Sleep Disturbances, Obesity, and Cognitive Function in Childhood: A Mediation Analysis.

BACKGROUND: Childhood cognitive development is influenced by biological and environmental factors. One such factor, obesity, impairs cognitive development and is associated with sleep disturbances.
OBJECTIVES: We aimed to examine the mediating role of sleep disturbances on the relation between BMI and cognitive function in children.
METHODS: A total of 9951 children aged 9-10 y were included in this cross-sectional study. Children were recruited from the longitudinal ABCD (Adolescent Brain Cognitive Development) Study. Cognitive development was assessed using metrics for fluid, crystallized, and total cognitive function. Mediation analyses were conducted via linear regression modeling, with adjustment for potential confounders (sex, age, ethnicity, household income, parental education, and self-reported physical activity) for each of the 3 outcomes. Mediation significance was determined by bootstrapping.
RESULTS: A statistically significant inverse association was found between BMI and total (β = -0.41, P < 0.001) and fluid (β = -0.49, P < 0.001) cognition, but not for crystallized cognition. Total sleep disturbances partially mediated the association between BMI and fluid cognition (indirect effect: -0.02, P = 0.002; proportion of the total effect: 0.05, P = 0.002), but no mediation was found in the association between BMI and total cognition.
CONCLUSIONS: Sleep disturbances partially mediate the effect of childhood obesity on cognitive function, particularly in fluid cognitions. Future work is necessary to understand the effects of sleep disturbances and obesity on reduced childhood cognition throughout time, predominantly across the life course.

Introduction

Neurocognitive development is critical in childhood. During this period of life, a dynamic restructuring of the neural structures takes place, where neurons undergo an initial overproduction, and later pruning of the synapses during puberty, resulting in more efficient neural processing (1). Optimal childhood cognitive development allows for the improvement and control of abilities within multiple domains, in addition to the successful foundation of the skills required in adulthood (2, 3). Within these domains, there are abilities associated to biological processes that comprise problem solving, immediate learning, and comprehension tasks, known as fluid cognition. Fluid cognition includes tasks related to executive function, attention, cognitive flexibility, working and episodic memory, and processing speed (4). The ability to utilize information that is acquired through experience, cultural exposure, and education, such as verbal knowledge and language skills, pertains to crystallized cognition (5). Thus, crystallized cognition is associated with long-term abilities, whereas fluid cognitive abilities serve as the building blocks of crystallized cognition.

Moreover, total cognition measures the composite of all neurocognitive abilities, which is important to understand adaptative functioning during childhood and across the life span (6).

Cognition is a developmental process that can be altered by the interaction of biological, genetic, and environmental factors (7–10). One such factor, obesity, has been associated with poorer cognitive performance and altered brain plasticity across the life span (11, 12). Obesity has also been linked to physiological and psychological consequences, including poor motor function, a higher risk of comorbidities (such as type 2 diabetes mellitus, hypertension, dyslipidemia, and obstructive sleep apnea), and emotional and social difficulties among children and adolescents (13–16). In the United States, ∼18.4% of children aged 6–11 y and 20.6% of children and adolescents aged 12–19 y have been classified with obesity (17). Owing to the deleterious effects of obesity on childhood cognitive development and the growing childhood obesity epidemic in the United States and worldwide, understanding the many elements of obesity's impact on cognition is needed.

Furthermore, obesity is associated with the development of sleep disorders (18). Poor sleep has increasingly become a common problem in children (19). Sleep is critical to supporting several biological functions, including cardiovascular, endocrine, and immune processes (20–22). Neural and cognitive processes, such as learning and memorizing, are directly related to healthy sleep patterns (18). Multiple studies have found that poor sleep patterns compromise cognition and mental health, as well as other developmental processes such as emotional control and metabolic regulation (23–25). Previous studies have examined the relation between obesity, obstructive sleep apnea, and executive function and attention, but findings have been inconclusive. This may be due to important study limitations including small sample sizes, lack of adjustment for confounders, as well as the inability to test for interaction and/or mediation effects (26–28).

Considering that previous studies have been limited to focusing on specific sleep disorders and cognitive tasks, this study examines the role of overall sleep disturbances in the association between obesity and 1) total, 2) crystallized, and 3) fluid cognition. Specifically, this study sought to understand the potential mediation effect of sleep disturbances on the obesity–cognition pathway.

Methods

Participants

This cross-sectional analysis included a total of 9951 children aged 9–10 y from the ABCD (Adolescent Brain Cognitive Development) Study, a longitudinal and demographically diverse data set with >11,800 children, recruited from 21 centers across the United States. Children were recruited through the school system, mailing lists at the catchment areas, enrollment of affiliates, referrals, and summer recruitment (29). The sampling scheme included considerations of age, gender, race, socioeconomic status, and urbanicity to reflect the distribution of the US population. The ABCD Study design and recruitment details are described elsewhere (29). For the analysis, children with missing data (n = 1924) and/or implausible values (BMI z score > 5, after reviewing height and weight values; n = 6) were excluded. This study received approval from the Indiana University Institutional Review Board.

