Sleep Duration and Hyperglycemia Among Obese and Nonobese Children Aged 3 to 6 Years

Objective  To investigate the association between sleep duration and risk of hyperglycemia among preschool Chinese children.

Design  A population-based cross-sectional study.

Setting  Seventy-one randomly selected kindergartens in Tianjin, China.

Participants  Six hundred nineteen obese (body mass index z score ≥1.65) and 617 nonobese (body mass index z score <1.65) children aged 3 to 6 years were recruited and matched by age.

Main Exposure  Sleep duration.

Main Outcome Measures  Hyperglycemia, defined as a fasting glucose level of 100 mg/dL or higher.

Results  Obese children were more likely to have shorter sleep duration (≤8 hours) compared with their nonobese counterparts (P < .001). Compared with those who slept for 9 or 10 hours per night, those who slept for 8 hours or less had a significantly higher likelihood of having hyperglycemia, controlling for age and sex (odds ratio [OR], 1.65; 95% confidence interval [CI], 1.12-2.45). After further adjustment for other potential confounders, the association still remained statistically significant (OR, 1.64; 95% CI, 1.09-2.46). In the stratified multivariable analyses, those who were obese and slept for 8 hours or less had an increased risk of having hyperglycemia (OR, 2.12; 95% CI, 1.06-4.21) compared with those who were nonobese and slept for 9 hours or more.

Conclusions  Shorter sleep duration is associated with an increased risk of having hyperglycemia among preschool Chinese children. Whether adequate sleep may help maintain euglycemia among children, especially for those who are overweight or obese, warrants further investigation.

The increasing prevalence of obesity and related cardiometabolic diseases, including cardiovascular disease and type 2 diabetes mellitus (T2D), is a critical health concern worldwide.¹ Childhood obesity is associated with being overweight or obese in adolescence and adulthood.² Importantly, obese children are more likely to have impaired glucose tolerance, a prediabetic phase, during childhood³ and increased risks for cardiovascular morbidity and mortality in adulthood.^4,5

Accumulative evidence indicates that sleep curtailment plays a role in the pathogenesis of both obesity and T2D.^6,7 There is a great body of evidence that short sleep duration is associated with both childhood^8,9 and adulthood⁹ obesity. Observational studies suggested that both short- and long-duration sleepers were associated with increased risks of T2D development among adults.^10,11 It has been argued that the sleep loss–associated progress of obesity and T2D may result from sleepiness-related physical inactivity and excess energy intake rather than direct physiological changes.¹² However, results from experimental studies⁶ have shown that sleep deprivation increases appetite and reduces glucose uptake in the brain, leading to excess energy intake and insulin resistance. A study by Nixon et al¹³ reported that certain environmental factors have a profound impact on variation of sleep duration in children. So far, data on how sleep duration is associated with glycemic status among preschool children are scarce. Furthermore, it remains unknown whether adiposity status and known environmental factors related to the risk of obesity and T2D may mediate the association of sleep duration and glycemic status. The aim of this study was to investigate the association between sleep duration and hyperglycemia among obese and nonobese Chinese children aged 3 to 6 years. Additionally, the effect modifications of adiposity on this association were assessed by controlling for children's anthropometric indexes, blood pressure, infant feeding modalities, current disease status, habitual food intake, physical activity, and parents' socioeconomic status and body mass index (BMI).

Tianjin, with a population of 11.5 million, is the fourth largest city in China and enjoys a provincial status. A cross-sectional survey was carried out in 71 kindergartens in Tianjin from March to September 2005. The city has 18 county-level administrative areas, including 9 urban districts and 9 counties, that govern towns and rural areas. There were a total of 269 kindergartens in Tianjin. A multistage cluster sampling was used to obtain a random sample of children aged 3 to 6 years in Tianjin. About 20% of kindergartens (29 kindergartens of 151) were randomly selected from 9 urban districts, and about 35% (42 kindergartens of 118) were randomly selected from 9 rural areas because of their small sizes. All children aged 3 to 6 years in the selected kindergartens were invited to participate in the survey. A total of 15 928 children completed the survey, with a participation rate of 95.6%.

