[Skip to Navigation]
Sign In
September 2010

Shortened Nighttime Sleep Duration in Early Life and Subsequent Childhood Obesity

Author Affiliations

Author Affiliations: Department of Health Services, School of Public Health, University of Washington, Seattle (Dr Bell); and Department of Health Services, University of California, Los Angeles (Dr Zimmerman).

Arch Pediatr Adolesc Med. 2010;164(9):840-845. doi:10.1001/archpediatrics.2010.143

Objective  To test associations between daytime and nighttime sleep duration and subsequent obesity in children and adolescents.

Design  Prospective cohort.

Setting  Panel Survey of Income Dynamics Child Development Supplements (1997 and 2002) from US children.

Participants  Subjects aged 0 to 13 years (n = 1930) at baseline (1997).

Main Exposures  Binary indicators of short daytime and nighttime sleep duration (<25th percentile of age-normalized sleep scores) at baseline.

Main Outcome Measures  Body mass index at follow-up (2002) was converted to age- and sex-specific z scores and trichotomized (normal weight, overweight, obese) using established cut points. Ordered logistic regression was used to model body mass index classification as a function of short daytime and nighttime sleep at baseline and follow-up, and important covariates included socioeconomic status, parents' body mass index, and, for children older than 4 years, body mass index at baseline.

Results  For younger children (aged 0-4 years at baseline), short duration of nighttime sleep at baseline was strongly associated with increased risk of subsequent overweight or obesity (odds ratio = 1.80; 95% confidence interval, 1.16-2.80). For older children (aged 5-13 years), baseline sleep was not associated with subsequent weight status; however, contemporaneous sleep was inversely associated. Daytime sleep had little effect on subsequent obesity in either group.

Conclusions  Shortened sleep duration in early life is a modifiable risk factor with important implications for obesity prevention and treatment. Insufficient nighttime sleep among infants and preschool-aged children may be a lasting risk factor for subsequent obesity. Napping does not appear to be a substitute for nighttime sleep in terms of obesity prevention.

During the last 3 decades, the prevalence of overweight and obesity has increased dramatically among children and adolescents. Obesity—defined as having age- and sex-specific body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) at or above the 95th percentile of national growth standards1—has doubled among children aged 2 to 5 years and adolescents aged 12 to 19 years and has tripled among those aged 6 to 11 years.2,3 In 2003 to 2004, 17% of children and adolescents were obese and 34% were overweight (defined as having BMI >85th and <95th percentiles of the same standards).3

Short sleep duration may increase the risk of obesity in children and adolescents. Evidence is accumulating from cross-sectional population studies to support a robust con temporaneous relationship between shortened sleep duration and unhealthy weight status in children and adolescents.4-15 In several studies, a strong dose-response relationship is evident with increasing odds of overweight/obesity associated with fewer hours spent sleeping.7-9,12

The mechanisms through which sleep operates to influence the balance between energy intake and expenditure are unknown. Shortened sleep duration has been hypothesized to influence weight status through decreased physical activity due to tiredness and increased energy intake given greater opportunity to eat.10,16 An equally compelling pathway may be through influence on the hypothalamic mechanisms that regulate body weight and metabolism via key hormones such as leptin and ghrelin.10 Hunger and appetite increase with lower leptin levels and higher ghrelin levels,10,17 and both low leptin and high ghrelin levels have been linked to short sleep duration in adults17-20; however, less is known about these relationships in children.

To date, most studies of sleep duration and obesity in children and adolescents have been cross-sectional and it is unknown whether the relationship is truly causal. Very few studies have assessed prospective associations of sleep and obesity.21-23 A US sample followed from birth to age 3 years generated 2 studies showing that sleep duration of less than 12 hours in infancy was associated with increased odds of overweight22,23 and with higher BMI z scores and skinfold thickness all measured at age 3 years.23 A third study of 150 children found that children's hours of sleep at ages 3 and 4 years were associated with increased odds of overweight at age 9.5 years. Finally, a study of 900 children in the United Kingdom found that shortened sleep duration at age 3 years was associated with obesity at age 7 years.16 Population-based studies are required to test these associations over longer follow-up periods and in larger samples. Moreover, longitudinal studies are needed to examine the independent effects of daytime and nighttime sleep on obesity as none of the prior studies made this distinction. Although the precise physiological functions of sleep are not fully understood, daytime sleep and nighttime sleep may have independent functions and thus distinct effects on subsequent obesity.10,24,25

The current study uses existing national, longitudinal, panel survey data collected for children and adolescents to test the following: (1) whether poor sleep in early life has lasting effects on obesity measured 5 years later; (2) whether daytime and nighttime sleep have independent effects on subsequent obesity; and (3) whether the relationship between sleep and obesity is different for adolescents and preadolescents.