Weight status

BMI (in kg/m2) was calculated for each child by dividing their weight in kilograms by their squared height in meters. WHO Anthro-plus was used to calculate the BMI-for-age and gender z scores (SD scores) (30, 31). Following the WHO recommendations, BMI-for-age was then categorized based on the z score cutoffs of <−2.0, >1.0, and >2.0 to classify children as underweight, overweight, and obese, respectively (32).

Sleep disturbances

Sleep disturbances were assessed using the Sleep Disturbance Scale for Children and were reported by parents. The Sleep Disturbance Scale for Children is a questionnaire designed to categorize sleep disorders. It includes 27 items, each rated on a 5-point Likert scale. These 27 items score 5 subdomains that include disorders of initiating and maintaining sleep, disorders of sleep somnolence, sleep breathing disorders, sleep-wake transition disorders, disorders of arousal, and sleep hyperhydrosis (33). The analysis focused on the sleep disturbance total score, which ranges from 26 to 130 points [based on the Sleep Disturbance Scale for Children punctuation system (1–5)], and summed the scores of all subdomains.

Cognitive function

The ABCD Study has a detailed neurocognitive battery that includes visual acuity, handedness, laterality, and neurocognitive development (34, 35). This study was focused on the neurocognitive development area, i.e., the scores of the cognitive tests that comprised the NIH Toolbox. The NIH Toolbox was used to determine cognitive functionality (36, 37). The NIH Toolbox consists of 7 cognitive tasks which cover the 6 cognitive domains: episodic memory (Picture Sequence Memory task), executive function (Flanker Inhibitory Control and Attention, Dimensional Change Card Sort tasks), attention (Flanker Inhibitory Control and Attention task), working memory (List Sorting Working Memory task), processing speed (Pattern Comparison Processing Speed task), and language abilities (Picture Vocabulary, Oral Reading Recognition tasks) (34, 35). The tasks were scored and used to generate 3 composite scores: fluid, crystallized, and total cognitive abilities (6). Fluid cognition includes 5 domains: episodic memory, executive function, attention, working memory, and processing speed; whereas crystallized cognition is particularly focused on language abilities.

For this analysis, the age-corrected cognitive scores available in the ABCD Study data set were used. The age-corrected standard scores, derived from a normative sample of 2917 children and adolescents, account for the performance of individuals in the same age group, and consider the effect of age, as the neurodevelopmental stages change during childhood (38).

Potential confounders

Based on previous literature, several potential confounders were included in the analysis (39, 40). More specifically, sex (female and male), age (as a continuous variable; mo), race [white, black, and other races (including Native American, Pacific Islanders, Asians, and others)], combined household income (categorized as ≤$15,999; $16,000–$34,999; $34,500–$49,999; $50,000–$74,999; $75,000–$99,999; $100,000–$199,999; and ≥$200,000), parental education (categorized as less than high school degree; completed high school/General Educational Development; some college; undergraduate/occupational degree; graduate degree), and self-reported physical activity (as the number of days that the participant was active for ≥60 min/d) were included in the analyses.

Statistical analysis

A mediation analysis was conducted to determine whether sleep disturbances mediate the relation between BMI and cognitive function.  shows a direct acyclic graph of the hypothesized relation. Potential mediation effects were assessed using the multivariate linear regression methods described by Baron and Kenny (41). Modeling included 1) BMI on cognitive function (separately modeled for the total, fluid, and crystallized composite scores); 2) BMI on the sleep disturbances total score; and 3) a regression of the 3 cognitive composite scores on the sleep disturbances total score. In addition, it was determined whether the sleep disturbances total score was a significant mediator of cognitive function in the full model (including BMI). Covariate significance was assessed through a backward elimination stepwise regression. All potential confounders, including age, sex, physical activity, family income, parents’ education, and race, were included as significant covariates in the full-sample and stratified models. Descriptive statistical analyses were performed for these covariates. A chi-square test was applied for the categorical covariates and a one-way ANOVA was applied for the continuous covariates. All analyses were conducted in R version 3.6.0, using the “mediation” package (42, 43). The mediation significance was assessed with the Hayes and Preacher method by bootstrapping (44), with 1000 simulation samples used to generate 95% CIs for the indirect effect [average causal mediator effect (ACME)] and the average direct effect (ADE). Standardized regression coefficients and their corresponding 95% CIs were reported in this study.

FIGURE 1

Direct acyclic graph of the potential mediation effect of sleep disturbances on the association between BMI and cognition composite scores.

