|Home | About | Journals | Submit | Contact Us | Français|
Previous studies have demonstrated an association between sleep duration and obesity, but few population-based studies have examined the association. We examined the relationship between recent and usual lifetime sleep duration with the odds of obesity in 5,549 women that participated in a population-based telephone survey.
The structured telephone interview included questions on usual sleep duration in adult life and the recent past, as well as height and weight and other demographic and lifestyle characteristics. We examined odds of overweight (BMI: 25-29.9 kg/m2), obesity (BMI: 30-39.9 kg/m2) and extreme obesity (BMI: ≥40 kg/m2) according to reported sleep duration.
Compared to women who slept 7-7.9 hours per night, women who slept an average of <6 hours per night in the recent past had significantly greater odds of obesity (Odds Ratio [OR]: 1.89; 95% Confidence Interval [CI]: 1.45-2.47) and extreme obesity (OR: 3.12; CI: 1.70-5.75), adjusting for potential confounding factors. Weaker associations were noted for short lifetime sleep duration. Current short sleep (< 7 hours) was associated with greater odds of obesity (≥ 30kg/m2) in those reporting less than 7 hours (OR: 1.59; 95% CI: 0.93-2.78) and in those reporting 8 or more hours (OR: 1.75; 95% CI: 1.33-2.32) of sleep throughout adult life.
Current short sleepers were more likely to be obese regardless of their usual sleep duration earlier in life. These findings do not support the hypothesis that sleep duration is a causal factor in obesity.
The prevalence of obesity has doubled during the last 25 years, and currently one-third of adults in the United States are obese . Obesity is associated with an increased risk of many chronic diseases including heart disease , diabetes  and cancers at multiple sites [3,4]. Short sleep duration may increase the risk of obesity by altering hormones involved in the regulation of appetite . In many cross-sectional studies, short sleep duration was associated with a higher body weight in adults and children (reviewed in [6-8]). A similar association has also been identified in several [9-12], though not all [13-15], prospective longitudinal studies. A U-shaped association has also been observed in several studies [11,16-19] wherein both short and long sleep durations were associated with the odds of obesity. We examined the relationship between sleep duration and obesity in a population-based sample of more than 5,000 adult US women. Unlike most previous cross-sectional studies, we were able to consider the association of both recent and usual lifelong sleep duration with the odds of obesity.
Women in this analysis were enrolled as controls in the Collaborative Breast Cancer Study, a population-based case-control study of determinants for breast cancer comprised of women ages 20-75 who resided in Wisconsin, Massachusetts and New Hampshire . The study was conducted according to protocols approved by the institutional review boards at each study site. Participants were randomly selected from drivers license lists if younger than age 65 or from a roster of Medicare beneficiaries if aged 65 years or older. Eligible controls had no prior history of breast cancer, had a listed telephone number, were able to participate in an English-speaking interview, and were licensed to drive for those under the age of 65. The present analysis is based on 5,706 women enrolled in the study between February 1997 and November 1999. Participation among the controls was estimated at 76% .
Data for sleep duration were collected during a structured telephone interview. Women were asked to report the usual number of hours they slept per night approximately 1-2 years before the date of the interview (the reference date) and over their adult lifetimes. Recent and lifetime sleep durations reported by women were highly correlated: (Spearman correlation coefficient = 0.64) . The interview also elicited information on demographic and lifestyle characteristics including age, education, marital status, alcohol consumption, cigarette smoking, leisure-time physical activity, menopause status, work history and height and body weight at the reference date. Body mass index (BMI) in kilograms per meters squared was calculated from self-reported height and weight at the reference date.
After excluding observations with missing data on BMI or sleep duration, a total of 5,549 women were included in analyses of sleep duration at the reference age and 5,487 women were included in analyses of average sleep duration during adult life. The median age of women in the analysis was 55 years (range: 20 – 75 years). The majority of women were Caucasian (95.5%).
Unconditional logistic regression was used to estimate odds ratios (ORs) and Wald 95% confidence intervals (CIs) for the relationship of sleep duration with excess body weight. Separate models were constructed for different definitions of excess body weight  including overweight (BMI = 25-29.9 kg/m2), obesity (BMI = 30-39.9 kg/m2), and extreme obesity (BMI ≥ 40 kg/m2), with a sleep duration of 7-7.9 hours per night serving as the referent category. All models included terms for age (in 5 year categories) and study site (Wisconsin, Massachusetts or New Hampshire). Multivariate models included terms for age, study site, education (no high school diploma, high school diploma, unknown), alcohol consumption (none, any, unknown), cigarette smoking (pack, years), average hours per week of leisure-time physical activity (< median hours per week, ≥ median hours per week), marital status (married, not married, unknown), menopause (pre-menopausal, post-menopausal, unknown), as well as employment status (working, not working) at the reference date.
