OBJECTIVE. Associations between short sleep duration and increased BMI have been found in children and adults. However, it is not known whether short sleep time during childhood has long-term consequences. We assessed the association between sleep time in childhood and adult BMI in a birth cohort.
METHODS. Study members were a general-population birth cohort of 1037 participants (502 female) who were born in Dunedin, New Zealand, between April 1972 and March 1973. Parental reports of bedtimes and rising times collected at ages 5, 7, 9, and 11 years were used to estimate childhood sleep time. Linear regression analysis was used to analyze the association between childhood sleep time and BMI measured at 32 years of age.
RESULTS. Shorter childhood sleep times were significantly associated with higher adult BMI values. This association remained after adjustment for adult sleep time and the potential confounding effects of early childhood BMI, childhood socioeconomic status, parental BMIs, child and adult television viewing, adult physical activity, and adult smoking. In logistic regression analyses, more sleep time during childhood was associated with lower odds of obesity at 32 years of age. This association was significant after adjustment for multiple potential confounding factors.
CONCLUSIONS. These findings suggest that sleep restriction in childhood increases the long-term risk for obesity. Ensuring that children get adequate sleep may be a useful strategy for stemming the current obesity epidemic.
Identifying the causes of the current global epidemic of obesity is an urgent public health need.1–3 Simple measurements of food intake and physical activity seem to explain only a small part of an individual's risk for obesity, and it is not clear whether the worldwide increase in obesity prevalence can be explained adequately by population-based changes in diet and exercise.4 It is plausible that other changes in lifestyle have increased the risk for obesity. One of these changes is sleep duration.4,5 There is evidence that we are sleeping less than previous generations, and it has been suggested that this may be contributing to the increasing prevalence of obesity.4,6,7
Several studies found that short sleep duration was associated with increased risk of obesity in children,5,8–15 adolescents,16,17 and adults.18–25 However, there have been few longitudinal studies,5,12,22,23,25 and none have assessed whether less sleep during childhood was associated with increased risk for obesity in adulthood.
We assessed the association between childhood sleep duration and adult BMI in a birth cohort of 1037 individuals who were followed up until 32 years of age. We hypothesized that less sleep during childhood would be associated with increased adult BMI and that this relationship would be independent of a number of established child and adult predictors of obesity.
The Dunedin Multidisciplinary Health and Development Study is a longitudinal study of health and behavior in an unselected birth cohort. The study was described in detail elsewhere.26 Briefly, study members were born in Queen Mary Hospital, the only maternity hospital in the city of Dunedin, New Zealand, between April 1972 and March 1973. All children still residing in the Otago province were invited to participate in the first follow-up assessment at 3 years of age. A total of 1037 children (91% of eligible births, including 535 boys, ie, 52%) participated in the first assessment and formed the base sample for the longitudinal study. Study members were assessed every 2 years up to 15 years of age, again at 18, 21, and 26 years of age, and most recently at 32 years of age, when we assessed 96% (n = 972) of the living study members. Study members represent the full range of socioeconomic status in the general population of the South Island of New Zealand and are mostly of New Zealand European ethnicity.
We obtained written informed consent for each assessment. The study was approved by the Otago Ethics Committee.
Information on time spent in bed was obtained from parental reports at ages 5, 7, 9, and 11 years. At ages 5, 7, and 9 years, the parents attending with the study members reported what time the study members went to bed the night before and what time they woke up on the morning of the assessment day. At 11 years of age, the parents reported what time the study members usually went to bed and what time they usually got up the morning. These times were used to estimate the total sleep time for each age, and the mean of the available sleep times across ages 5 to 11 years was used as a composite measure of childhood sleep time (in hours).
Time in bed was also assessed at 32 years of age, when the study members themselves reported what time they usually went to bed and what time they usually got up. These values were used to estimate adult sleep time.
At 32 years of age, weight (in light clothing, without shoes) was measured by using calibrated scales (BC418; Tanita, Tokyo, Japan). Height was measured to the nearest 1 mm. These measures were used to calculate BMI (in kg/m2). Measurements obtained at 5 years of age (n = 893)27 were used as a measure of early BMI. Missing BMI values at 5 years of age were imputed from measurements taken at 3 years of age for an additional 120 participants.28
A number of potential confounding factors were included as covariates in the analyses, including sex, socioeconomic status,29 parental BMI, television viewing,28 smoking,30 parental control,31,32 and adult physical activity. These factors were chosen because they were either known or plausible mediating/confounding factors between childhood/adult lifestyle choice and adult BMI.