Of them, an age-matched case-control study was further conducted to investigate neonatal and postnatal factors associated with childhood obesity and obesity-related metabolic abnormalities. We used the World Health Organization child growth reference¹⁴ to define obesity and nonobesity. The BMI z score cutoff point we used was 1.65, which is the 95th percentile of age- and sex-specific distribution. The cases and controls were those with a BMI z score of 1.65 or higher (n = 1258) and their age-matched counterparts with a BMI z score less than 1.65, respectively. The matching was undertaken in each sampled kindergarten. No participant reported a diagnosis of diabetes mellitus and/or current use of any hypoglycemic medication. In total, 1282 children (632 cases and 650 controls) were successfully recruited for the study and agreed to provide peripheral blood samples. Forty-six children were subsequently excluded from the present analysis because of missing data on waist circumference (n = 21) or having the following diseases diagnosed by a physician and/or use of the corresponding medications during the past 30 days: hepatitis (n = 1), tuberculosis (n = 1), asthma (n = 11), and chronic bronchitis (n = 12). The final sample in the present analysis comprised 619 obese and 617 nonobese children. The study was approved by the institutional review board of the Tianjin Women and Children's Health Center.

A self-administered questionnaire was given to the children's parents to be completed at home. The questionnaire included questions on the child's birth date, birth weight and length, and gestational age (for questions related to birth, parents were asked to copy from the child's birth certificate); history of illness status; current health status; feeding modalities during infancy; current health behaviors including habitual food intake, duration of usual sleep, television viewing, and physical activity; and parents' sociodemographic factors, body weight and height, and medical history. Health workers, who were from the Women and Children's Health Center at city and local district levels, conducted the survey. All health workers were intensively trained in meetings and in practical sessions. Trained health workers checked the questionnaire at the kindergartens for uncertain and incomplete questions. All participants were asked to fast overnight. They underwent a physical examination and provided a 0.5-mL peripheral blood sample from the middle finger between 7:30 AM to 8:30 AM at the kindergarten clinic unit.

To determine usual sleep duration, the question “How many hours of sleep does this child usually get?” was asked. The parents were asked to choose 1 of the following options: less than 8 hours, 8 hours, 9 hours, 10 hours, 11 hours, or 12 or more hours. Parents' educational attainment was categorized into 3 groups: 9 or fewer years, 10 to 12 years, and 13 or more years. Based on the responses to relevant questions, variables on breastfeeding at age 6 months, complementary food introduction before age 6 months, sweetened beverage consumption more than 500 mL/wk, high-fat meat intake, and everyday intake of vegetables and fruit were created and dichotomized as yes or no. Disease status was classified as yes for 13 participants who reported having pneumonia, cold, or fever during the past 30 days. Duration of television viewing and any type of physical activity was recorded and grouped into 3 categories: less than 30 min/d, 30 to 59 min/d, and 60 or more min/d.

Body weight was measured with a beam-balance scale with subjects wearing light indoor clothing without shoes. Body height was measured by a stadiometer. Weight was measured to the nearest 0.1 kg and height to the nearest 0.1 cm. Body mass index was calculated as weight in kilograms divided by height in meters squared. Body mass index and weight-for-age and height-for-age z scores were calculated with the World Health Organization child growth reference.¹⁴ Waist circumference was measured at the level of the umbilicus to the nearest 0.1 cm.

Blood pressure was measured using a standardized mercury sphygmomanometer with a cuff bladder width of 8 cm. The fourth Korotkoff sound was adopted for diastolic blood pressure recording. The measurement was taken on the right arm of the participant in a comfortable sitting position after at least 5 minutes' rest. Mean blood pressure was calculated from 2 readings unless the difference between these readings was greater than 10 mm Hg, in which case a third measurement was taken and the mean of the last 2 measurements was used.