We considered potential effect modification of the association between obesity and recent and adulthood sleep duration by including cross-product terms for sleep duration categories with age (<50 vs. ≥50), physical activity (any vs. none), smoking (current vs. past/never) and employment status (currently working vs. not currently working) in multivariate models. We tested for interaction using a likelihood ratio test that compared model log likelihoods with and without cross-product terms included in the multivariate models.
Characteristics of participants by sleep duration at the reference date are shown in Table 1. Older women reported sleeping fewer hours. Any alcohol consumption was less often reported by women reporting both short (< 6 hours) and long durations (≥ 9 hours) of sleep. Higher pack-years of smoking were also associated with shorter sleep intervals. Women reporting fewer hours of sleep were more likely to be postmenopausal, not currently married, to have less than a high school education and to not participate in leisure-time physical activity. Current employment was less often reported by women reporting both shorter (< 6 hours) and longer (≥ 9 hours) durations of sleep. The mean BMI was significantly different across sleep duration categories for both recent (p <0.0001) and lifetime (p <0.0001) sleep. The median BMI for each sleep duration category (<6, 6-6.9, 7-7.9, 8-8.9 and ≥9 hours per night) was 26.5, 25.8, 24.9, 25.0, and 25.6 kg/m2, respectively, for recent sleep and 26.6, 25.8, 24.9, 25.1, and 25.0 kg/m2, respectively, for average adult lifetime sleep duration (not shown).
Table 2 presents the odds of overweight (BMI = 25-29.9 kg/m2), obesity (30-39.9 kg/m2), and severe obesity (BMI ≥ 40kg/m2) when compared to a normal body weight (BMI < 25 kg/m2) according to usual sleep duration over adult life and in the recent past. ORs are shown adjusted for age and study center and for potential confounders. All associations between sleep duration and obesity were attenuated after adjustment for potential confounders including education, marital status, alcohol consumption, cigarette smoking, leisure-time physical activity, menopause status, and current employment status.
For overweight (BMI = 25-29.9 kg/m2), we observed an approximately 30% excess odds associated with a shorter sleep duration (<7 hours per night) over adult life which was statistically significant for 6-6.9 hours (multivariate OR: 1.30; 95% CI: 1.09-1.56) when compared to the referent (7-7.9 hours). No similar excess was observed for short sleep duration in the recent past. Odds of overweight were unrelated to long sleep duration (≥ 8 hours) in adult life or the recent past.
For obesity (BMI = 30-39.9 kg/m2), we observed increasing odds with shorter durations of sleep both in adult life and in the recent past. For recent sleep duration, the multivariate-adjusted odds ratios were 1.52 (95% CI: 1.23-1.89) for 6-6.9 hours of sleep and 1.89 (95% CI: 1.45-2.47) for less than 6 hours of sleep when compared to 7-7.9 hours (the referent). The corresponding odds ratios for sleep duration throughout adult life were 1.46 (95% CI: 1.17-1.82) and 1.54 (95% CI: 1.07-2.21), respectively. A long duration of sleep (≥ 8 hours) was unrelated to the odds of obesity.
For extreme obesity (BMI ≥ 40kg/m2), we again observed positive associations with a shorter sleep duration, with somewhat stronger and more consistently significant associations noted for sleep duration in the recent past. The odds of extreme obesity were also increased for longer sleep durations. A significant U-shaped association (p <.001) was demonstrable for sleep duration in the recent past, with significantly elevated odds observed both for short (multivariate OR: 3.12; 95% CI: 1.70-5.75 for less than 6 hours) and long (multivariate OR: 2.53; 95% CI: 1.10-5.78 for 9 or more hours) reported sleeping hours 1-2 years prior to the interview.
We further evaluated associations in a subset of women in the survey who were asked about use of antidepressant medications, which may influence sleep patterns. Restricting analysis to 1,864 women who reported never using any antidepressant medication in the past did not materially change the results (data not shown), though all results were based on sparse data.
Associations between recent sleep duration and obesity were not materially different in younger (<50) and older (≥50) women (p for interaction = 0.54) (not shown). Likewise, associations were similar regardless of levels of recent physical activity (p for interaction = 0.19), smoking status (p for interaction = 0.54) and current employment (p for interaction = 0.17) (not shown). Similarly, null results were obtained when we considered these potential interactions with usual sleep duration in adult life (not shown).
We next examined prevalence of obesity in women who reported stable or changing durations of sleep in the recent past when compared to lifetime sleep durations (Table 3). Among lifetime long sleepers (≥8 hours per night), current short sleepers (< 7 hours per night) had a significantly higher mean BMI than women who reported longer current sleep durations (≥8 hours per night) (26.4 kg/m2 and 25.0 kg/m2, respectively (p<0.001)). Odds of obesity (≥ 30kg/m2) were 75% higher in women currently sleeping fewer hours (OR: 1.75; 95% CI: 1.33-2.32) than in the past. A similar pattern was noted among lifelong short sleepers (< 7 hours per night): current short sleepers had a slightly higher mean BMI when compared to lifelong short sleepers who recently increased their sleep time to a minimum of 8 hours per night (26.2 kg/m2 and 25.7 kg/m2, respectively (p=0.24)); consistent short sleepers had approximately 60% greater odds of obesity than women currently sleeping longer hours (OR: 1.59; 95% CI: 0.93-2.78), though the difference was not significant.