During the assessment at 11 years of age, the attending parents reported heights and weights for themselves and for the other parent. Parental BMIs when the study members were 11 years of age were calculated for 839 mothers (81%) and 798 fathers (77%). We also interviewed both parents when the study members were ∼32 years of age and obtained self-reported heights and weights, which were used to calculate BMIs for 844 mothers (81%) and 731 fathers (70%). The BMIs from the 2 ages were standardized by using z scores for each age, and the means of these were used as measures of mothers' (n = 961; 92%) and fathers' (n = 916; 88%) BMIs.
The socioeconomic status of the study members' families was measured by using parental occupational status assessed from birth to 15 years of age, as described previously.29 Each parent was assigned an occupational code (ranging from 1 = professional to 6 = unskilled laborer) based on educational level and income for that occupation from data in the New Zealand census. Final socioeconomic status scores were obtained by taking the highest scores of either parent and calculating the mean of those scores from birth to 15 years of age.
To adjust for parental control, we used the 9-item control subscale from the Family Environment Scale.33 The study members' mothers completed this subscale when the study members were 7 and 9 years of age; the scale was used to assess the extent to which set rules and procedures were used to run family life. The scale included items such as “family members are rarely ordered around,” “there is a strong emphasis on following rules in our family,” and “you can't get away with much in our family,” which were scored as either true or false. Responses indicative of parental control were summed, with scores ranging from 0 (low parental control) to 9 (high parental control). The mean of the summed scores from the 2 ages was used.
Parental estimates of weekday television viewing were at obtained at ages 5, 7, 9, and 11 years, and self-report estimates of television viewing were obtained at ages 13, 15, 21, and 32 years. These estimates were averaged and used as an overall measure of television viewing throughout the life course.
At 32 years of age, we obtained self-reported estimates of physical activity. Study members were asked to consider the last 7-day period that they had spent at home and to report the amounts of time they slept and engaged in physical activity. These values were recorded by trained interviewers, who elicited details about the amount of exertion involved in each activity. Activities were then coded as moderate, hard, or very hard, and these results were used to calculate total metabolic equivalent intensities.34,35 Current smoking at 32 years of age was defined as smoking ≥1 cigarette per day for ≥1 month during the past year.
Associations between childhood and adult sleep times and BMI at 32 years of age were analyzed by using linear regression. There were small but significant differences in the mean sleep times reported for boys and girls; therefore, analyses were adjusted for sex and tested for sex-sleep time interactions. Further analyses also adjusted for BMI at 5 years of age, childhood socioeconomic status, parental control, mothers' and fathers' BMIs, physical activity at 32 years of age, television viewing from 5 to 32 years of age, and current smoking. The regression models were checked by inspection of the residuals and, where appropriate, the analyses were repeated after logarithmic transformation of the variables. Because the mean sleep time decreased with age, additional checks of the regression models were undertaken by standardizing the childhood sleep times at each age and repeating the analyses. We also repeated the analyses after excluding those with missing sleep data for ≥1 age during childhood.
Logistic regression analyses were used to assess the association between childhood sleep time and obesity at 32 years of age. These analyses initially were adjusted for sex only and then were repeated with adjustments for the other covariates.
Data were analyzed by using SPSS 15.0 (SPSS, Chicago, IL). Statistical significance was set at P < .05. Pregnant women (n = 31) were excluded from all analyses.
Sleep time at ages 5, 7, 9, and 11 years were all positively correlated with each other, and values tended to decrease from 5 to 11 years of age. Of the 4 childhood sleep time estimates, only sleep time at 11 years of age was significantly correlated with sleep time at 32 years of age (Table 1).
In sex-adjusted linear regression analyses, mean childhood sleep time (in hours) predicted adult BMI (Table 2). There was no significant sex-sleep time interaction, and the regression coefficients were similar when the sex were analyzed separately. The inverse association between sleep time and adult BMI remained after controlling for early BMI, childhood socioeconomic status, parental control, parental BMI values, television viewing, physical activity, smoking, and adult sleep time (Table 2). The residuals from these regression models were slightly skewed. Analyses using logarithmically transformed BMI scores corrected this problem but did not materially alter the significant inverse associations between childhood sleep time and adult BMI.
In contrast, sleep time at 32 years of age was not associated with adult BMI (unstandardized regression coefficient: −0.01; 95% confidence interval: −0.31 to 0.29; P = .95). This association remained nonsignificant with controlling for childhood sleep time and all of the covariates (Table 2).
Analyses using the mean of the standardized sleep times (z scores) for each age provided similar, statistically significant, inverse associations between childhood sleep time and adult BMI. Analyses restricted to study members with sleep data for all of the childhood assessment ages (n = 777) provided similar, statistically significant results.