Serum glucose level was measured enzymatically on an automatic analyzer (RX Daytona; Randox Laboratories Ltd, Antrim, Ireland) with reagents purchased from the manufacturer. There were 11 children (7 boys and 4 girls) whose fasting glucose level was 126 mg/dL or higher (to convert to millimoles per liter, multiply by 0.0555). Hyperglycemia was defined as a fasting glucose level of 100 mg/dL or higher (138 children; 101 boys and 37 girls).¹⁵

A general linear model for continuous variables and a logistical regression model for categorical variables were used for the comparison between obese and nonobese children, appropriately adjusted for sex. For BMI and weight-for-age and height-for-age z scores, the results were confirmed with nonparametric analysis but only parametric analyses are reported. Multiple linear regression models were used to evaluate the associations of anthropometric indexes and sleep duration with fasting glucose level. Multivariable logistic regression models were used to assess the associations of obesity status and sleep duration with the risk of hyperglycemia. The analyses were performed with adjustment for age and sex and then further for other potential confounding factors. When the analyses were performed using the age- and sex-matched subsample (n = 1096 [458 girls and 638 boys]), excluding those having disease during the past month (n = 13) and defining hyperglycemia as a fasting glucose level of 110 mg/dL or higher, a similar pattern of results was observed. We therefore report the results for the original analyses. The statistical inference was made when P < .05 (2-sided). All statistical analyses were performed with SAS version 9.1 (SAS Institute, Cary, North Carolina).

Compared with nonobese children, anthropometric indexes, including birth weight, body weight and height, BMI, waist circumference, and weight-for-age, height-for-age, and BMI z scores, were significantly higher among obese children (all P < .001) (Table 1). Parents of obese children had higher BMIs (all P < .001), and the fathers were less likely to be highly educated (P = .03). Obese children were more likely to have shorter sleep duration, drink more than 500 mL of sweetened beverages per week, and have longer duration of television viewing (all P < .001). Meanwhile, obese children had significantly higher systolic and diastolic blood pressure and fasting glucose levels (all P < .01).

Sleep duration was inversely associated with fasting glucose level (Table 2). There was no significant association of anthropometric indexes with fasting glucose level and risk of hyperglycemia (Table 2 and Table 3). However, compared with those who slept for 9 or 10 hours per night, those who slept for 8 hours or less had a significantly higher likelihood of having hyperglycemia, controlling for age and sex (odds ratio [OR], 1.65; 95% confidence interval [CI], 1.12-2.45). The association was only slightly attenuated (OR, 1.64; 95% CI, 1.09-2.46) after further adjustment for birth weight; gestational age; BMI; systolic blood pressure; parents' educational status and BMI; breastfeeding at age 6 months; timing of complementary food introduction; self-reported disease during the past month; sweetened beverage consumption; intake of vegetables, fruit, and high-fat meat; and duration of television viewing and physical activity. Additionally adjusted for waist circumference, the association still remained statistically significant (OR, 1.64; 95% CI, 1.09-2.46). When the analysis was stratified according to obesity status, the significant association was only observed among obese children but not nonobese children. The ORs were 2.05 (95% CI, 1.19-3.55) for the simple adjusted analysis (model 1) and 2.22 (95% CI, 1.24-3.95; model 2) and 2.15 (95% CI, 1.20-3.84; model 3) for the multivariable-adjusted analyses.

Among obese children, the P value for those who slept for more than 11 hours was .03 in the full-model analysis (model 3), indicating a possibly monotonic association between sleep duration and risk of hyperglycemia. In this regard, we selected those who slept for 8 hours or less as the reference group (OR, 1) (Figure). In the whole sample, the multivariable-adjusted ORs (model 3) for hyperglycemia were 0.61 (95% CI, 0.41-0.92) and 0.47 (95% CI, 0.26-0.86) for those who slept for 9 or 10 hours and 11 hours or more, respectively (P for trend = .003). Among obese children, the corresponding ORs were 0.47 (95% CI, 0.26-0.83) and 0.23 (95% CI, 0.08-0.63), respectively (P for trend <.001). The association was not significant among nonobese children. When multivariable analyses were performed according to obesity status (obese/nonobese) and sleep duration (≤8 hours/≥9 hours), those who were obese and slept for 8 hours or less had a significantly higher likelihood of having hyperglycemia (OR, 2.12; 95% CI, 1.06-4.21) compared with those who were nonobese and slept for 9 hours or more (Table 4). The P value for interaction (χ²₁ = 2.4434; P = .12), however, did not reach statistical significance.

