participants were children enrolled in a longitudinal study of stress and eating behavior; the present study is a secondary analysis of the recruitment visit, which occurred during the school year (October at the earliest) from 2009-2011. Families were recruited from Head Start, a federally-funded preschool program for low-income families. The study was approved by the University of Michigan Institutional Review Board; parents/legal guardians (typically mothers) provided written informed consent and families were compensated $90 for participating. Inclusion criteria were that the child was enrolled in Head Start (either in the first or second year), not in foster care, born at ≥ 35 weeks gestation without serious perinatal/neonatal complications, and had no serious medical problems affecting growth or appetite. Additional criteria were that the parent and child spoke English and that one parent did not have a 4-year college degree. All questionnaires were interviewer-administered.

Of the full sample of 380 children, 25 were excluded either due to missing or implausible dietary data.

Sleep Measures

A parent reported their child’s usual bedtime and wake time on weekdays and weekends, as well as the number of days per week their child napped and the typical nap duration. From these data, we calculated average weeknight sleep duration, average weekend-night sleep duration, difference between weekend and weeknight sleep duration, and total average 7-day sleep duration including naps. These were all reported in hours and minutes.

The midpoint of sleep, calculated as the median time of the sleep period (bedtime to wake time), was the measure of sleep timing. We also calculated the difference in midpoint from weekend to weekday.

Parents completed the 25-item Children’s Sleep-Wake Scale, which has been validated in preschoolers with sleep diaries and actigraphy, to assess sleep quality (Cronbach’s α based on the present sample=.89). The scale includes five components: Going to bed, falling asleep, maintaining sleep, reinitiating sleep, and returning to wakefulness; individual questions pertaining to these components were rated from 1=never to 6=always and a mean sleep quality score was generated.²⁶ Higher scores indicated better sleep quality, and included items such as whether the child typically resists bedtime, has difficulty falling asleep, is restless during the night, or difficult to get out of bed in the morning.

Parents also completed the 22-item Children’s Sleep Hygiene Scale (Cronbach’s α based on present sample=.73), which has been assessed for internal consistency and content validity in preschoolers; items were rated from 1=never to 6=always and a mean sleep hygiene score was generated.²⁷ Higher scores indicated more optimal sleep hygiene practices such as avoiding caffeine and maintaining a consistent sleep-wake schedule.

Child Dietary Pattern Scores

Parents completed the Harvard Service Food Frequency Questionnaire which estimates usual intake habits of preschool children via parental proxy and has been validated in low-income populations, with energy-adjusted correlation coefficients ranging from 0.26 for dietary fiber to 0.63 for magnesium.²⁸ Parents were asked to recall how often in the past month the child typically consumed one serving of a standard portion size of 84 different food items. The dietary recall period included preschool meals that a parent was not present for, but per state regulations, Head Start classrooms were required to inform parents of the daily food offerings ²⁹. Response options to the dietary frequency questions ranged from never to ≥6 times per day, which we converted into times/day (e.g. once per month=1/30.4= 0.03 times/day). United States Department of Agriculture Food Composition tables³⁰ were also used to convert the portions/day into average daily total energy intake, and a trained nutritionist evaluated the plausibility of each participant’s total energy intake (<500 or >3500 calories deemed implausible) and response option variance (<1.27 overall variation deemed implausible).