By 32 years of age, 492 (53%) of 930 study members were overweight (BMI of ≥25 kg/m2), of whom 164 (18%) were obese (BMI of ≥ 30 kg/m2). In logistic regression analyses, greater childhood sleep time predicted lower odds of adult obesity (Table 2).
On the basis of the data, we created categories of childhood sleep time to compare BMIs at each age (Fig 1). Children who spent a mean of up to 11 hours in bed between ages 5 and 11 years were defined as short sleepers (n = 301), children who were in bed between 11 and 11.5 hours were defined as moderate sleepers (n = 400), and children who were in bed for >11.5 hours were defined as long sleepers (n = 311). At ≥7 years of age, the BMIs for short childhood sleepers were consistently higher than those for moderate and long sleepers. Independent-sample t tests, comparing short sleepers with moderate and long sleepers at each age, were significant (all P < .05). There were no consistent differences in BMIs between moderate and long sleepers.
In this population-based birth cohort, we showed that short sleep time in childhood was associated with increased adult BMI. This association was independent of a number of important familial and lifestyle predictors of adult BMI, including early childhood BMI, parental BMIs, physical activity, television viewing, and smoking. These findings support the hypothesis that less sleep during childhood increases the long-term risk for obesity.
Our findings are consistent with a number of cross-sectional studies that reported associations between childhood sleep duration and BMI,8–10 and they are also consistent with short-term follow-up studies of sleep and BMI in childhood.5,12,13 Unlike other studies,18–24 we did not find an association between adult sleep time and BMI. Also, we did not find evidence for a curvilinear association between adult sleep time and BMI (data not shown), as shown by some.36–39 The reasons for the differences between our findings and those reported by others are not clear; they may be attributable to the younger age of our adult participants, compared with most other adult studies. However, our findings do indicate that the long-term association between childhood sleep time and adult BMI is not likely to be mediated by the continuity of sleep habits from childhood into adulthood.
The correlation coefficient for the correlation between childhood sleep time and adult BMI was −0.11. This is generally regarded as consistent with a weak effect. However, this effect size is of a similar magnitude as those found for the more-obvious predictors of BMI such as diet and physical activity.4,40,41 A short childhood sleep time was also a significant predictor of increased risk for adult obesity. In view of the increasing prevalence of obesity, an effect of this size may have important public health implications.
Although data on sleep times in previous generations are scarce, there is some evidence to suggest that children of the same age may be going to bed as much as 2 hours later today than they did 20 years ago.7 This decrease in sleep time seems to coincide with the dramatic increase in the prevalence of obesity.1 Neither the association between childhood sleep time and adult BMI in our study nor the apparent ecological association between reduced sleep time and increased obesity necessarily implies causation. Nevertheless, our findings suggest that sleep restriction could be a plausible contributing factor to the current epidemic of obesity.
Our study does not indicate a mechanism for the association between short sleep time and risk for obesity, but it is important to note that the association was independent of a number of other lifestyle factors, including television viewing, childhood socioeconomic status, adult smoking, and adult physical activity. The association was also independent of the parental control measure, which was used to assess the extent to which rules and procedures were used to run family life. This suggests that the association between sleep time and obesity is unlikely to be explained by the possibility that families who have lax rules regarding bedtimes also have lax rules with respect to other health-related behaviors. The association also was independent of early BMI, and the findings were similar if we excluded those who were overweight or obese42 at age 5. This suggests that the association is unlikely to be attributable to reverse causation. In other words, it is unlikely that children slept less because they were already overweight.
Several explanations for the association between short sleep times and obesity have been proposed.43 Observational36,39 and experimental studies44–46 found that shorter sleep times and sleep restriction were associated with elevated levels of ghrelin, an appetite-stimulating hormone, and decreased levels of leptin, an appetite-suppressing hormone. This suggests that sleep debt may disrupt hunger and appetite regulation. Another explanation may be that tiredness, resulting from shorter sleep times, alters behavior, including a reduction in physical activity. Tiredness also may affect dietary habits. People who are tired may seek fast-release, high-energy foods to compensate for perceived low energy levels. It also has been suggested that more time spent sleeping simply reduces the opportunity to eat.43 Unfortunately, we do not have sufficient information on food consumption and physical activity during childhood to explore whether the association between childhood sleep time and adult BMI is mediated by these factors.
