The association between sleep duration, body mass index and metabolic measures in the Hordaland Health Study

Transcription

1 J. Sleep Res. (2007) 16, The association between sleep duration, body mass index and metabolic measures in the Hordaland Health Study BJØRN BJORVATN 1,5, INA MARIE SAGEN 1, NICOLAS ØYANE 1, SIRI WAAGE 1, ARNE FETVEIT 2, STA LE PALLESEN 3,5 and REIDUN URSIN 4,5 1 Department of Public Health and Primary Health Care, University of Bergen, Bergen, Norway, 2 Department of General Practice and Community Medicine, University of Oslo, Oslo, Norway, 3 Department of Psychosocial Science, University of Bergen, Bergen, Norway, 4 Department of Biomedicine, University of Bergen, Bergen, Norway and 5 Norwegian Competence Center for Sleep Disorders, Haukeland University Hospital, Bergen, Norway Accepted in revised form 30 October 2006; received 16 June 2006 SUMMARY Several studies show that short self-reported sleep duration is associated with elevated body mass index (BMI). Short sleep duration may change appetite hormones, but whether this also influences metabolic measures like cholesterol and triglycerides is less clear. Furthermore, obesity is linked to increases in blood pressure, and recently, short sleep duration has been shown to be an independent risk factor for hypertension. This is a population-based cross-sectional study (The Hordaland Health Study). A subgroup of 8860 subjects, aged years, answered a sleep questionnaire. Body weight, height and blood pressure were measured, and non-fasting blood samples were collected and analyzed for total cholesterol, HDL-cholesterol and triglycerides. was divided into the following subgroups: <5, , , , and 9 h. The results show that short sleep duration was associated with elevated BMI and increased prevalence of obesity. Similar to BMI, levels of cholesterol, triglycerides, systolic and diastolic blood pressure were higher in subjects with short sleep duration. This covariation seemed to be attributed to variables like gender, smoking and BMI. In conclusion, our study confirms a clear association between short sleep duration and elevated BMI and obesity. Furthermore, levels of total cholesterol, HDL-cholesterol, triglycerides and blood pressure were associated with sleep duration. keywords blood pressure, bmi, cholesterol, obesity, sleep, triglycerides INTRODUCTION Elevated body mass index (BMI; calculated as weight in kilograms divided by the square of height in meters) has in several previous reports been associated with a decrease in sleep duration across many different age groups (Gangwisch et al., 2005; Gupta et al., 2002; Hasler et al., 2004; Heslop et al., 2002; von Kries et al., 2002; Kripke et al., 2002; Locard et al., 1992; Singh et al., 2005; Vioque et al., 2000; Vorona et al., 2005). The US National Health and Nutritional Examination Survey found that, in 1960, an estimated 13.4% of adult men and women were obese (BMI 30), whereas, by Correspondence: Bjørn Bjorvatn MD, PhD, Department of Public Health and Primary Health Care, University of Bergen, Kalfarveien 31, N-5018 Bergen, Norway. Tel.: ; fax: ; bjorn.bjorvatn@isf.uib.no 2000, the proportion of obese adults had increased to 30.5% (Flegal et al., 2002) with a parallel increase in obese-related healthcare costs (Colditz, 1999). There has been a discussion whether sleep duration has decreased in the same period. Several studies from the USA indicate this (see discussion in Gangwisch et al., 2005; Singh et al., 2005) whereas a study from England indicated no statistically reliable reduction in sleep duration from 1969 to 1997 (Groeger et al., 2004). In a recent prospective study, Hasler et al. (2004) identified a longitudinal association between short sleep duration and future weight gain. Whether short sleep per se interact with body weight is unknown, but several metabolic functions may be involved, for example, hormones regulating appetite and energy expenditure (Spiegel et al., 1999, 2004; Taheri et al., 2004). Taheri et al. (2004) demonstrated that short sleep duration was associated with elevated BMI and with reduced 66 Ó 2007 European Sleep Research Society

