Several studies have demonstrated an association between sleep duration and BMI, but without consensus with regard to the optimal sleep duration with respect to BMI and waist circumference. A methodological problem is represented by the fact that some studies have defined short sleep duration as ≤6 hours while others as ≤5 hours. Moreover, the reference sleep group has in some studies had a sleep duration between 7 and 8 hours 
, while others have used 7 to 9 hours sleep as the reference group 
In this latter part of The Tromsø Study, we found that even if 7–7.9 hour sleepers had the lowest WHR and the lowest prevalence of abdominal obesity, the optimal sleep length with regard to all body composition measures was actually between 8 and 9 hours, when all the life style factors, health factors, and biological factors were accounted for in the multivariate regression analyses. As shown in , exercise, education, self-evaluated good health, smoking and HDL-cholesterol level lowered the risk of being overweight. On the contrary, alcohol use, heart disease and total cholesterol and triglyceride levels were associated with increased overweight risk. More or less the same pattern was found for WHR and male abdominal obesity (results are not shown). Magee et al. 
underlined the need to control for potentially confounding variables in this type of studies. In their study, Fogelholm et al. 
included sex, age, mental health, smoking and education as confounding factors, while Kohatsu et al. 
adjusted for sex, age, educational achievement, physical job demand, household income, depressive symptoms, marital status, alcohol consumption and snoring. In our study, we also included physical activity in leisure time, self-rated physical health, former and current heart disease and diabetes, blood pressure, use of medication, HSCL-10 score, coping, sleep variables and biological factors. One possible explanation for the difference in optimal sleep length in our study compared these other studies 
could be that we have included more relevant variables, which is a major strength of our study. Other strengths of the present study were the high number of participants and that height, weight, and measures of waist and hip were done by trained personnel, as self-reported height is often overestimated and weight underestimated 
, which could lead to underestimation of BMI.
More people with elevated body weight were found in the short sleeping group, and the risk of having BMI≥25 kg/m2 was about 80% higher in this group compared to moderate sleepers. Short sleeping also doubled the abdominal obesity risk for men, but not for women.
Roenneberg et al. found that beyond sleep duration, social jetlag was associated with increased BMI 
. This was a large epidemiological study including more than 64,000 people aged 16–65 years, but the study did not include important predictors of body weight like educational level, health factors or diet. In another study, Baron et al. found that calories consumed after 8:00 PM may increase the risk of obesity, independent of sleep timing and duration 
. In our study we did not find an association between elevated BMI and social jetlag or chronotype. In the present study we did not report data on diet, anyway some of the biological variables that were included (triglycerides, total cholesterol, HDL-cholesterol and glucose) could –at least in part- be taken to reflect participants' diet.
The associations between sleep duration and BMI are far from fully understood, and studies have looked for explanations other than short sleeping, sleepiness-related inactivity and excess energy intake. Short sleep (<5 h) has been associated with reduced levels of leptin (“anorexigenic hormone”) and elevated levels of ghrelin (“hunger hormone”), which could increase appetite and also BMI 
. Spiegel et al. found the same leptin and ghrelin changes, together with alterations in glucose metabolism in persons sleeping <4 hours 
. In our study, we excluded persons who reported sleeping <3 hours since BMI elevation always develops over time, and habitual sleeping <3 hours per night is likely to occur only over a short period of time. We have no data on appetite hormone levels, but the significantly higher BMI levels among short sleepers was found after controlling for confounding factors including (non-fasting) total-cholesterol, HDL-cholesterol, triglycerides and glucose levels.
In our study, associations between short sleep and elevated risk of overweight (BMI≥25 kg/m2) was found for both genders, and the association was slightly stronger for men. Studies have shown gender differences in sleep patterns, with women sleeping longer than men 
, as we have previously reported 
Stranges et al. found no association between short sleep duration and increased BMI or waist circumference in prospective analyses of data from the Whitehall II Study 
. This is comparable to the findings from the CARDIA Sleep Study where there were no longitudinal associations between sleep measurements and BMI 
It has been pointed out that although there has been evidence for associations between obesity and short sleep duration, most short sleepers are unlikely to be obese, and most obese people are unlikely to be short sleepers 
. In our study, as well as most of the studies referred to by Horne, the number of participants was high, sleep duration was self-reported by a questionnaire and BMI was measured. We found that people with normal weight slept significantly longer than those who were in the overweight group, but the difference was only about 5 minutes. Even if an inverse relationship between sleep duration and the risk of having an elevated BMI was found for both genders in our study, the difference in age corrected BMI was very small (1 BMI unit for men and 0.8 BMI units for women), similar to the difference in BMI between sleep groups referred by Horne. This difference in BMI is possibly not large enough to have a clinical significance.
The present study had some limitations that should be mentioned. The MCTQ is based on self-reported data, and the recordings for sleeping behavior were the participants' own estimates and probably not the exact times. It is important to keep this in mind when analyzing and comparing the sleep duration data. Cournot et al. found independent and positive associations between BMI and the number of naps per week in women 
. Napping during the day was not reported in our study, but in a study by Ursin et al., 26.5% of the men and 28.9% of the women reported napping several times a week or more 
. Women had slightly longer naps than men, 48 minutes in women versus 45 minutes in men. We have no reason to believe that this phenomenon was less common among the participants in the Tromsø study than in the Hordaland study. If we anticipate that the napping is more common in short sleepers, and that napping was more common and slightly longer in women, short sleep would be even more strongly associated with an increased BMI in women. Unfortunately, we do not know if this is the case in our study.
The Tromsø study had low response rates in the age group 30–34 and in subjects aged 70 years or more, but findings from a study on non-responders evidenced that selection according to sociodemographic variables had little impact on prevalence estimates for participants compared to non-participants 
. This, together with our exclusion of participants over 65 years, suggests that the low response rate in the age group 30–34 years had a minor impact on the health variables in our study.
A possible hypothesis for the difference in optimal sleep duration in this study compared to other studies is that subarctic people have different sleeping patterns than other populations because of the extreme photic conditions in this area, and that this could have an impact on BMI. This ought to be investigated in further longitudinal studies in subarctic populations.