There are a number of limitations to our analyses. Bedtimes were reported by the parents of the study members and might contain inaccuracies. We do not have information on sleep times before 5 years of age, sleep times during adolescence, or differences between weekday and weekend sleep times. For ages 5, 7, and 9 years, the parents were asked about the previous night, which might not have been representative of general sleep patterns. At age 11 years, however, parents were asked about usual sleep times, and analyses using only those sleep time estimates provided similar results. We also do not have information on sleep latency or quality of sleep, and time in bed does not necessarily equate to time asleep. However, these limitations are most likely to reduce the magnitude of any association between sleep time and BMI and are unlikely to explain the significant associations that we observed.
Our study also has a number of strengths. Data were collected prospectively throughout childhood and adolescence and into adulthood from a large, general-population, birth cohort, with a high rate of participation. Study members' heights and weights were measured directly, to provide objective assessments of BMI at all ages. We also were able to control for important covariates that signal a predisposition to increased BMI, including early BMI and parental BMIs. Finally, because we assessed sleep times at multiple ages, our data are likely to represent a pattern of sleep behavior during a large part of childhood.
We found an inverse association between childhood sleep time and adult BMI. This association was independent of a number of important covariates. Although the mechanism for this association is unclear, the findings indicate that sleep restriction in childhood may have long-term consequences for risk of obesity. Ensuring that children get adequate sleep may be a useful strategy for stemming the current obesity epidemic.
The study was funded by the Health Research Council of New Zealand. This research was also supported by the US National Institute of Mental Health (grant MH 49414). Mr Landhuis was supported by a Dunedin School of Medicine Strategic Initiative Award. The sponsors of the study had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, and approval of the manuscript.
We thank the study members and their families and friends for their participation and ongoing support. We also thank Dr Phil Silva, founder of the study, and Profs Malcolm Sears and Avshalom Caspi for facilitating access to data.
- Accepted February 6, 2008.
- Address correspondence to Bob Hancox, MD, Dunedin Multidisciplinary Health and Development Research Unit, Department of Preventive and Social Medicine, Dunedin School of Medicine, University of Otago, PO Box 913, Dunedin, New Zealand. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject
Short sleep duration has been associated with increased BMI and obesity in children and adults in cross-sectional and short-term follow-up studies.
What This Study Adds
Short sleep time during childhood is associated with a higher BMI in adulthood and an increased long-term risk for obesity. Restricted sleep time in childhood is a possible contributor to the current epidemic of obesity.
- ↵Kim S, Popkin BM. Commentary: understanding the epidemiology of overweight and obesity: a real global public health concern. Int J Epidemiol.2006;35 (1):60– 67
- ↵Reilly JJ, Armstrong J, Dorosty AR, et al. Early life risk factors for obesity in childhood: cohort study. BMJ.2005;330 (7504):1357
- ↵Iglowstein I, Jenni OG, Molinari L, Largo RH. Sleep duration from infancy to adolescence: reference values and generational trends. Pediatrics.2003;111 (2):302– 307
- ↵Lumeng JC, Somashekar D, Appugliese D, Kaciroti N, Corwyn RF, Bradley RH. Shorter sleep duration is associated with increased risk for being overweight at ages 9 to 12 years. Pediatrics.2007;120 (5):1020– 1029
- ↵Patel SR, Malhotra A, White DP, Gottlieb DJ, Hu FB. Association between reduced sleep and weight gain in women. Am J Epidemiol.2006;164 (10):947– 954
- ↵Silva PA, Stanton WR, eds. From Child to Adult: The Dunedin Multidisciplinary Health and Development Study. Auckland, New Zealand: Oxford University Press; 1996
- ↵Molarius A, Seidell JC, Kuulasmaa K, Dobson AJ, Sans S. Smoking and relative body weight: an international perspective from the WHO MONICA Project. J Epidemiol Community Health.1997;51 (3):252– 260
- ↵Moos R, Moos B. Family Environment Scale Manual. Palo Alto, CA: Consulting Psychologists Press; 1981
- ↵Sallis JF, Haskell WL, Wood PD, et al. Physical activity assessment methodology in the Five-City Project. Am J Epidemiol.1985;121 (1):91– 106
- ↵Berkey CS, Rockett HR, Field AE, et al. Activity, dietary intake, and weight changes in a longitudinal study of preadolescent and adolescent boys and girls. Pediatrics.2000;105 (4). Available at: www.pediatrics.org/cgi/content/full/105/4/e56
- ↵Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ.2000;320 (7244):1240– 1243
- ↵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– 884
- ↵Spiegel K, Leproult R, L'Hermite-Baleriaux M, Copinschi G, Penev PD, Van 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– 5771
- Copyright © 2008 by the American Academy of Pediatrics