We conducted a longitudinal analysis of the Panel Survey of Income Dynamics (PSID) Child Development Supplement (CDS) collected for the same children and adolescents in 1997 and again in 2002. Data are in the public domain and the study was approved by the Institutional Review Board of Children's Hospital and Regional Medical Center in Seattle, Washington.

Data source

The PSID is a longitudinal survey directed by the National Science Foundation. Since 1968, 4800 families have been followed. In 1997 (ie, baseline) and 2002 (ie, follow-up), a CDS funded by the National Institute of Child Health and Human Development was administered to the primary caregivers of 3563 children aged 0 to 13 years.26 The CDS questionnaire included detailed demographic data as well as psychological and behavioral assessments of parents and children. Time-use diary data were collected from 1 randomly chosen weekday and 1 randomly chosen weekend day during a school year (September-May). The time diaries recorded both primary and secondary activities for each child during a 24-hour period. Such time diaries have been used extensively in research and have excellent validity when compared with direct observation of activities.27,28


Of the 2569 children in the PSID at follow-up, 990 were aged 0 to 59 months (younger cohort) and 1579 were aged 60 to 154 months (older cohort) at baseline. Children were excluded from the study if they were underweight (n = 24) or had implausibly low BMI (<12) or missing BMI at follow-up or if they had missing time diary data that provide the estimates of sleep time. The total final sample size was 822 children in the younger cohort and 1108 children in the older cohort at baseline measurement. We selected age 5 years as the sample cut point for assignment to the younger vs older cohort to account for differences in sleep patterns between preschool-aged and school-aged children. By age 5 years, the vast majority of children do not take naps.25 Of the older children, 125 were missing BMI at baseline. Accordingly, the models that included baseline BMI were estimated with a sample size of 983. The BMI at baseline was not available for the younger children because height and weight were not assessed in the 1997 survey for children younger than 5 years.

Dependent variable

The dependent variable was BMI in 2002, converted to age- and sex-specific z scores using the 2000 growth charts published by the Centers for Disease Control and Prevention3 and trichotomized based on established cut points: normal weight (BMI ≤85th percentile), overweight (BMI >85th and <95th percentiles), or obese (BMI ≥95th percentile). Both height and weight were measured at follow-up, whereas at baseline the height of children aged 5 years and older was measured and their weight was recorded from parental report.

Independent variables

Family time data were used to calculate the average duration of daytime and nighttime sleep in hours at baseline and at follow-up. Sleeping hours differ with normal growth and development, with younger age associated with longer sleeping hours.25,29,30 Accordingly, we developed age-normalized sleep scores for each child in the data set using external norms derived in a large sample of Swiss children.29 While Swiss children tend to sleep a little more on average than the children in this study, using an external data set has the advantage of avoiding the additional estimation error that would come from using within-sample norms, particularly given the relatively small sample size. No other data sources were found that could be used to compute sleep norms in US children across our study's age range, and no articles describing such norms were found. The age-specific sleep scores were used to develop dichotomous indicators of low sleep for age—identifying those children below the 25th percentile in sleep for age—separately by category of daytime sleep and nighttime sleep. The 25th percentile was chosen to identify children and adolescents who get a low amount of sleep for their age while classifying enough of them as having low sleep to retain adequate statistical power.

More than 90% of children in the sample did not nap at follow-up when all were at least aged 5 years, and most of those who napped did so for a very short duration (<30 minutes). The results were completely robust to the inclusion of napping at follow-up; however, as this variable contained little information and was not significant, it was subsequently dropped from the analysis.