We found that shorter sleep duration was associated with increased risks of having hyperglycemia among Chinese children as young as 3 to 6 years of age. This association was more apparent for obese children than their nonobese counterparts. For those who were obese, the inverse association between sleep duration and risk of hyperglycemia seemed to be monotonic. To our knowledge, this is the first study reporting the association between sleep duration and glycemic status among preschool children. Importantly, this association was not able to be appreciably attenuated by children's anthropometric indexes, blood pressure, infant feeding modalities, habitual food intake, physical activity, and parents' socioeconomic status and BMI.

Few studies have examined the association between sleep duration and glycemic status among children. A study by Flint et al¹⁶ reported that, among 40 obese children, shorter sleep duration was associated with higher levels of insulin resistance indexes, including fasting insulin level and homeostatic assessment model algorithm for insulin resistance. Among adults, both short and long sleep durations are associated with increased risks of T2D^9,10,15-17 or impaired glucose tolerance.^17,18 A recent cross-sectional analysis carried out among middle-aged Finnish men and women found that short (≤6 hours) or long (≥8 hours) sleep duration was associated with increased risks of T2D among women.¹⁹ Among 1486 individuals who participated in the Sleep Heart Health Study, a sleep duration of 6 hours or less or 9 hours or more was associated with increased prevalence of diabetes mellitus and impaired glucose tolerance.¹⁷ Similar findings were observed among 940 participants of the Quebec Family Study.¹⁸ Based on data obtained from the Nurse Health Study, Ayas et al¹⁰ evaluated the relationship between sleep duration and the 10-year risk of diabetes development. In multivariable analyses, short- or long-duration sleepers were associated with increased risks of diabetes diagnosis. After further adjustment for BMI, the significant relationship remained for long-duration sleepers only. However, when the analyses were performed for symptomatic diabetes, modestly increased risks were observed for both short- and long-duration sleepers, indicating an independent effect of sleep restriction on developing symptomatic diabetes. In the Massachusetts Male Aging Study,¹¹ 1139 middle-aged and elderly men were followed up for 15 years. A U-shaped association between self-reported sleep time and risk of diabetes development was observed with short (≤6 hours) or long (≥8 hours) sleep duration. In a random sample of 2663 middle-aged Swedish men and women, men who slept for 5 hours or less per night at baseline had increased incident diabetes during a 12-year follow-up.²⁰ Along with conventional diabetes risk factors, these studies examined and controlled for certain sleep complaints, including indexes of sleep apnea^17,19 and hypopnea,¹⁷ snoring, and difficulties initiating and maintaining sleep,²⁰ suggesting an independent effect of sleep loss on the risk of diabetes development.

Although the physiological function of sleep has not been well understood, experimental studies have suggested that sleep restriction may induce altered glucose metabolism and reduced insulin sensitivity.^6,7 Data from laboratory studies suggest that the sleep-debt condition (4 hours per night for 6 days) induces impaired glucose tolerance and insulin resistance among healthy young men.²¹ During intravenous glucose infusion, glucose effectiveness, which measures the ability of glucose to mediate its own disposal independently of insulin, is reduced remarkably compared with that in the sleep-recovery period (12 hours per night for 7 days).²¹ The brain is the major site of non–insulin-dependent glucose uptake.⁶ Reduced brain glucose uptake during sleep restriction may result in increased glucose concentrations in peripheral tissues.^21,22 Our findings observed among obese children are in line with these results. In addition, partial sleep deprivation may upregulate the activation of the hypothalamic-pituitary axis, leading to disturbances in the secretory profiles of the counterregulatory hormones, growth hormone and cortisol.²¹ The changed hormonal profile may also contribute to the altered glucose metabolism.^22,23 Meanwhile, sleep loss is associated with increased activity of the sympathetic nervous system, resulting in downregulated pancreatic beta-cell function and insulin release.^6,21 Moreover, modest sleep deprivation is associated with elevated proinflammatory cytokine levels, ie, interleukin 6 and tumor necrosis factor α,²⁴ that stimulate the hepatic synthesis of C-reactive protein, an acute-phase reactant and a sensitive indicator of low-grade systemic inflammation, resulting in the progress of insulin resistance.²⁵