Covariates

Parents provided relevant sociodemographic information, including child’s age and race/ethnicity, his/her highest level of education attained, and household income from all sources. Trained research assistants measured height (in meters) and weight (in kilograms) of children during home visits using calibrated equipment (Seca 213/217 portable stadiometer, Seca, Chino, California and Detecto Portable Scale Model #DR550C, Detecto, Webb City, Missouri, respectively). Child race/ethnicity was classified as non-Hispanic/Latino White, non-Hispanic/Latino Black, other non-Hispanic/Latino, and Hispanic/Latino of any race. Body mass index (BMI) z-scores for sex and age were calculated and weight status categories generated (overweight/obese as BMI ≥ 85^th percentile for sex and age, vs. not overweight as BMI < 85^th percentile for sex and age) based on the Center for Disease Control reference norms.³¹ Parental education was classified into three categories: did not complete a high school education or General Equivalency Diploma (GED), completed a high school diploma or GED, or completed at least some post-high school education. Income-to-needs ratio was calculated as the total reported household income divided by the federal poverty line for a family of the same size during the corresponding year. Household chaos was measured as the sum of 15 true/false items on the Confusion, Hubbub, and Order Scale³² (CHAOS; Cronbach’s α based on present sample=.79; theoretical range 0 to 15), with higher scores indicating more chaotic environments. Household routines were measured with the 14-item Child Routines Inventory³³ which assesses how regularly the child engages in family routines that involve interaction with or supervision by a parent (e.g., eating together), on a 5-point scale (from 0 = “never” to 4 = “nearly always”), with higher scores indicating greater presence of routines (Cronbach's alpha based on present sample = .70).

Statistical Analyses

First, dietary patterns were identified based on how the intake of food groups correlate, using an often-used methodology.³⁴ Figure 1 provides a flowchart of the process followed. Briefly, similar food items were first grouped into food groups based on nutritional similarity (Table 1), and total-energy adjusted food group intakes were computed using the residual method. ³⁵ Next, a principal component analysis of the food groups was performed and the components were transformed with Varimax rotation to obtain uncorrelated components and improve interpretation. The number of components to retain was determined based on visual inspection of the Scree plot, eigenvalues>1, and interpretability of the components. The food groups with the highest component loadings enabled interpretation of the pattern; groups with component loadings ≥∣0.25∣ are shown in Table 2. To determine how closely each child’s diet aligned with the dietary pattern, scores were computed by multiplying the component loadings of each food group by the child’s frequency of intake in that food group and then summing (because there were food groups with negative component loadings, this meant that the overall raw scores could have been <0). Each child received a score for each of the dietary patterns, with higher scores representing a closer correspondence to the dietary pattern. Finally, each of the scores were converted to Z-scores with a mean of 0 and standard deviation of 1 to facilitate comparisons between the 3 dietary patterns.

Figure 1.

Flowchart depicting the creation of dietary pattern scores in a sample of 355 low-income Head Start preschoolers enrolled in a study on stress and eating behavior

^a FFQ= Food Frequency Questionnaire

^b TEI= total energy intake

Table 1 (Suggested Online-Only Table).

Foods in each food group, derived from a semi-quantitative food frequency questionnaire^a completed by parental proxy for 355 low-income preschoolers attending Head Start

Food Groups	Foods
Potatoes	Potatoes; sweet potatoes or yams
French fries	French fries, fried potatoes, tater tots
Soup	Vegetable soup; other soup
Vegetables	Corn; peas; tomatoes, tomato sauce, salsa; peppers; carrots; broccoli; green beans; spinach; mixed vegetables; squash; zucchini; cabbage, coleslaw, cauliflower; lettuce salad
Rice	Rice
Legumes	Beans
Fruit	Banana; peaches; fruit cocktail, mixed fruit; orange or grapefruit; apple or pear; applesauce; grapes; strawberries; melon; pineapple; raisins or prunes
Juice	orange juice or grapefruit juice; other juice
Cheese	Cheese
Yogurt	Yogurt
Milk	Milk
Peanut butter and other nuts	Nuts; peanut butter
Hot Cereal	Hot cereal, grits
Cold Cereal	Cold cereal
Bread	Bread, toast, roll or pita; English muffin or bagel; biscuit; cornbread or tortilla
Salty snacks	Chips; popcorn or pretzels; crackers;
Mixed dish	Macaroni and cheese; pizza; spaghetti or other pasta; tacos, burritos
Processed meat	Hot dogs; sausage; cold cuts; fried chicken or turkey; bacon
Non-processed meat	Hamburger; other chicken or turkey; pork or ham; roast beef or steak; liver, organ meats
Fish	Canned tuna; fried fish, fish sticks; other fish
Spreads with fat	Mayonnaise; salad dressing; butter; margarine
Eggs	Eggs
Breakfast pastries	Donut; sweet roll or muffin; pancake, waffle or French toast
Sweetened dairy	Hot chocolate; ice cream; pudding
Sweets	Cookies or brownies; cake or cupcake; pie; chocolate or candy bar; other candy
Sugar-sweetened beverages	Fruit drinks; soda, soft drink, pop (not sugar free); other juice
Diet soda	Soda, soft drink, pop (sugar free)