Results

A total of 9951 children were included in the analysis.  shows the demographic characteristics of the sample. Mean ± SD age in the overall sample was 9.91 ± 0.62 y, with statistically significant differences between BMI categories. In this analysis, 52.23% of the participants were male and 66.2% of the children in the overall sample were identified as white; however, distribution by BMI categories varied between sex and race (P < 0.01). Children with normal BMI (≤1 SD) had a higher mean number of hours spent in physical activity than children with overweight or obesity (P < 0.01). Statistically significant differences were also found regarding family income and parents’ education between BMI categories (P = 0.69 and P = 0.37, respectively). Compared with children with a normal BMI, children with overweight (>1 SD to ≤2 SD) or obesity (>2 SD) had a statistically significantly higher total score in sleep disturbances. Cognition scores were statistically higher in children with normal BMI than in children with overweight and obesity (P < 0.01).

TABLE 1

Descriptive characteristics of the ABCD (Adolescent Brain Cognitive Development) Study participants

		BMI-for-age SD scores
	Overall (n = 9951)	≤1 SD (n = 6208)	>1 SD to ≤2 SD (n = 2005)	>2 SD (n = 1738)	P value
Age, mo	118.96 ± 7.43	119.12 ± 7.45	118.78 ± 7.43	118.57 ± 7.35	0.01
Physical activity, d	3.55 ± 2.31	3.68 ± 2.31	3.46 ± 2.29	3.20 ± 2.29	<0.01
Sex					<0.01
Male	5197 (52.23)	3178 (51.19)	1047 (52.22)	972 (55.93)
Female	4754 (47.77)	3030 (48.81)	958 (47.78)	766 (44.07)
Race					<0.01
Black	1434 (14.41)	674 (10.86)	329 (16.41)	431 (24.80)
Other white	1928 (19.37)	1132 (18.23)	407 (20.30)	389 (22.38)
White	6589 (66.21)	4402 (70.91)	1269 (63.29)	918 (52.82)
Family income, $					<0.01
≤15,999	960 (9.45)	425 (6.85)	218 (10.87)	317 (18.24)
16,000–34,999	1047 (10.52)	524 (8.44)	244 (12.17)	279 (16.05)
35,000–49,999	838 (8.43)	442 (7.12)	208 (10.37)	188 (10.82)
50,000–74,999	1388 (13.95)	823 (13.26)	287 (14.31)	278 (16.00)
75,000–99,999	1458 (14.65)	930 (14.98)	297 (14.81)	231 (13.29)
100,000–199,999	3087 (31.03)	2167 (34.91)	560 (27.93)	360 (20.71)
≥200,000	1173 (11.79)	897 (14.45)	191 (9.53)	85 (4.89)
Parental education					<0.01
Less than high school	507 (5.09)	224 (3.61)	120 (5.98)	163 (9.38)
High school	941 (9.46)	451 (7.26)	192 (9.58)	298 (17.15)
Some college	1602 (16.10)	886 (14.27)	368 (18.35)	348 (20.02)
College degree	4215 (42.36)	2745 (44.22)	820 (40.90)	650 (37.40)
Graduate degree	2686 (26.99)	1902 (30.64)	505 (25.19)	279 (16.05)
Sleep disturbances total score	36.47 ± 8.04	35.97 ± 7.39	37.17 ± 9.09	37.47 ± 8.82	<0.01
Total cognition score	101.34 ± 17.80	103.20 ± 17.74	99.87 ± 17.46	96.40 ± 17.28	<0.01
Fluid cognition score	96.24 ± 17.27	97.72 ± 17.23	95.10 ± 17.04	92.25 ± 16.96	<0.01
Crystallized cognition score	106.46 ± 18.24	108.07 ± 18.40	105.14 ± 17.85	102.22 ± 17.25	<0.01

Values are mean ± SD or n (%) unless indicated otherwise. P values are for any difference across race (white, African American, and other races) using ANOVA or chi-square test, as appropriate.

This analysis explored the association of BMI with total, fluid, and crystallized cognition, and the effect of sleep disturbances as a mediator.  shows these results. In the preliminary analysis, it was found that higher BMI was positively associated with sleep disturbances. After adjusting for potential confounders, attenuated inverse statistically significant associations of BMI with fluid cognition and total cognition were found. When sleep disturbances total score was included in the model, no mediation effect was found in the associations of BMI with total and crystallized cognition, but a small partial mediation effect of sleep disturbance was found in the association between BMI and fluid cognition. Sleep disturbances had an ACME of −0.02 (P < 0.01) and an ADE of −0.46 (P < 0.01) on the association between BMI and fluid cognition, and the proportion of the total effect caused by sleep disturbances was 0.05 (P < 0.01).  shows the mediation significance of sleep disturbances in the overall sample.