2 Sleep is related to obesity and metabolic measures 67 levels of leptin and elevated levels of ghrelin, which are hormonal changes known to increase appetite. Recently, an epidemiological study by Gangwisch et al. (2006) showed that short self-reported sleep duration increased the risk for hypertension incidence, even after controlling for obesity and diabetes. The follow-up period was 8 10 years. One possible explanation is that short sleep duration may increase sympathetic nervous system activity (Gangwisch et al., 2006; Spiegel et al., 1999). Whether short sleep duration may be related to metabolic measures like cholesterol and triglycerides is unknown. Body fat and obesity are obviously related to these measures, but an independent association of short sleep is possible. In a pioneering study from 1999, sleep restricted to 4 h per night for 1 week led to endocrine and metabolic changes associated with diabetes (insulin resistance) and weight gain (Spiegel et al., 1999). Short sleep duration, or chronic sleep debt, is associated with increased hunger and appetite, especially for calorie-dense foods with high carbohydrate content (Spiegel et al., 2004). In a recent review, Spiegel et al. (2005) provide convincing evidence that chronic sleep debt may represent a novel risk factor for weight gain, insulin resistance and type 2 diabetes. Although several epidemiological studies link short sleep duration to elevated BMI, the design of these studies varies., weight and height are usually self-reported, introducing possible bias. In the present large populationbased study of 40- to 45-year olds, both weight and height were objectively measured. was defined based on self-reported bedtime and rise time. Blood samples were collected from all subjects, measuring biochemical markers like cholesterol and triglycerides. In addition, blood pressure was measured. As far as we know, no epidemiological study has previously investigated sleep duration and BMI in relation to such metabolic variables. METHODS The cross-sectional Hordaland Health Study (HUSK) was conducted during as a collaboration between the National Health Screening Service, the University of Bergen and the local health services. The study population included all individuals in Hordaland county, Norway, born between 1953 and 1957 (n ¼ ). A total of 8598 men and 9983 women participated, yielding a participation rate of 63% (57% for men and 70% for women). The study protocol was approved by The Regional Ethics Committee and by the Norwegian Data Inspectorate. The participants completed self-administered questionnaires with information on various health behaviors. SubjectsÕ height and weight were measured (shoes and outer garment were taken off) by specially trained study personnel. In addition, specially trained nurses measured blood pressure and collected blood samples. Systolic and diastolic blood pressures were measured in a relaxed sitting position following 2 min of rest. During the rest period the participants were instructed that three automated blood pressure measurements were to be performed, and that no talking was allowed during the measurements. The average of the last two measurements was used as the recorded blood pressure. Blood samples were collected non-fasting, and later analyzed for total cholesterol, HDL-cholesterol and triglycerides. A subgroup of 8860 participants (3531 men and 5329 women) answered a questionnaire with detailed information on sleep habits and problems, based on a revised version of the Karolinska Sleep Questionnaire (A kerstedt et al., 2002; Ursin et al., 2005). The questionnaire was given to the participants when they met for the health screening (BMI, blood pressure and blood sampling) and was submitted by mail when completed. A detailed analysis of the sleep habits in this population is published elsewhere (Ursin et al., 2005). In the present study, sleep duration was defined as time in bed (calculated from bedtime and rise time) minus selfreported sleep latency (Ursin et al., 2005). s during the workweek and during free time were analyzed separately. We decided to focus on sleep durations during the Body mass index (BMI) Obesity prevalence % 20% 15% 10% Figure 1. Mean values and confidence intervals of body mass index (BMI) (top panel) and prevalence of obesity (BMI 30) (lower panel) in different sleep duration groups (workweek data).

3 68 B. Bjorvatn et al Total cholesterol (mmol L 1 ) HDL-cholesterol (mmol L 1 ) Triglycerides (mmol L 1 ) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) Figure 2. Mean values and confidence intervals of total cholesterol, HDL-cholesterol, triglycerides, systolic and diastolic blood pressure in different sleep duration groups (workweek data).