^aFood frequency questionnaire used was the Harvard Service Food Frequency Questionnaire

Table 2.

Component loadings for dietary patterns derived from semi-quantitative food frequency questionnaire data completed by parental proxy for 355 low-income preschoolers attending Head Start

	Component 1a	Component 2b	Component 3c
Food Groups	Vegetables, Healthy Proteins & Sides	Breads & Spreads	Processed & Fried
Vegetables	0.63
Non-processed meat	0.49
Legumes	0.48
Fish	0.48
Sugar-sweetened beverages	−0.48
Potato	0.46
Juice	−0.43
Eggs	0.43
Rice	0.39
Cheese	−0.27
Mixed dish	0.25
Bread		0.68
Peanut butter & other nuts		0.51
Fruit		−0.50
Spreads with fat		0.41
Yogurt		−0.35
Breakfast pastry		0.27
Milk			−0.58
French fries			0.51
Processed meat			0.45
Salty snacks			0.43
Sweets			0.33
Cold cereal			−0.29
% Variance Explained	9%	7%	6%

^aComponent 1, called Vegetables, Healthy Proteins & Sides, is characterized by high energy-adjusted intake of vegetables, non-processed meat, legumes, fish, potatoes, eggs, and rice, and low frequency of sugar sweetened beverage and juice.

^bComponent 2, called Breads & Spreads, is characterized by high energy-adjusted intake of bread, peanut butter and other nuts, spreads with fat, and breakfast pastry intake and low intake of fruit.

^cComponent 3, called Processed & Fried, is characterized by high energy-adjusted intake French fries, processed meat, salty snacks, sweets, and diet soda, and low energy-adjusted intake of milk.

In bivariate analysis, the associations between dietary patterns and potential confounders were first examined by comparing the distributions or means ± standard deviation (SD) of the potential confounders in lower versus higher dietary pattern categories (split at the median of dietary pattern scores). To evaluate the primary study questions on sleep and diet, linear regression models were run with dietary pattern scores as the continuous dependent variable and sleep measures as a continuous independent variable (separate models for each sleep measure and for each dietary pattern). In multivariable models, child age, sex, race/ethnicity, parental education, and sleep hygiene were added to each of the models. These variables were selected as confounders based on the results of the bivariate confounder analysis and prior research.^36,37

Three sets of sensitivity analyses were implemented. First, to evaluate the dose-response relationship between the sleep variables and dietary pattern scores, linear regression models were run with each sleep measure as a categorical variable (in quartiles). Because there was evidence of dose-response relationships, only the continuous β estimates are presented. Second, the following variables were added to the multivariable models separately to evaluate their potential confounding role: income-to-needs ratio, household CHAOS and household routines. Because the addition of these variables did not substantially alter the estimates, they were not included in the final multivariable models. Third, to evaluate whether associations with sleep duration and timing were independent, regression models were run with mutual adjustment for all sleep variables (separately for weekend versus weekday variables). Because estimates were not substantially different in mutually-adjusted models, analyses from separate models are presented. Analyses were conducted using Stata 14.0.³⁸ The criterion of statistical significance was P<0.05.

RESULTS

The mean (SD) age of the children was 4.2 (0.5) years. The sample was nearly equally divided by sex (49.9% male). Other descriptive statistics are shown in Table 3.

Table 3.

