This is one of the few studies to examine the relationship between sleep and body composition variables beyond body weight and BMI in adults. Furthermore, it is the only study to date to examine the relationship between sleep and body composition in a bi-racial population of Blacks and Whites. As such, this is the first study to examine whether sleep duration is related to waist circumference and percent body fat in Black and White adults.
Few studies to date have examined the relationship between waist circumference and sleep in adults. In the Better Health for Better Hong Kong campaign, waist circumference was inversely related to sleep duration [5
], in agreement with our data in women. In men, although the association was not statistically significant, the magnitude and direction of the association between waist circumference and sleep duration was similar to that observed in women. However, in a sample of older Spanish adults, waist circumference was not related to sleep duration [23
]. Results from the Quebec Family Study also showed larger waist circumference in men and women who reported sleeping <7
h/night compared to those who reported sleeping 7-8
]. It is possible that the age difference between the Spanish cohort and our young adult cohort is responsible for the different results. This relationship between waist circumference and sleep duration is of importance since well-known metabolic disturbances are associated with increased waist circumference [25
In our study, BMI and waist circumference were inversely associated with sleep duration in women before including physical activity in our model. In men, associations tended to be in the same direction and of similar magnitude, but were not statistically significant. This gender difference has not been consistently reported in adults. In fact, no gender effect on the association of sleep and BMI was observed in an elderly Spanish cohort [23
], a primary care U.S. sample [26
], or the Quebec Family Study [24
]. However, in a sample of Southern France, sleep duration was found to be related to BMI in women but not men [27
]. Gangswisch et al. [9
] have previously reported that BMI was higher in men and women who slept less than 7 hours compared to those who slept 7 hours or more only in individuals age 32–49. This age group includes the age group of our cohort (33–45 years). In the NHANES data, gender differences were also observed in the association between BMI and sleep duration, with women being progressively more likely to be obese as sleep duration was reduced below 7
h/night whereas men were more likely to be obese with 6 or fewer hours of sleep/night. In our study, the association between BMI and sleep duration was significant in women but a trend in men. Our data add to the body of literature suggesting that sleep may have a slightly greater impact on body composition in adult women than in men. This potential gender difference in the relationship between sleep and adiposity deserves further exploration, especially since discordant data have also been observed in children and adolescents [10
]. Previous metabolic studies have been conducted in young men only [30
] or in a study with few women [32
]. Future metabolic studies should be done in women to determine if they have a different hormonal response to short sleep duration than men.
Our finding that percent body fat was not significantly associated with sleep duration is novel. One previous study has examined the relationship between this body composition measure and sleep duration in adults [24
]. In our cohort, we did not find any significant association between these two variables. In the Quebec Family Study, Chaput et al. [33
] found lower percent body fat in men and women who reported sleeping 7-8
h/night compared to those reporting 5-6 hours of sleep per night. It is possible that our sample size was not sufficient to detect an association. In fact, only 320 subjects participated in the DXA measurement, leaving few participants within each race-gender category. In comparison, the Quebec Family Study had more than twice our sample size with only one racial group.
In this study, adding physical activity to our models attenuated or abolished the relationship between sleep duration and body composition. We can speculate that sleep may in part act on body composition via changes in physical activity. It is possible that adults who sleep less are too tired to be physically active or those who sleep very long have too little time to be physically active. In fact, in our data, physical activity tended to increase with increasing sleep duration in women (data not shown). Moreover, being physically active can improve sleep quality and duration [34
] such that physical activity and sleep can be positively related.
Conservation of energy is a well-known principle that applies to obesity research. In this principle, energy is neither created nor lost. Therefore, in order to lose weight, one must expend more energy than is consumed or consume less energy than is expended; to gain weight, one must expend less energy than is consumed or consume more energy than is expended. It thus follows that our hypothesis explaining the relationship between sleep duration and body composition must (short of variations in energy partitioning) be due to an effect of sleep on energy expenditure or energy intake. If one controls for energy intake (food consumption) and energy expenditure (not equivalent, but closely related to physical activity), much of the variance in body composition variables is accounted for. This was the case when we added physical activity level to our models. In addition, food intake may be increased in those who sleep less. An increase in food intake with shorter sleep duration may result from various factors; longer hours awake create more time for eating, and hunger may increase with shorter sleep duration. Increased hunger ratings have been reported with short sleep duration and these have been related to reduced leptin and increased ghrelin levels [33
]. Furthermore, in a recent report from the Quebec Family Study, the relationship between sleep duration and adiposity measures disappeared after controlling for leptin levels [24
]. In fact, 88% of short sleepers had low leptin levels, confirming a previous clinical study showing that short sleep decreases leptin concentrations [33
]. Dietary information was not taken at Y15 in the CARDIA study and we could not include this variable in our models.
Our study has several limitations. First, our sleep data are self-reported. Sleep was objectively measured in a subsample of our cohort [38
]. In that study it was shown that the correlation between self-reported sleep duration and actigraph-measured sleep duration was 0.47. Furthermore, there was a systematic overestimation of sleep duration that was greater in short sleepers than in those sleeping more than 7
h/night. Also, self-reported sleep duration by obese individuals tended to be more closely related to actigraph-measured sleep duration than lean individuals, who tended to overestimate their sleep duration in self-reports relative to the actigraph measurement. Thus, obese individuals may provide more accurate self-reports of their sleep duration than nonobese individuals and nonobese individuals over-report their sleep times. These data suggest that longer sleep hours would falsely include a greater proportion of lean individuals than obese individuals. This might then lead one to erroneously conclude that BMI decreases with increasing sleep duration even if this were not actually the case. In fact, Knutson and Lauderdale [36
] have reported that the odds ratio for overweight in adolescents using time-diary sleep times was not significantly affected by sleep duration whereas the odds ratio for overweight was significantly increased with reduced sleep duration when sleep duration was self-reported. On the other hand, Nixon et al. [37
] found a significant inverse association between actigraphy-measured sleep duration and overweight/obesity in 7-year old children. Similarly, Taheri et al. [7
] found a significant U-shaped relationship between sleep duration and BMI when sleep was measured using polysomnography. More research is needed to determine the degree of error introduced by biased sleep reporting into the association of sleep duration and obesity.
Finally, our data show that the relationship between sleep duration and body composition is not affected by race but rather differs according to gender. In fact, women show a more consistent effect of sleep duration on body composition than men. Our data thus suggest that short sleep duration may be a risk factor for overall and abdominal obesity in young Black and White women. However, our overall small sample size, self-reported measures of sleep, and the cross-sectional nature of our data do not allow us to make definitive conclusions on the impact of gender on the relationship between sleep duration and obesity. More studies are needed to examine the gender difference on the relationship between sleep and obesity and consideration must be given to objectively-measured sleep duration.