4 Sleep is related to obesity and metabolic measures 69 workweek, because such data are more stable and comprise most days during the week. Statistics Data were analyzed by the Statistical Package for Social Sciences (SPSS) software program for Windows (version 14.0). was dummy coded into six groups: <5, , , , and 9 h). One-way anovas were conducted to investigate differences in BMI, total cholesterol, HDL-cholesterol, triglycerides as well as systolic and diastolic blood pressure between the six sleep duration groups (separate analyses were conducted for workweek and free time). Least significant difference tests were subsequently conducted to test whether any of the sleep duration groups differed from the mean of the sleep duration group, which comprised the reference. Hierarchical linear regression analyses were then conducted to investigate the relationship between sleep duration and BMI (criterion variable). (with as the reference group) was entered in the first step, and gender and smoking (yes/no) were entered in the second step. s during workweek and during free time were analyzed separately. A total of five hierarchical linear regression analyses were subsequently conducted using cholesterol, HDL-cholesterol, triglycerides, systolic and diastolic blood pressure as the criterion variables respectively. In these analyses, sleep duration (workweek) was entered in the first step, gender, smoking and sleep duration free time (continuous) were entered in the second step, and BMI was entered in the third and final step. Finally, crude logistic regression analyses using sleep duration (workweek) as the predictor variable were performed for the following six dichotomized criterion variables: (1) obesity (BMI 30 versus BMI <30), (2) systolic blood pressure (>140 versus 140), (3) diastolic blood pressure (>90 versus 90), (4) total cholesterol (>5 versus 5), (5) HDL-cholesterol (<1 versus 1) and (6) triglycerides (>2 versus 2). Adjusted logistic regression analyses were subsequently performed to control for gender and smoking (for all six criterion variables) and BMI (for the latter five criterion variables). RESULTS Figures 1 and 2 show the mean and confidence intervals for BMI, prevalence of obesity, cholesterol, triglycerides and blood pressure in the different sleep duration groups. Table 1 shows the mean and standard deviations for all variables in the different sleep duration groups during workweek. Table 2 shows the corresponding data during free time. Table 3 shows the results for the hierarchical regression analysis predicting BMI. The results from step 1 show that sleep duration <5, , and 9 h during workweek were positively related to BMI compared with the reference sleep duration. When gender and smoking were entered in step 2, sleep duration did not retain its significant relationship with BMI. For sleep duration during free time it was shown Table 1 Mean values and standard deviations of body mass index, metabolic measures and blood pressure across sleep duration groups (during workweek) < (n ¼ 164) (n ¼ 634) (n ¼ 2543) (n ¼ 3208) (n ¼ 850) 9 h (n ¼ 131) Body mass index (4.30)* (4.04)* (3.75)* (3.74) (4.09) (4.10)* Total cholesterol (mmol L )1 ) 5.69 (1.07)* 5.61 (1.01)* 5.52 (1.00) 5.49 (0.97) 5.48 (1.00) 5.51 (0.99) HDL-cholesterol (mmol L )1 ) 1.25 (0.78)* 1.22 (0.31)* 1.26 (0.34)* 1.30 (0.34) 1.32 (0.33) 1.27 (0.34) Triglycerides (mmol L )1 ) 2.14 (2.05)* 1.86 (1.31)* 1.71 (1.27)* 1.62 (1.19) 1.61 (1.13) 1.66 (0.98) Systolic blood pressure (mmhg) (15.0)* (12.7) (13.6) (14.6) (14.4)* (14.5) Diastolic blood pressure (mmhg) 75.8 (10.7)* 73.9 (9.9)* 73.3 (10.0)* 72.7 (10.4) 72.1 (10.3) 73.0 (10.3) *Significantly different (at the 0.05 level) from the mean score in the sleep duration group. Table 2 Mean values and standard deviations of body mass index, metabolic measures and blood pressure across sleep duration groups (during free time) < (n ¼ 83) (n ¼ 155) (n ¼ 667) (n ¼ 2156) (n ¼ 3055) 9 h (n ¼ 1535) Body mass index (4.45) (4.03)* (4.12)* (3.75) (3.72) (4.07) Total cholesterol (mmol L )1 ) 5.71 (1.01) 5.62 (1.06) 5.57 (1.04) 5.54 (0.99) 5.51 (1.00) 5.51 (0.96) HDL-cholesterol (mmol L )1 ) 1.23 (0.37) 1.22 (0.34) 1.23 (0.34)* 1.27 (0.39) 1.29 (0.34) 1.31 (0.34)* Triglycerides (mmol L )1 ) 2.12 (1.67)* 1.96 (2.05)* 1.86 (1.43)* 1.71 (1.37) 1.65 (1.14) 1.60 (1.05)* Systolic blood pressure (mmhg) (12.9) (12.9) (14.4) (13.7) (14.4) (14.0)* Diastolic blood pressure (mmhg) 73.8 (11.4) 74.0 (9.9) 74.0 (10.6)* 73.0 (10.1) 73.1 (10.2) 72.5 (10.2) *Significantly different (at the 0.05 level) from the mean score in the sleep duration group.