Sociodemographic and sleep characteristics of a sample of 355 low-income Head Start preschoolers enrolled in a study on stress and eating behavior

Variable	Mean (SD) or %
Child Age (years)	4.23 (0.53)
Child Sex
Male	49.9
Female	50.1
Child Race/Ethnicity
White, non-Hispanic/Latino	57.8
Black, non-Hispanic/Latino or othera	31.6
Hispanic/Latino, any race	10.7
Child body mass index z-scoreb	0.80 (1.04)
Child Weight Status
Not overweight	60.5
At risk for overweight (>85th percentile)	39.6
Parental Education
<High school	15.5
High school diploma or GEDc	30.4
Some education beyond high school	54.1
Household Income-to-Needs Ratio (1.0=poverty)	0.87 (0.76)
Household CHAOSd (higher=more chaos; range= 0-15)	4.09 (3.29)
Household Routine (higher=more routine; range=0-56)	45.43 (6.23)
Weeknight Sleep Duration	10 hr 22 min (47 min)
Weekend-night Sleep Duration	10 hr 45 min (1 hr 8 min)
Total Sleep per day (hours per night + naps during the day)	10 hr 39 min (2 hr 33 min)
Child’s Usual Bedtime (weekdays)	8:35 pm (45 min)
Child’s Usual Waketime (weekdays)	6:57 am (35 min)
Child’s Usual Midpoint (Median) of Sleep (weekday)	1:45 am (33 min)
Child’s Usual Bedtime (weekend nights)	9:40 pm (1 hr 7 min)
Child’s Usual Waketime (weekend days)	8:25 am (1 hr 11 min)
Child’s Usual Midpoint (Median) of Sleep (weekend)	3:01 am (60 min)
Sleep Hygiene (higher=better sleep hygiene; range= 0-10)	4.84 (0.50)
Sleep Quality (higher=higher quality sleep; range= 0-10)	4.13 (0.73)

^aIncludes Asian, Asian Pacific Islander, and Biracial, non-Hispanic

^bFrom Centers for Disease Control reference²⁵

^cGED=General Equivalency Diploma

^dCHAOS= Confusion, Hubbub, and Order Scale

Using PCA, three dietary patterns that together explained 22% of the variance in the energy-adjusted food groups were identified: Vegetables, Healthy Proteins & Sides; Breads & Spreads; and Processed & Fried (Table 3). The Vegetables, Healthy Proteins & Sides pattern was characterized by high energy-adjusted intake of vegetables, non-processed meat, legumes, fish, potatoes, eggs, and rice, and low frequency of sugar sweetened beverage and juice. The Breads & Spreads pattern was marked by high energy-adjusted intake of bread, peanut butter and other nuts, spreads with fat, and breakfast pastry intake and low intake of fruit. The Processed & Fried pattern included high energy-adjusted intake French fries, processed meat, salty snacks, sweets, and diet soda, and low energy-adjusted intake of milk. Parental education was associated with the Breads & Spreads pattern, such that children with higher scores on the Breads & Spreads pattern were more likely to have parents who completed at least some post-high school education (Table 4) compared to children with lower scores on this pattern. In addition, Black, non-Hispanic (or other race/ethnicity) children were more likely to be in the upper half of the Processed & Fried pattern scores compared to White, non-Hispanic/Latino children.

Table 4 (Suggested Online-Only Table).

Cross-sectional associations between sociodemographic characteristics and dietary patterns among 355 low-income Head Start preschoolers enrolled in a study on stress and eating behavior