In this study, we did not observe a significant association between childhood obesity and risk of having hyperglycemia. However, the increased risk of having hyperglycemia among short-duration sleepers seemed to be more pronounced among obese rather than nonobese children, indicating that obesity might play a role in the altered glucose metabolism. Adipose tissue dysfunction, ie, the altered systemic balance of inflammatory factors and adipocytokines, is an important scenario in the etiology of T2D.²⁶ The physiological consequence of the unbalance-induced alteration may result from a long-term progression. Spiegel et al²¹ pointed out that predisposed individuals may be more vulnerable to the sleep loss–altered glucose metabolism leading to insulin resistance. Given that our study participants were so young, we hypothesize that, among our study participants, the modifying effect of obesity on the observed association might result from a combined effect of chronic sleep curtailment with obesity.

Our analyses are based on a population-based case-control sample from kindergartens from all 18 administrative districts of Tianjin. Hence, it is unlikely that our findings were obtained by chance. In the multivariable analyses, we simultaneously controlled for a variety of previously reported covariates^27-29 related to both childhood obesity and possibly altered glucose metabolism. Notwithstanding, we were not able to make a causal inference given the cross-sectional nature of this study design. Because of the limited numbers of girls who had hyperglycemia, sex-stratified analyses could not be performed because of lack of statistical power. In our study, parent-reported sleep duration was adopted for the association assessment, which may be subject to recall bias. The use of an 8-cm cuff for blood pressure measurement might underestimate blood pressure for some children 3 years of age. The parents' BMIs were calculated based on self-reported body weight and height. Individuals who are overweight or obese are more likely to underreport their body weight. These pitfalls may lead to the underestimation of the association. Moreover, in this study, we did not measure the presence of other sleep disorders, eg, sleep apnea, hypopnea, and snoring, that might confound the observed association between sleep duration and hyperglycemia.

In conclusion, our findings suggest that shorter sleep duration (≤8 hours) is associated with an increased risk of having hyperglycemia among Chinese children aged 3 to 6 years. Obese children are more likely to have shorter sleep duration. The inverse association between sleep time and risk of hyperglycemia appears to be more pronounced as well as monotonic for obese children. Appropriately designed prospective studies or intervention trials are warranted to evaluate the effect of adequate sleep on maintaining euglycemia among children. To avoid the adverse influence of a lifelong behavior on the development of obesity and related metabolic disorders in children, an adequate sleep duration should be promoted along with a prudent diet and physical activity as the keys to a healthy lifestyle. This may be particularly important for children who are overweight or obese.

Correspondence: Zhijie Yu, MD, PhD, MPH, Key Laboratory of Nutrition and Metabolism, Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200031, China (zjyu@sibs.ac.cn).

Accepted for Publication: July 10, 2009.

Author Contributions: The authors had full access to the data in this study and take complete responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Ye, W. Wang, P. Wang, G. Liu, Hu, and Yu. Acquisition of data: Tian, Ye, Zhang, E. Liu, W. Wang, P. Wang, G. Liu, and Yu. Analysis and interpretation of data: Tian, Ye, Zhang, E. Liu, G. Liu, Yang, Hu, and Yu. Drafting of the manuscript: Tian, W. Wang, P. Wang, and Yu. Critical revision of the manuscript for important intellectual content: Ye, Zhang, E. Liu, G. Liu, Yang, Hu, and Yu. Statistical analysis: Tian, Ye, Zhang, E. Liu, W. Wang, P. Wang, G. Liu, Yang, and Yu. Obtained funding: Ye, Zhang, E. Liu, P. Wang, and G. Liu. Administrative, technical, and material support: Tian, Ye, W. Wang, and G. Liu. Study supervision: Ye, G. Liu, Hu, and Yu.

Financial Disclosure: None reported.

Funding/Support: This study was financially supported by the Tianjin Women and Children's Health Center.

Additional Contributions: We thank all study participants and health workers who participated in the study for their excellent cooperation.