All regression models control for child and family attributes that could affect sleep and obesity. Covariates include child age at baseline (linear and quadratic terms in months), sex (female; male [reference]), race/ethnicity (white, non-Hispanic [reference]; black, non-Hispanic; Hispanic), birth order (only child [reference]; youngest; middle; oldest), family income in 1997 (in dollars, log-transformed), maternal education at baseline (in years, log-transformed), the child's birth weight (in grams), and the mother's and father's BMI (each self-reported in 1999). The parents' BMI is a proxy for both the diet and physical activity patterns in the household as well as genetic factors that might influence the child's BMI. Indicator variables were included for low birth weight (<2500 g), macrosomia (>4500 g), and presence of the father in the household (each coded 1 for yes or 0 for no). Television viewing and other media may disrupt sleep patterns of children and adolescents31; therefore, continuous variables representing the hours per day of entertainment television and noncommercial television (eg, educational programs or DVDs) watched at baseline and at follow-up were included as controls. For the older cohort only, time diary data were used to compute the hours per day in which the child engaged in physical activity at baseline and at follow-up.

For children older than 5 years, the child's baseline BMI z score was controlled to account for the possibility that an association between sleep duration and subsequent obesity reflects an unmeasured tendency of obese children to sleep fewer hours. As height and weight were not measured for the younger cohort at baseline, it was not possible to include baseline BMI in regression models for this age group.

None of the covariates had more than 10% missing data. For each, missing data points were imputed with the sample mean or mode. An alternative approach of dropping observations from the sample if covariates were missing produced similar findings.

Statistical analysis

All statistical analysis was conducted with Stata version 10.1 statistical software (StataCorp LP, College Station, Texas) and developed separately for the younger and older cohorts. Two models were estimated in the older cohort, one including baseline BMI and one without this variable. Results from the 2 models were compared to assess the contribution of baseline BMI with the estimated association between sleep duration and subsequent BMI. This comparison allows us to gauge whether missing baseline BMI in the younger cohort is likely to affect the results. Because the dependent variable included 3 ordered categories, ordered logistic regression32 was used to model child weight status at follow-up as a function of sleep duration at baseline and at follow-up, controlling for all covariates and, for children older than 5 years, the child's baseline BMI. The assumption of proportional odds required for ordered logistic regression was tested formally. In all models, sampling weights were used to account for the complex sampling design and to allow inferences valid for the population.


The characteristics of the study sample are summarized in Table 1. At follow-up, 33% of the younger cohort and 36% of the older cohort were overweight or obese. On average, the younger cohort slept approximately 10 hours per night in 1997 and 2002 and napped for about an hour each day at baseline. Sleep duration in the older cohort averaged 9.7 hours per night at baseline, 9.2 hours per night at follow-up, and only 12 minutes of daytime sleep at baseline.

Table 1. 
Sample Characteristics
Sample Characteristics

The sleep variables were not highly correlated (results not shown), with respective correlation coefficients in the younger and older cohorts of r = −0.23 and −0.14 for baseline napping and baseline nighttime sleep, 0.08 and 0.02 for baseline napping and nighttime sleep at follow-up, and 0.08 and 0.21 for nighttime sleep at baseline and follow-up.

Figure 1 and Figure 2 describe the cutoff points used to identify low sleep (defined as <25th percentile) in the younger and older cohorts, respectively. A χ2 test statistic failed to reject the null hypothesis of parallel odds, with P values of .45 for the younger cohort and .70 for the older cohort. Further investigation confirmed that the sleep variables did not individually violate the assumption of parallel odds. Therefore, results are reported for the ordered logit models in Table 2. Because the parallel odds assumption is met, the ordered logit results can be interpreted as the odds of a shift from normal weight to overweight or as the odds of a shift from overweight to obesity.

Figure 1. 
Low sleep duration (<25th percentile) cutoff points in the younger cohort.

Low sleep duration (<25th percentile) cutoff points in the younger cohort.

Figure 2. 
Low sleep duration (<25th percentile) cutoff points in the older cohort.

Low sleep duration (<25th percentile) cutoff points in the older cohort.

Table 2. 
Ordered Logistic Regression Models of Weight Classification at Follow-upa
Ordered Logistic Regression Models of Weight Classification at Follow-upa

Younger cohort

For the younger children (Table 2, model 1), low nighttime sleep at baseline was significantly associated with increased odds of overweight (vs normal weight) and increased odds of obesity (vs overweight) at follow-up (odds ratio = 1.80; 95% confidence interval, 1.16-2.80). Daytime sleep at baseline and nighttime sleep at follow-up were not associated with subsequent obesity in this cohort.