		Vegetables, Healthy Proteins & Sides Pattern		Breads & Spreads Pattern		Processed & Fried Pattern
Sociodemographic variables	n	Lower scores	Higher scores	Lower scores	Higher scores	Lower scores	Higher scores
Child’s sex
Male	177	50.0	49.7	51.7	48.0	47.8	52.0
Female	178	50.0	50.3	48.3	52.0	52.3	48.0
P valuea		0.96		0.49		0.43
Child’s race/ethnicity
White, non-Hispanic/Latino	205	55.6	60.0	53.9	61.6	64.0	51.4
Black, non-Hispanic/Latino or otherb	112	36.0	27.1	31.5	31.6	25.3	37.9
Hispanic/Latino, any race	38	8.4	13.0	14.6	6.8	10.7	10.7
P value		0.12		0.05		0.03
Child’s overweight statusc
Not overweight	214	58.4	62.5	64.6	56.3	60.5	60.5
Overweight	140	41.6	37.5	35.4	43.8	39.6	39.6
P value		0.43		0.11		0.99
Parental education
< high school	55	16.9	14.1	14.0	17.0	12.9	18.1
High school diploma or GEDd	108	33.7	27.1	36.5	24.3	29.2	31.6
Some education beyond high school	192	49.4	58.8	49.4	58.8	15.5	50.3
P, trend		0.21		0.04		0.27
Child’s age, per year	355	4.2 ± 0.5	4.3 ± 0.5	4.2 ± 0.5	4.2 ± 0.5	4.2 ± 0.6	4.3 ± 0.5
P valuee		0.10		0.79		0.40
Household income-to-needs ratio, per unit	339	0.9 ± 0.9	0.9 ± 0.6	0.9 ± 0.9	0.9 ± 0.6	0.8 ± 0.6	0.9 ± 0.9
P value		0.85		0.81		0.47
Household CHAOS (higher=more chaos)e	355	4.2 ± 3.2	4.0 ± 3.3	3.7 ± 3.1	4.4 ± 3.4	4.1 ± 3.3	4.1 ± 3.3
P value		0.67		0.05		0.83
Household routine (higher=more routine)	343	45.2 ± 5.8	45.7 ± 6.6	45.6 ± 6.1	45.3 ± 6.3	46.1 ± 6.0	44.8 ± 6.4
P value		0.40		0.71		0.06

^aP values for categorical variables are from Chi-square tests (for more than 3 categories, a significant P value indicates at least 2 of the categories are different from one another)

^bIncludes Asian, Asian Pacific Islander, and Biracial, non-Hispanic

^cFrom Centers for Disease Control reference²⁵

^dGED=General Equivalency Diploma

^eP values for continuous variables are from Wald tests

^dCHAOS= Confusion, Hubbub, and Order Scale

In regression analysis, one hour longer sleep duration during the weekend was related to a −0.13 lower Vegetables, Healthy Proteins & Sides SD after adjustment for potential confounders (95% Confidence Interval (CI) −0.22 to −0.04), and every one-hour greater difference in sleep duration from the weekend to weekdays was associated with −0.14 SD scores (95% CI −0.23 to −0.04, respectively; Table 5). Later sleep midpoint during the weekday was related to lower Vegetables, Healthy Proteins & Sides pattern scores, such that every one hour later in the sleep midpoint was related to a −0.20 SD lower scores (95% CI −0.40 to −0.01). Later sleep midpoints on the weekend was associated with higher Processed & Fried pattern scores; every one-hour later sleep midpoint was associated with a 0.18 SD higher Processed & Fried pattern score (95% CI 0.06 to 0.29). Similarly, each hour of difference in weekend-weekday midpoints was associated with a 0.21 SD higher Processed & Fried pattern score (95% CI 0.07 to 0.35). Sleep quality was not associated with dietary pattern scores.

Table 5.

Cross-sectional associations between sleep duration, timing, and quality and dietary patterns among 355 low-income Head Start preschoolers enrolled in a study on stress and eating behavior