5 70 B. Bjorvatn et al. Step and predictor variable Workweek r 2 Dr 2 B SE b Step ** 0.006** < ** ** * >9 h * Step ** 0.047** Sleep < ** ** >9 h ** (1 ¼ male, 2 ¼ female) ) )0.199** (1 ¼ no, 2 ¼ yes) ) )0.088** Free time Step ** 0.006** Sleep < ** ** ) )0.014 >9 h ) )0.024 Step ** 0.043** Sleep < ** ** ) )0.007 >9 h ) )0.001 (1 ¼ male, 2 ¼ female) ) )0.191** (1 ¼ no, 2 ¼ yes) ) )0.081** Table 3 Hierarchical regression analysis predicting body mass index with sleep duration (for workweek and free time separately), gender and smoking comprises the comparison group. that sleep duration and were positively related to BMI. These results remained unchanged in step 2 where gender and smoking were entered. The results from the hierarchical regression analysis predicting total cholesterol showed that sleep duration <5 and h were positively related to total cholesterol compared with the reference group (Table 4). When gender, smoking and sleep duration free time were entered in step 2 the association between sleep duration and total cholesterol was no longer significant. Total cholesterol was positively associated with male gender and with smoking. The results from step 2 did not change in step 3. BMI was positively associated with total cholesterol. The hierarchical regression analysis predicting HDL-cholesterol showed that sleep duration and were negatively related to HDL-cholesterol (Table 5). However, when gender, smoking and sleep duration free time were entered in step 2, sleep duration was no longer significantly related to HDL-cholesterol. HDL-cholesterol was positively associated with female gender and negatively with smoking. When entering BMI in step 3, sleep duration < became positively associated with HDL-cholesterol. The relationships between gender and smoking and HDL-cholesterol from step 2 were retained. BMI was negatively associated with HDL-cholesterol. The results for triglycerides showed that sleep duration <5, and were positively related compared with the reference group (Table 6). When gender, smoking and sleep duration free time were entered in the regression analysis in step 2, sleep duration < retained its positive relationship and sleep duration became positively related to triglycerides. Triglycerides was positively related to male gender and with smoking. Upon entering BMI in step 3, sleep duration < was no longer significantly related, but sleep duration retained its significant relationship with triglycerides. and gender also retained their relationships with triglycerides, and BMI was shown to be positively related to triglycerides. The hierarchical regression analysis predicting systolic blood pressure showed that sleep duration < was positively and sleep duration was negatively related to systolic blood pressure compared with the reference group (Table 7). When gender, smoking and sleep duration

6 Sleep is related to obesity and metabolic measures 71 Table 4 Hierarchical regression analysis predicting total cholesterol Step and predictor variable r 2 Dr 2 B SE b Step * 0.002* Sleep < * ** ) )0.005 >9 h Step ** 0.029** Sleep < ) ) >9 h (1 ¼ male, 2 ¼ female) ) )0.161** (1 ¼ no, 2 ¼ yes) ** free time ) )0.002 Step ** 0.036** Sleep < ) ) >9 h ) )0.002 (1 ¼ male, 2 ¼ female) ) )0.122** (1 ¼ no, 2 ¼ yes) ** free time BMI ** Workweek data are used. comprises the comparison group. Table 5 Hierarchical regression analysis predicting HDL-cholesterol Step and predictor variable r 2 Dr 2 B SE b Step ** 0.006** Sleep < ) )0.022 ) )0.063** ) )0.052** >9 h ) )0.007 Step ** 0.168** Sleep < ) ) ) )0.019 >9 h ) )0.014 (1 ¼ male, 2 ¼ female) ** (1 ¼ no, 2 ¼ yes) ) )0.124** free time Step ** 0.072** Sleep < * ) )0.013 >9 h ) )0.003 (1 ¼ male, 2 ¼ female) ** (1 ¼ no, 2 ¼ yes) ) )0.149** free time ) )0.004 BMI ) )0.276** Workweek data are used. comprises the comparison group.