Older cohort

In the older cohort (Table 2, model 2), low nighttime sleep at baseline was not associated with any shift in weight status at follow-up. Contemporaneous sleep, however, was associated with increased odds of a shift from normal weight to overweight or from overweight to obesity at follow-up (odds ratio = 1.80; 95% confidence interval, 1.16-2.81). The association of nighttime sleep at follow-up and obesity was attenuated with the inclusion of baseline BMI, while the estimate for nighttime sleep at baseline remained unchanged (Table 2, models 2 and 3).


In a national sample, we found a robust longitudinal association between duration of nighttime sleep in early life and subsequent obesity measured at ages 5 to 9 years. These findings persisted when controlling for contemporaneous sleep and other important confounding variables including parents' BMI, family socioeconomic status (parental education, income, single parent), and hours of television viewed. For the younger cohort, duration of sleep at baseline was more closely associated with obesity at follow-up than was duration of nighttime sleep at follow-up. In the older cohort, nighttime sleep at follow-up was associated with marginally increased odds of obesity at follow-up, while sleep duration 5 years prior had no significant effect. These findings suggest that there is a critical window prior to age 5 years when nighttime sleep may be important for subsequent obesity status.

Our findings also indicate that daytime sleep had little effect on subsequent obesity at any age. This result suggests that napping is not a substitute for nighttime sleep. There is some evidence that nighttime sleep and naps serve different physiological functions. Naps may reduce daytime psychosocial stress, increase attention span, and increase alertness for learning, while nighttime sleep involves complex biological, psychosocial, and restorative functions.10,24,25 Problem napping and disruptive behaviors are associated with higher cortisol levels and shorter nap duration.24

Interestingly, some variables associated with obesity in other work,2 notably race/ethnicity, were not significant in our models. Significant racial/ethnic disparities have been reported in nighttime sleep duration.33 Consistent with our results, sleep duration may be a plausible contributor to well-documented racial/ethnic disparities in obesity.

Study strengths include a large, nationally representative sample and data collected at 2 points 5 years apart, which permit assessment of causal ordering in the relationship between sleep and obesity. Interpretation of the study findings is, however, subject to limitations. The PSID CDS collects sleep data for only 2 days in a year—a random weekday and a random weekend day. With only 2 measurements, the reported sleep patterns may not reflect the children's usual sleep patterns throughout the year. The resulting measurement error may introduce a conservative bias into the analyses. Second, BMI at baseline was not measured for the younger children. We did control for parents' BMI, which is positively associated with child's BMI34 and is expected to capture much of the variance associated with the tendency of children to be overweight at such a young age. More importantly, the estimates of the contribution of nighttime sleep at baseline in the older cohort were similar with and without baseline BMI, suggesting that the exclusion of baseline BMI in the younger cohort would be unlikely to alter the main study findings. Third, physical activity and diet are potential confounders of the associations of sleep with obesity. Few variables are available in the PSID to measure diet, and those available were of insufficient quality to be included in the regression models. Our results were robust to the inclusion of the number of hours per day the child engaged in physical activity. Furthermore, inclusion of parents' BMI in the models serves as a proxy for the diet and physical activity patterns in the household. Fourth, weight for children at baseline was obtained by parent report, which could lead to misclassification, especially for the older children in the cohort for whom parents may minimize problems with overweight.35,36 Accordingly, the associations we report may be conservative.

Sleep duration is a modifiable risk factor with potentially important implications for obesity prevention and treatment. Insufficient nighttime sleep among infants and preschool-aged children appears to be a lasting risk factor for subsequent obesity, while contemporaneous sleep appears to be important to weight status in adolescents. Napping had no effects on the development of obesity and is not a substitute for sufficient nighttime sleep.

Correspondence: Janice F. Bell, PhD, MPH, Department of Health Services, School of Public Health, University of Washington, Box 357230, Seattle, WA 98198-7230 (jfbell@u.washington.edu).

Accepted for Publication: February 9, 2010.