	Vegetables, Healthy Proteins & Sides		Breads & Spreads		Processed & Fried
Sleep measures	Unadjusted betaa (95% CI)	Adjusted betab,c,d (95% CI)	Unadjusted beta (95% CI)	Adjusted betab,c,d (95% CI)	Unadjusted beta (95% CI)	Adjusted betab,c,d (95% CI)
Sleep duration
Weeknight Sleep (hr/night)	0.05 (−0.09, 0.18)	0.01 (−0.13, 0.14)	0.00 (−0.13, 0.14)	−0.02 (−0.16, 0.11)	−0.05 (−0.19, 0.08)	−0.01 (−0.14, 0.13)
Weekend-night Sleep (hr/night)	−0.12 (−0.21, −0.02)*	−0.13 (−0.22, −0.04)**	−0.04 (−0.14, 0.05)	−0.04 (−0.13, 0.06)	0.00 (−0.09, 0.10)	0.02 (−0.08, 0.11)
Weekend-weekday duration difference	−0.14 (−0.23, −0.05)**	−0.14 (−0.23, −0.04)**	−0.05 (−0.14, 0.05)	−0.02 (−0.12, 0.07)	0.03 (−0.06, 0.13)	0.03 (−0.07, 0.12)
Total average sleep with naps (hr/day)	−0.04 (−0.18, 0.11)	−0.08 (−0.22, 0.07)	−0.04 (−0.19, 0.11)	−0.06 (−0.20, 0.09)	−0.03 (−0.17, 0.12)	0.02 (−0.13, 0.16)
Sleep timing
Midpoint of sleep during weeke	−0.23 (−0.42, −0.04)*	−0.20 (−0.40, −0.01)*	0.01 (−0.18, 0.21)	0.05 (−0.14, 0.25)	0.20 (0.01, 0.39)*	0.13 (−0.06, 0.33)
Midpoint of sleep during weekende	−0.15 (−0.25, −0.04)**	−0.11 (−0.22, 0.01)	−0.08 (−0.19, 0.02)	−0.04 (−0.15, 0.08)	0.21 (0.11, 0.32)***	0.18 (0.06, 0.29)**
Weekend-weekday midpoint difference	−0.13 (−0.26, 0.00)	−0.06 (−0.20, 0.08)	−0.15 (−0.28, −0.01)*	−0.10 (−0.24, 0.04)	0.25 (0.12, 0.39)***	0.21 (0.07, 0.35)**
Sleep quality (higher is better)	0.12 (−0.03, 0.26)	0.08 (−0.10, 0.25)	−0.05 (−0.19, 0.09)	0.00 (−0.17, 0.18)	−0.05 (−0.19, 0.10)	0.02 (−0.15, 0.20)

^aAll estimates are from linear regression models with dietary pattern as the continuous outcome and sleep measure as a continuous predictor, separately for each sleep measure and dietary pattern combination

^bAdjusted for child age, sex, race, parental education level, and sleep hygiene

^cSample sizes range from 345 to 352 in complete-case analysis

^dR² for the adjusted models range from 0.02 to 0.06

^eMidpoint refers to the median time between bedtime and wake time

^*=P value<0.05

^**=P value<0.01

^***=P value<0.001

DISCUSSION

In this sample of low-income preschool children, three dietary patterns were identified - Vegetables, Healthy Proteins & Sides; Breads & Spreads; and Processed & Fried - that were related to several dimensions of sleep health reported by parents. In particular, after accounting for potential confounders, longer sleep duration on weekends, later sleep midpoints on both weekends and weekdays, and greater differences in sleep duration and timing from weekends to weekdays were related to less healthy diet quality- either lower Vegetables, Healthy Proteins & Sides dietary pattern scores or higher Processed & Fried pattern scores.

This study extends prior literature as it specifically focused on sleep and diet associations in preschool-aged children, and evaluated sleep duration, timing, and quality. One of the primary findings was that timing of sleep in young children may be associated with obesogenic eating behaviors. These associations are in line with recent findings among older children and young adults, showing that later sleep timing relates to lower intake of specific healthy foods and/or to higher intake of energy-dense foods.^12,39–41 Similarly, a greater mismatch in sleep timing from weekend to weekday, also referred to as social jetlag, has also been related to obesity and unhealthy eating in older children⁴² and undergraduate students.⁴³