7 72 B. Bjorvatn et al. Step and predictor variable r 2 Dr 2 B SE b Step ** 0.006** Sleep < ** ** ** ) )0.000 >9 h Step ** 0.106** Sleep < ** ) ) * >9 h (1 ¼ male, 2 ¼ female) ) )0.325** (1 ¼ no, 2 ¼ yes) ** free time ) )0.022 Step ** 0.084** Sleep < ) )0.004 ) ) * >9 h (1 ¼ male, 2 ¼ female) ) )0.266** (1 ¼ no, 2 ¼ yes) ** free time ) )0.015 BMI ** Table 6 Hierarchical regression analysis predicting triglycerides Workweek data are used. comprises the comparison group. free time were entered in step 2, none of the sleep duration groups were significantly related to systolic blood pressure. Systolic blood pressure was found to be positively related to male gender and negatively related to smoking. When entering BMI in step 3, sleep duration and became negatively associated with systolic blood pressure compared with the reference sleep duration., but not smoking, retained the significant relationship with systolic blood pressure. BMI was positively associated with systolic blood pressure. The results from the hierarchical regression analysis predicting diastolic blood pressure showed that sleep duration <5 and were positively related to diastolic blood pressure compared with the reference (Table 8). When gender, smoking and sleep duration free time were entered in step 2, only sleep duration < retained its significant relationship with diastolic blood pressure. Diastolic blood pressure was positively related to male gender. Upon entering BMI in step 3, only gender retained its significant relationship with diastolic blood pressure. BMI was positively associated with diastolic blood pressure. The results from the logistic regression analysis showed that sleep durations <5, , and 9 h were positively related to being obese compared with the reference sleep duration. These findings were maintained also when controlling for gender and smoking (Table 9). Investigating whether sleep duration was related to having systolic blood pressure >140 mmhg (Table 9) the results from the logistic regression analysis showed that none of the sleep duration groups were significantly related to this criterion variable. However, when controlling for gender, smoking and BMI the results showed that sleep duration and were associated with a decreased risk of having high systolic blood pressure compared with the reference sleep duration (). Investigating whether sleep duration was related to having diastolic blood pressure >90 mmhg (Table 9), the results from the logistic regression analysis showed that only sleep duration < was significantly related to this criterion variable (using sleep duration as reference). When controlling for gender, smoking and BMI, none of the sleep duration groups were significantly related to high diastolic blood pressure. For total cholesterol the results from the logistic regression analysis showed that sleep duration < was a risk factor for having cholesterol >5 mmol L )1, compared with the reference sleep duration h. However, when controlling for gender, smoking and BMI, sleep duration was no longer significantly related to cholesterol (see Table 9). For HDL-cholesterol the results from the logistic regression analysis showed that sleep duration <5, and were associated with increased risk of HDL-cholesterol <1.00 mmol L )1 compared with sleep duration. When controlling for

8 Sleep is related to obesity and metabolic measures 73 Table 7 Hierarchical regression analysis predicting systolic blood pressure Step and predictor variable r 2 Dr 2 B SE b Step ** 0.003** Sleep < ** ) )0.028* >9 h ) )0.007 Step ** 0.097** Sleep < ) )0.013 ) )0.023 ) )0.003 >9 h (1 ¼ male, 2 ¼ female) ) )0.316** (1 ¼ no, 2 ¼ yes) ) )0.028* free time Step ** 0.075** Sleep < ) )0.025* ) )0.027* ) )0.010 >9 h ) )0.011 (1 ¼ male, 2 ¼ female) ) )0.261** (1 ¼ no, 2 ¼ yes) ) )0.003 free time BMI ** Workweek data are used. comprises the comparison group. gender, smoking and BMI, none of the sleep duration groups were a significant risk factor for low HDL-cholesterol (see Table 9). For the last criterion variable, triglycerides, the logistic regression analysis showed that sleep duration <5, and were risk factors for having >2.00 mmol L )1 of triglycerides compared with the reference sleep duration. These associations did not remain significant when controlling for