Author Contributions:Study concept and design: Bell and Zimmerman. Acquisition of data: Zimmerman. Analysis and interpretation of data: Bell and Zimmerman. Drafting of the manuscript: Bell and Zimmerman. Critical revision of the manuscript for important intellectual content: Bell. Statistical analysis: Bell and Zimmerman. Obtained funding: Bell. Study supervision: Bell.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grant 1R40MC08965-01-00 from the Health Resources and Services Administration/Maternal and Child Health Bureau (Dr Bell).

Kuczmarski  RJOgden  CLGuo  SS  et al.  2000 CDC Growth Charts for the United States: methods and development.  Vital Health Stat 11 2002; (246) 1- 190Http://Www.Cdc.Gov/Growthcharts. Accessed February 4, 2010PubMedGoogle Scholar
Koplan  JPedLiverman  CTedKraak  VIed Preventing Childhood Obesity: Health in the Balance.  Washington, DC National Academies Press2005;
Ogden  CLFlegal  KMCarroll  MDJohnson  CL Prevalence and trends in overweight among US children and adolescents, 1999-2000.  JAMA 2002;288 (14) 1728- 1732PubMedGoogle ScholarCrossref
Locard  EMamelle  NBillette  AMiginiac  MMunoz  FRey  S Risk factors of obesity in a five year old population: parental vs environmental factors.  Int J Obes Relat Metab Disord 1992;16 (10) 721- 729PubMedGoogle Scholar
Kagamimori  SYamagami  TSokejima  S  et al.  The relationship between lifestyle, social characteristics and obesity in 3-year-old Japanese children.  Child Care Health Dev 1999;25 (3) 235- 247PubMedGoogle ScholarCrossref
Vioque  JTorres  AQuiles  J Time spent watching television, sleep duration and obesity in adults living in Valencia, Spain.  Int J Obes Relat Metab Disord 2000;24 (12) 1683- 1688PubMedGoogle ScholarCrossref
von Kries  RToschke  AMWurmser  HSauerwald  TKoletzko  B Reduced risk for overweight and obesity in 5- and 6-y-old children by duration of sleep: a cross-sectional study.  Int J Obes Relat Metab Disord 2002;26 (5) 710- 716PubMedGoogle ScholarCrossref
Sekine  MYamagami  THanda  K  et al.  A dose-response relationship between short sleeping hours and childhood obesity: results of the Toyama Birth Cohort Study.  Child Care Health Dev 2002;28 (2) 163- 170PubMedGoogle ScholarCrossref
Chaput  JPBrunet  MTremblay  A Relationship between short sleeping hours and childhood overweight/obesity: results from the ‘Quebec en Forme’ Project.  Int J Obes (Lond) 2006;30 (7) 1080- 1085PubMedGoogle ScholarCrossref
Taheri  S The link between short sleep duration and obesity: we should recommend more sleep to prevent obesity.  Arch Dis Child 2006;91 (11) 881- 884PubMedGoogle ScholarCrossref
Gupta  NKMueller  WHChan  WMeininger  JC Is obesity associated with poor sleep quality in adolescents?  Am J Hum Biol 2002;14 (6) 762- 768PubMedGoogle ScholarCrossref
Padez  CMourao  IMoreira  PRosado  V Prevalence and risk factors for overweight and obesity in Portuguese children.  Acta Paediatr 2005;94 (11) 1550- 1557PubMedGoogle ScholarCrossref
Knutson  KL Sex differences in the association between sleep and body mass index in adolescents.  J Pediatr 2005;147 (6) 830- 834PubMedGoogle ScholarCrossref
Chen  MYWang  EKJeng  YJ Adequate sleep among adolescents is positively associated with health status and health-related behaviors.  BMC Public Health 2006;659PubMedGoogle ScholarCrossref
Jiang  FZhu  SYan  CJin  XBandla  HShen  X Sleep and obesity in preschool children.  J Pediatr 2009;154 (6) 814- 818PubMedGoogle ScholarCrossref
Reilly  JJArmstrong  JDorosty  AR  et al. Avon Longitudinal Study of Parents and Children Study Team, Early life risk factors for obesity in childhood: cohort study.  BMJ 2005;330 (7504) 1357PubMedGoogle ScholarCrossref