There are several potential mechanisms to explain the findings with sleep timing, and it is important to consider the possibility of bi-directional associations.⁴⁴ Regarding potential pathways linking sleep to diet, there is evidence that delayed sleep timing alters hormonal appetite regulators,⁴⁵ which might affect food preferences.⁴⁶ In addition, sleep timing may affect mood and stress levels,⁴⁷ which have also been associated with eating behaviors in young children.⁴⁸ Sleep timing can also affect the timing of food intake, and studies in adults show that the timing of food intake may have implications for energy metabolism and weight loss.⁴⁹ In contrast, individual components in the dietary patterns could affect sleep. For example, the Vegetables, Healthy Proteins & Sides pattern contains sources of dietary fiber and polyunsaturated fat, which have been related to more favorable sleep measures.^50,51 These nutrients may affect sleep by promoting melatonin production and regulation.⁵² In contrast, the Processed & Fried pattern contains foods high in saturated fat, which has been related to shorter sleep duration⁵³ and less slow wave sleep⁵⁰ in some studies.

Another finding was that longer sleep duration during the weekend only was associated with lower Vegetables, Healthy Proteins & Sides pattern scores. Although the finding was not in the hypothesized direction, there are some potential explanations. Children with longer sleep on the weekends may be compensating for shorter or lower quality sleep during the weekdays⁵⁴, and these weekday sleep characteristics may be responsible for the association. Indeed, associations with sleep quality and duration during the weekdays were in this expected direction but were not statistically significant. Further, the finding that a greater difference in duration from weekend to weekday was associated with lower Vegetables, Healthy Proteins & Sides, and greater Processed & Fried pattern scores is in line with this explanation.

It remains a possibility that the sleep and diet associations we observed were due to confounding factors that are associated with dietary patterns as well as sleep health. For example, parents who promote healthy eating may be more likely to also maintain consistent bedtime routines. Residual confounding due to socioeconomic status (SES) is also a possibility. However, many of the associations between sleep and diet persisted in spite of adjusting for parental education as well as sleep hygiene. Additionally, in models adjusting for measures of routine and household chaos, the estimates were unaltered.

There are both strengths and limitations of this study. Strengths included querying multiple aspects of children’s sleep and examining several potential confounders (e.g., household chaos). In addition, the examination of dietary patterns is novel in this age group, and the percent variance explained by the components was in line other studies of adolescents and adults.^55–58 The cross-sectional study design was a limitation as it means that temporality of the sleep and diet associations are unclear. There is a possibility of recall bias of self-reported measures. Parents may have reported their children's diet and/or sleep to appear more favorably (social desirability bias). Parent characteristics such as weight status or SES may also influence recall.^59,60 Further, although menus of all meals and snacks provided at Head Start are made available, ²⁹ parents may not be fully aware of food actually consumed at preschool. Although this is a notable limitation to our capacity to accurately measure child intake in the field, prior work has shown that parent FFQs even for children attending preschool show reasonable validity. ^28,61–64 As well, although we cannot rule out systematic error, it may be that parents’ lack of knowledge on their child’s diet primarily introduces random measurement error, which would decrease precision of the study estimates but should not unduly bias the results.⁶⁵ Finally, it is important to note that the generalizability of these findings may be limited, as the study sample consisted of low-income preschool-aged children attending Head Start in southeast Michigan.

CONCLUSION

In summary, in a sample of low-income preschoolers, delayed sleep timing, longer sleep duration on weekends, and greater differences in sleep timing and duration from weekends to weekdays were associated with less healthy dietary pattern scores. Longitudinal investigations are needed to further elucidate the relationship between sleep health and diet in young children.

RESEARCH SNAPSHOT.

Research Question:

Are measures of sleep- nightly and total sleep duration, sleep timing, and sleep quality- associated with dietary patterns in low-income preschool-aged children?

Key Findings:

Longer average weekend sleep duration and a greater difference in weekend-to-weekday sleep duration was related to lower Vegetables, Healthy Proteins & Sides pattern scores. Later sleep midpoint during the weekday was related to lower Vegetables, Healthy Proteins & Sides pattern scores, while later sleep midpoint on the weekend was associated with higher Processed & Fried pattern scores. Similarly, a larger weekend-weekday midpoint difference was associated with higher Processed & Fried pattern scores.