TABLE 2

β Coefficients from the linear regression approach for the mediation effect of sleep disturbances in the association of BMI and cognitive scores

Sample	Cognitive test	Exposure to mediator (β,α)	Exposure to outcome (β,α)	Mediator to outcome (β,α)	Exposure to outcome, adjusted by mediator (β,α)
Total sample	Fluid cognition	0.38 (<0.001)	−0.49 (<0.01)	−0.07 (<0.01)	−0.47 (<0.01)
	Crystallized cognition		−0.20 (0.12)	0.05 (0.01)	−0.21 (0.09)
	Total cognition		−0.41 (<0.01)	−0.01 (0.71)	−0.41 (<0.01)

All models were adjusted for age, sex, physical activity, family income, parents’ education, and race.

FIGURE 2

Sleep disturbances’ mediation significance in the association between BMI and fluid cognition, for the overall sample. ACME, average causal mediation effect/indirect effect; ADE, average direct effect.

Discussion

A statistically significant, although partial, mediation effect of sleep disturbances on the association between BMI and fluid cognition was found. There was no significant mediation for total or crystallized cognition. These results can be explained by the influence of biological and external factors on fluid cognition, which comprise a series of cognitive processes that develop during late childhood. Fluid cognition is sensitive to changes in neurobiological integrity and other factors, such as sleep quality and weight status, which can alter the brain structure and its function (6). Previous studies have shown that obesity was associated with the impairment of inhibition, cognitive flexibility, working memory, and planning. Incremental increases in BMI have been associated with a decline in individual cognitive performance among both children and adults (45–49). Moreover, neuroimaging literature has shown that an increased BMI is associated with reduced cortical thickness in the prefrontal region, which is involved in fluid cognition, and with lower executive function, processes involved in fluid cognition (39, 50). In addition, sleep quality and quantity are also linked to metabolic and endocrine disorders affected by obesity (39, 50–53). The quality and quantity of sleep are fundamental for memory consolidation, and sleep deprivation can affect the brain's reaction time and information processing (36). Poor sleep can negatively affect cognitive processes related to fluid cognition that include learning, cognitive performance, IQ measures, and self-regulatory abilities (54–56). These findings align with the results presented in the current study, suggesting the importance of sleep disturbances for fluid cognition. Moreover, the results in this analysis support and complement the evidence found in a recent study conducted with the ABCD data set, which found that fewer hours of sleep were related to reduced fluid cognition in children (57).

Finally, the results in this study are supported by previous studies that have explored the BMI, sleep, and cognition axis. A previous small-sample study showed that apnea, a specific sleep disturbance, can worsen the effect of obesity on cognitive function, particularly mental processing (58). In addition, other studies have shown that worse sleep behaviors are associated with reduced gray matter volume and cognitive performance (59, 60). One advantage of the present study is that it incorporates several types of sleep disturbances in the analysis, which strengthen the findings from previous studies.

This study has several limitations that may limit the interpretation of our findings. First, the cross-sectional study design does not permit the examination of causal relations owing to the lack of temporality, owing to measurement of the exposure and outcome at the same time point, and not measuring the exposure before the occurrence of the outcome. Second, sleep disturbances were collected through the widely used Sleep Disturbance Scale for Children. Nonetheless, the sleep outcomes might be underestimated because the scale was based on parental report and not clinical measurements. In addition, the use of gold-standard techniques for body composition, such as DXA, was unavailable. However, BMI is a valid surrogate estimate that allows a representation of the population weight distribution in large samples, like the ABCD Study. Furthermore, this analysis was developed with the baseline measurements of the ABCD Study. The baseline sample only included children aged 9–10 y, which limits the generalizability of the results beyond these ages. In addition, nutrient and energy intake measurements were not available in the baseline data set, which limits the potential adjustment for energy and nutrient intakes. Other potential confounders such as caffeine were not included in the study, because the available information was not detailed enough and could potentially have introduced bias into the analysis. Finally, physical activity was self-reported in this data release, which could potentially have created report bias.

This analysis suggests that sleep disturbances partially mediate the effect of childhood obesity on cognitive function, particularly as it pertains to fluid cognition. Future work is needed to understand the effects of sleep disturbances and obesity on reduced childhood cognition over time, predominantly across the life course, and particularly on fluid cognition.

Fluid cognition is critical for neurodevelopment. The fluid abilities developed during childhood support crystallized cognition. During childhood, individuals acquire the knowledge that is needed for long-term memory, message decoding, and comprehension, which are abilities that are fundamental in life. Understanding the effect of sleep disturbance patterns in the association between obesity and fluid cognition in childhood may lead to appropriate interventions to improve the developmental conditions that could enhance neurodevelopment in children.