gender, smoking and BMI (see Table 9). DISCUSSION This study demonstrates a relation between short night-time sleep and elevated BMI and obesity. Thus, these data confirm several previous studies in different age groups (Gangwisch et al., 2005; Gupta et al., 2002; Hasler et al., 2004; Heslop et al., 2002; von Kries et al., 2002; Kripke et al., 2002; Locard et al., 1992; Singh et al., 2005; Vioque et al., 2000; Vorona et al., 2005). Also in concordance with previous reports, the relation between sleep duration (during workweek) and BMI shows a U-shaped curvilinear pattern, where BMI increases in subjects with sleep duration below 6 h and above 9 h (Kripke et al., 2002; Singh et al., 2005; Taheri et al., 2004). During free time, sleep durations were clearly prolonged compared with those during the workweek. Still, there was a positive association between short sleep and elevated BMI. The increase in BMI with longer sleep durations was, however, not seen during free time sleep. To our knowledge, this study is the first to demonstrate how metabolic variables like total cholesterol, HDL-cholesterol and triglycerides are associated with sleep duration. A similar relation as for BMI was seen, with higher total cholesterol and triglyceride levels with shorter sleep durations. For HDLcholesterol high values are considered protective for cardiovascular disease. Thus, not surprisingly, an opposite relation compared with total cholesterol was demonstrated, with reduced levels of HDL-cholesterol in subjects with short sleep duration. When controlling for gender, smoking and BMI, most of the relations between sleep duration and these metabolic measures were no longer significant. This may not be surprising, as such variables are linked to levels of serum lipids. Chronic sleep debt has been associated with increased hunger and appetite, especially for calorie-dense foods (Spiegel et al., 2004), which offers one possible explanation for the relation between short sleep duration and these metabolic measures. Both systolic and diastolic blood pressures were higher with shorter sleep durations. When controlling for gender, smoking and BMI, these associations were no longer significant. Surprisingly, sleep duration and became significantly and negatively related to systolic blood pressure compared with the reference sleep duration (). Why

9 74 B. Bjorvatn et al. Step and predictor variable r 2 Dr 2 B SE b Step ** 0.004** Sleep < ** ** ) )0.021 >9 h Step ** 0.083** Sleep < * ) ) >9 h (1 ¼ male, 2 ¼ female) ) )0.293** (1 ¼ no, 2 ¼ yes) ) )0.011 free time Step ** 0.041** Sleep < ) )0.003 h ) )0.014 ) )0.003 >9 h ) )0.002 (1 ¼ male, 2 ¼ female) ) )0.252** (1 ¼ no, 2 ¼ yes) free time BMI ** Table 8 Hierarchical regression analysis predicting diastolic blood pressure Workweek data are used. comprises the comparison group. the associations changed this way when controlling for these variables are unclear. In a recent longitudinal study, short selfreported sleep duration increased the risk of developing hypertension, even after controlling for obesity (Gangwisch et al., 2006). The link between sleep duration and hypertension may go through an increase in sympathetic nervous system activity following sleep deprivation (Gangwisch et al., 2006; Spiegel et al., 1999). One may argue that the differences in BMI present in the different sleep duration groups were not large. Other studies have reported larger variations in BMI across different sleep durations (Singh et al., 2005). Still, we argue that the current findings of sleep duration related changes in metabolic variables and blood pressure underline the importance of obtaining sufficient amounts of sleep, as well as pointing at short sleep duration as a possible health threat. Several large studies show that mortality is increased in short sleepers (Amagai et al., 2004; Kripke et al., 2002; Patel et al., 2004; Tamakoshi and Ohno, 2004), but whether this is related to elevated BMI or other factors remains uncertain. Studies like ours point to possible associations, but do not establish a cause-and-effect relationship. Interestingly, Amagai et al. (2004) showed that short sleep duration in males was associated with elevated risk of death even after controlling for BMI, systolic blood pressure, total cholesterol, smoking habits, alcohol