Spiegel  KTasali  EPenev  PVan Cauter  E Brief communication: sleep curtailment in healthy young men is associated with decreased leptin levels, elevated ghrelin levels, and increased hunger and appetite.  Ann Intern Med 2004;141 (11) 846- 850PubMedGoogle ScholarCrossref
Taheri  SLin  LAustin  DYoung  TMignot  E Short sleep duration is associated with reduced leptin, elevated ghrelin, and increased body mass index.  PLoS Med 2004;1 (3) e62PubMedGoogle ScholarCrossref
Spiegel  KKnutson  KLeproult  RTasali  EVan Cauter  E Sleep loss: a novel risk factor for insulin resistance and type 2 diabetes.  J Appl Physiol 2005;99 (5) 2008- 2019PubMedGoogle ScholarCrossref
Spiegel  KLeproult  RL'Hermite-Baleriaux  MCopinschi  GPenev  PDVan Cauter  E Leptin levels are dependent on sleep duration: relationships with sympathovagal balance, carbohydrate regulation, cortisol, and thyrotropin.  J Clin Endocrinol Metab 2004;89 (11) 5762- 5771PubMedGoogle ScholarCrossref
Agras  WSHammer  LD McNicholas  FKraemer  HC Risk factors for childhood overweight: a prospective study from birth to 9.5 years.  J Pediatr 2004;145 (1) 20- 25PubMedGoogle ScholarCrossref
Gillman  MWRifas-Shiman  SLKleinman  KOken  ERich-Edwards  JWTaveras  EM Developmental origins of childhood overweight: potential public health impact.  Obesity (Silver Spring) 2008;16 (7) 1651- 1656PubMedGoogle ScholarCrossref
Taveras  EMRifas-Shiman  SLOken  EGunderson  EPGillman  MW Short sleep duration in infancy and risk of childhood overweight.  Arch Pediatr Adolesc Med 2008;162 (4) 305- 311PubMedGoogle ScholarCrossref
Ward  TMGay  CAlkon  AAnders  TFLee  KA Nocturnal sleep and daytime nap behaviors in relation to salivary cortisol levels and temperament in preschool-age children attending child care.  Biol Res Nurs 2008;9 (3) 244- 253PubMedGoogle ScholarCrossref
Weissbluth  M Healthy Sleep Habits, Happy Child.  New York, NY Ballantine Books2003;
Mainieri  T The Panel Study of Income Dynamics Child Development Supplement: User Guide for CDS-II.  Ann Arbor, MI Institute for Social Research2006;
Juster  FTOno  HStafford  FP An assessment of alternative measures of time use.  Sociol Methodol 2003;3319- 54Google ScholarCrossref
Vandewater  EABickham  DSLee  JH Time well spent? relating television use to children's free-time activities.  Pediatrics 2006;117 (2) e181- e191PubMedGoogle ScholarCrossref
Iglowstein  IJenni  OGMolinari  LLargo  RH Sleep duration from infancy to adolescence: reference values and generational trends.  Pediatrics 2003;111 (2) 302- 307PubMedGoogle ScholarCrossref
Bayer  ORosario  ASWabitsch  Mvon Kries  R Sleep duration and obesity in children: is the association dependent on age and choice of the outcome parameter?  Sleep 2009;32 (9) 1183- 1189PubMedGoogle Scholar
Zimmerman  FJ Children's Media Use and Sleep Problems: Issues and Unanswered Questions.  Menlo Park, CA Kaiser Family Foundation2008;
Greene  WH Econometric Analysis. 4th ed. Upper Saddle River, NJ Prentice-Hall2000;
Crosby  BLeBourgeois  MKHarsh  J Racial differences in reported napping and nocturnal sleep in 2- to 8-year-old children.  Pediatrics 2005;115 (1) ((suppl)) 225- 232PubMedGoogle ScholarCrossref
Davis  MM McGonagle  KSchoeni  RFStafford  F Grandparental and parental obesity influences on childhood overweight: implications for primary care practice.  J Am Board Fam Med 2008;21 (6) 549- 554PubMedGoogle ScholarCrossref
Parry  LLNetuveli  GParry  JSaxena  S A systematic review of parental perception of overweight status in children.  J Ambul Care Manage 2008;31 (3) 253- 268PubMedGoogle ScholarCrossref
Maynard  LMGaluska  DABlanck  HMSerdula  MK Maternal perceptions of weight status of children.  Pediatrics 2003;111 (5, pt 2) 1226- 1231PubMedGoogle Scholar