We observed no association of sleep duration with low body weight (BMI <18.5 kg/m2)  (not shown). When compared to women reporting 7-7.9 hours in the recent past, the odds of being underweight (<18.5 kg/m2) were similar for women reporting <7 hours (OR = 0.97; 95% CI: 0.59-1.59) and ≥8 hours (OR = 0.91; 95% CI: 0.58-1.44) of sleep per night. Few women in the population were underweight (110 women had a BMI <18.5 kg/m2) and these associations were imprecise.
In this population-based study, we found that obese and severely obese women were more likely to report shorter durations of sleep in the recent past and over their adult lifetimes. The odds of being overweight (defined as a body mass index of 25-29.9 kg/m2) were also modestly increased among women reporting habitual shorter sleep durations in adult life. This association was consistently observed regardless of a woman's age, physical activity levels, smoking, and employment status. Obesity was more common among current short sleepers regardless of reported sleep times in the past. A further finding was that long sleep duration in the recent past was related only to severe obesity.
Our findings are consistent with many previous studies that have demonstrated an association between short sleep duration and obesity [6-8]. In a meta-analysis of 18 cross-sectional studies encompassing 604,509 men and women from around the world , the pooled odds ratio for short sleep duration (generally less than 5 or 6 hours per night) and obesity was 1.55 (95% confidence interval: 1.43 to 1.68; P < 0.0001). Similar patterns were noted in children  and have been reported in adult men and women in many studies [6,7]. Results of several studies suggest that obese individuals systematically underestimate time asleep [23,24] when comparing self-report to objective measures of sleep duration. Thus, the association between short sleep duration and obesity in many cross-sectional studies could reflect systematic over-reporting of short sleep times among the obese. Nevertheless, the fact that it has also been demonstrated in studies based on objectively measured sleep duration using wrist actigraphy [15,18,25] and polysomnography  suggests that the finding is not attributable solely to reporting bias.
Prospective analyses have produced conflicting results, some [9-12] though not all [13-15] demonstrating an increased risk of subsequent weight gain in short sleepers. In a recent longitudinal analysis—the first to be based on objectively measured sleep duration (average results from 6 days of actigraphy)—Lauderdale et al.  were able to demonstrate the expected cross-sectional relationship  but did not identify a longitudinal association over an average follow up of 5 years. Of note, snoring was a significant predictor of weight gain in the Lauderdale study. As obesity is a risk factor for obstructive sleep apnea  (a symptom of which is snoring), which disrupts sleep, it has been suggested that the cross-sectional relationship between short sleep and obesity may be due to the effects of obesity on sleep rather than vice versa .
Several findings from the current study favor this explanation. The odds of obesity were consistently more elevated for short/recent duration than lifetime sleep duration, as would be expected if obesity (as assessed at the time of the survey) was causing disrupted sleep. (The observed association with “lifetime” sleep duration may be a reflection of the high correlation between lifetime and recent sleep duration). Moreover, current short sleepers were more likely to be obese regardless of their usual sleep duration earlier in life (Table 3). The lack of evidence of a temporal relationship in our data (e.g., short sleep preceding current obesity) supports the argument that short sleep is not causal for obesity but rather is a consequence of excess body weight.
The relationship of long duration sleep with obesity has been less consistent in published reports. Several studies have found a U-shaped association in which obesity was related both to short and long durations of sleep [11,16-19]. In the prospective Nurses’ Health Study, a higher mean BMI was observed at both extremes of sleep duration, but only short sleepers were at greater risk of becoming obese in the follow up period . In the present study, only severely obese women were more likely to report sleeping for long durations and only in the recent past. This pattern would be observed if women extremely overweight at the time of the survey tended to overestimate their time spent asleep because of reduced mobility and longer hours spent in bed. The absence of a similar association for lifetime sleep duration in our data is consistent with this interpretation.
We observed no association of sleep duration and low body weight (not shown). This is in contrast to results from the National Health and Nutrition Examination Survey (NHANES), in which a higher prevalence of underweight status was observed in those reporting both short and long sleep durations . The present data was based on limited numbers of underweight women and the results should be interpreted with caution.
In this large population-based sample of women, we observed significantly increased odds of obesity among women reporting short durations of sleep and severe obesity in women reporting long durations of sleep in the recent past. Our findings, taken together, are most likely attributable to the effects of sleep apnea in the former and immobility in the latter on perceived or actual durations of sleep. These findings combined with recent prospective studies tend to rule out a causal role of sleep duration on obesity.
This study was supported by National Cancer Institute grants R01 CA47147, R01 CA47305, and R01 CA69664.
Financial support: This study was supported by National Cancer Institute grants R01 CA47147, R01 CA47305, and R01 CA69664.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflicts of interest: none.