drinking habits, education and marital status. Summarizing the current knowledge in this field, sufficient sleep seems to be of importance for several health-related aspects. A number of limitations of our study need to be addressed. An important limitation is that the blood samples were collected non-fasting. Thus, the values of triglycerides are less reliable, as they could be affected by recent food intake. Total cholesterol and HDL-cholesterol are considered stable variables in relation to food intake, making them more reliable. Another concern is that subjects with high levels of cholesterol may have received medical treatment, to lower pathological values. We do not believe that this is a major problem, as such medications were not commonly used when these data were collected (8 10 years ago). Another limitation of the present study is that we only have data from 40- to 45-year olds, making generalization to other age groups not possible. Gangwisch et al. (2005) showed that the relation between sleep duration and BMI was not significant above 50 years of age, and suggested that this may be because of increased mortality in the obese subjects. Having a narrow age group may also be considered an advantage in epidemiological studies, possibly making the subjects more homogeneous. A limitation of our study is that in the lowest and highest sleep duration groups the number of subjects is fairly low (below 200), making these data less reliable and

10 Sleep is related to obesity and metabolic measures 75 Table 9 (workweek), obesity, blood pressure, cholesterol and triglycerides (logistic regression analyses) Crude Adjusted Obesity (BMI 30) < 1.87 ( ) 1.97 ( ) 1.37 ( ) 1.42 ( ) 1.16 ( ) 1.17 ( ) ( ) 1.32 ( ) 9 h 2.14 ( ) 2.25 ( ) Women 0.84 ( ) Yes 0.69 ( ) Systolic blood pressure >140 mmhg < 1.31 ( ) 0.96 ( ) 0.88 ( ) 0.68 ( ) 0.96 ( ) 0.83 ( ) ( ) 0.90 ( ) 9 h 0.79 ( ) 0.75 ( ) Women 0.43 ( ) Yes 0.89 ( ) BMI 1.14 ( ) Diastolic blood pressure >90 mmhg < 2.22 ( ) 1.56 ( ) 1.06 ( ) 0.81 ( ) 0.98 ( ) 0.84 ( ) ( ) 1.03 ( ) 9 h 1.09 ( ) 1.07 ( ) Women 0.38 ( ) Yes 1.03 ( ) BMI 1.14 ( ) Total cholesterol >5.00 mmol L )1 < 1.65 ( ) 1.28 ( ) 1.17 ( ) 0.94 ( ) 1.09 ( ) 0.98 ( ) ( ) 0.99 ( ) 9 h 0.98 ( ) 0.91 ( ) Women 0.67 ( ) Yes 1.38 ( ) BMI 1.10 ( ) HDL-cholesterol <1.00 mmol L )1 < 2.26 ( ) 1.35 ( ) 1.37 ( ) 0.86 ( ) 1.27 ( ) 0.97 ( ) Table 9 (Continued) Crude Adjusted 0.93 ( ) 1.16 ( ) 9 h 1.21 ( ) 1.26 ( ) Women 0.19 ( ) Yes 1.77 ( ) BMI 1.17 ( ) Triglycerides >2.00 mmol L )1 < 1.93 ( ) 1.18 ( ) 1.48 ( ) 1.00 ( ) 1.19 ( ) 0.93 ( ) ( ) 1.18 ( ) 9 h 1.04 ( ) 1.00 ( ) Women 0.24 ( ) Yes 1.36 ( ) BMI 1.21 ( ) Values are given as OR (95% CI). also limiting the statistical power for these sleep duration groups. Furthermore, sleep was not objectively measured. We calculated sleep duration based on self-reported bedtime and rise times, and subtracted self-reported sleep latency (Ursin et al., 2005). We believe this may give a more reliable measure of sleep duration than just asking subjects to report sleep length. However, awakenings during sleep and early morning awakenings were not reported in our study. This means that the reported sleep duration may be somewhat overestimated. Furthermore, naps during the day were not included in our calculated sleep durations. Such detailed data were given by only 20% of the subjects, making them less reliable, as reported previously (Ursin et al., 2005). We did separate analysis for sleep durations during the workweek and during free time. s during the workweek are more stable and accurate, as bed and rise times do not differ much. During free time, however, the subjects tended to report bed and rise times in whole hours, as these variables probably vary more during free time than during the workweek (Ursin et al., 2005). Workweek data also constitute most of the days of the week, and we believe that these sleep durations may be of most importance when examining possible health risks. In conclusion, our study confirms the clear association between short sleep duration and elevated BMI and obesity. Furthermore, total cholesterol, HDL-cholesterol, triglycerides and blood pressure were associated with sleep duration. These latter associations seemed to be attributed to variables like gender, smoking and BMI.

11 76 B. Bjorvatn et al. ACKNOWLEDGEMENTS The study was supported financially by the Norwegian Research Council and by AstraZeneca AS, Aventis Pharma AS and Scand-Med AS. REFERENCES Åkerstedt, T., Knutsson, A., Westerholm, P., Theorell, T., Alfredsson, L. and Kecklund, G. Sleep disturbances, work stress and work hours: a cross-sectional study. J. Psychosom. Res., 2002, 53: Amagai, Y., Ishikawa, S., Gotoh, T., Doi, Y., Kayaba, K., Nakamura, Y. and Kajii, E. and mortality in Japan: the Jichi Medical School Cohort Study. J. Epidemiol., 2004, 14: Colditz, G. A. Economic costs of obesity and inactivity. Med. Sci. Sports Exerc., 1999, 31: S663 S667. Flegal, K. M., Carroll, M. D., Ogden, C. L. and Johnson, C. L. Prevalence and trends in obesity among US adults, JAMA, 2002, 288: Gangwisch, J. E., Malaspina, D., Boden-Albala, B. and Heymsfield, S. B. Inadequate sleep as a risk factor for obesity: analyses of the NHANES I. Sleep, 2005, 28: Gangwisch, J. E., Heymsfield, S. B., Boden-Albala, B., Buijs, R. M., Kreier, F., Pickering, T. G., Rundle, A. G., Zammit, G. K. and Malaspina, D. Short sleep duration as a risk factor for hypertension: analyses of the first National Health and Nutrition Examination Survey. Hypertension, 2006, 47: Groeger, J. A., Zijlstra, F. R. and Dijk, D. J. Sleep quantity, sleep difficulties and their perceived consequences in a representative sample of some 2000 British adults. J. Sleep. Res., 2004, 13: Gupta, N. K., Mueller, W. H., Chan, W. and Meininger, J. C. Is obesity associated with poor sleep quality in adolescents? Am. J. Hum. Biol., 2002, 14: Hasler, G., Buysse, D. J., Klaghofer, R., Gamma, A., Ajdacic, V., Eich, D., Rossler, W. and Angst, J. The association between short sleep duration and obesity in young adults: a 13-year prospective study. Sleep, 2004, 27: Heslop, P., Smith, G. D., Metcalfe, C., Macleod, J. and Hart, C. Sleep duration and mortality: the effect of short or long sleep duration on cardiovascular and all-cause mortality in working men and women. Sleep Med., 2002, 3: von Kries, R., Toschke, A. M., Wurmser, H., Sauerwald, T. and Koletzko, B. Reduced risk for overweight and obesity in 5- and 6-yold children by duration of sleep a cross-sectional study. Int. J. Obes. Relat. Metab. Disord., 2002, 26: Kripke, D. F., Garfinkel, L., Wingard, D. L., Klauber, M. R. and Marler, M. R. Mortality associated with sleep duration and insomnia. Arch. Gen. Psychiatry, 2002, 59: Locard, E., Mamelle, N., Billette, A., Miginiac, M., Munoz, F. and Rey, S. Risk factors of obesity in a five year old population. Parental versus environmental factors. Int. J. Obes. Relat. Metab. Disord., 1992, 16: Patel, S. R., Ayas, N. T., Malhotra, M. R., White, D. P., Schernhammer, E. S., Speizer, F. E., Stampfer, M. J. and Hu, F. B. A prospective study of sleep duration and mortality risk in women. Sleep, 2004, 27: Singh, M., Drake, C. L., Roehrs, T., Hudgel, D. W. and Roth, T. The association between obesity and short sleep duration: a populationbased study. J. Clin. Sleep Med., 2005, 1: Spiegel, K., Leproult, R. and Van Cauter, E. Impact of sleep debt on metabolic and endocrine function. Lancet, 1999, 354: Spiegel, K., Tasali, E., Penev, P. and Van 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: Spiegel, K., Knutson, K., Leproult, R., Tasali, E. and Van Cauter, E. Sleep loss: a novel risk factor for insulin resistance and type 2 diabetes. J. Appl. Physiol., 2005, 99: Taheri, S., Lin, L., Austin, D., Young, T. and Mignot, E. Short sleep duration is associated with reduced leptin, elevated ghrelin, and increased body mass index. PLoS Med., 2004, 1: e62. Tamakoshi, A. and Ohno, Y. Self-reported sleep duration as a predictor of all-cause mortality: results from the JACC study, Japan. Sleep, 2004, 27: Ursin, R., Bjorvatn, B. and Holsten, F., subjective sleep need, and sleep habits of 40- to 45-year-olds in the Hordaland Health Study. Sleep, 2005, 28: Vioque, J., Torres, A. and Quiles, J. Time spent watching television, sleep duration and obesity in adults living in Valencia, Spain. Int. J. Obes. Relat. Metab. Disord., 2000, 24: Vorona, R. D., Winn, M. P., Babineau, T. W., Eng, B. P., Feldman, H. R. and Ware, J. C. Overweight and obese patients in a primary care population report less sleep than patients with a normal body mass index. Arch. Intern. Med., 2005, 165: