This study adds to the limited literature investigating relationships between measures of body composition other than BMI in older adolescent populations. In this cohort of older adolescents, we found that longer sleep duration was associated with lower levels of all body composition variables, including lower levels of fat free mass, in boys only. While significant results were seen for measures of central adiposity (waist circumference and waist-to-height ratio), the results were most pronounced for FMI, a measure of total adiposity. A one hour decrease in sleep on weeknights was associated with a 4.5% increase in FMI, after adjustment for age, deprivation, ethnicity, whether participants met recommendations for fruit and vegetable consumption and the number of screens in the bedroom. A major strength of this study is the use of several measures of body composition. The majority of previous studies in adolescents have used only BMI  and have categorised both sleep and BMI, with no consistent cut-points for short sleep or overweight being used, which makes direct comparisons of effect size between studies difficult. However, results from studies in younger children suggest that boys may be at greater risk of sleep-associated obesity than girls . Our results for BMI are similar to the only other study in adolescents  where data were analysed in a similar way, where an hour decrease in sleep duration in adolescents boys was associated with a 0.1 increase in BMI z score.
Fewer studies have investigated the relationship between sleep and other measures of body composition and results from studies in adolescents have been inconsistent, with some studies finding significant results for both boys and girls and some only in boys . However none of these studies have investigated these relationships in older adolescents, who have more autonomy with regards to lifestyle choices, including bedtimes and food choice, compared to younger children . Even the most recent studies in this age group that have advantages such as using longitudinal rather than cross-sectional data , or have large, nationally representative samples  rely on using BMI measures only. While BMI is a measure of excess weight and not necessarily excess body fat, it has been shown to have good specificity , and may be a good proxy for fat mass in obese adolescents . However, in normal-weight or overweight adolescents higher BMI may be more reflective of increased lean mass rather than fat mass . For this reason BMI may not be a good proxy for body fat in this cohort, as a very small number of participants were obese. More consistent evidence has been found between sleep duration and more direct measures of fat mass as assessed by BIA or DXA (Duel-energy X-ray Absorbance) , however all of these previous studies have been in younger children. Therefore our results provide unique insight into the relationships between sleep and body composition in older adolescents.
Data were collected during the New Zealand summer and autumn, meaning that there were differences in daylight hours over the course of data collection. For example, in February, sunrise in the Otago region is at around 7 am and sunset at around 9 pm, whereas sunrise in June is at around 8 am and sunset at 5 pm. While it is possible that adolescent sleep patterns may be related to hours of daylight, previous research in adolescents suggests that factors other than hours of daylight, such as hormonal and physical changes, and shifting of circadian rhythms may potentially increase their risk of not gaining sufficient sleep [6, 13].
In agreement with previous research [4, 5] we found that the relationship between sleep and body composition was independent of ethnicity and deprivation. While ethnicity and deprivation were themselves significant in the models, they made no meaningful difference to effect sizes for relationships between sleep and body composition. However, the low numbers of Māori and Pacific participants and those from more deprived schools may have impaired our ability to detect the true effects of these variables. We also found that whether participants met recommendations for fruit and vegetable consumption and number of screens in the bedroom played no significant part in any relationships.
While we found significant negative relationships between sleep duration and all measures of body composition in boys, the largest effect sizes were seen for specific measures of fat and lean mass, rather than overall body composition. Our results show that while longer sleep duration was associated with both lower lean and fat mass, the effects were most pronounced for fat mass. Previous research in twin cohorts has shown that while an individual’s absolute amount of lean tissue may be determined in the fetal period, changes in the proportion of fat to lean tissue are influenced by their physical and social environment [31, 32]. Changes in muscle mass can only be achieved by hypertrophy, whereas adipocytes can undergo both hypertrophy and hyperplasia, meaning that, in agreement with our findings, an individual’s environment may influence their fat mass more than their lean mass .
It is outside the scope of this study to determine the mechanisms by which short sleep duration affects body composition. Results from previous studies indicate that reduced sleep may increase dietary intake due to increased wakefulness leading to more opportunity to eat . Hormonal changes may also play a role. Lower growth hormone levels have been observed in those who sleep less, which could result in adolescents not attaining their optimal genetically determined height, resulting in a higher BMI . Sleep deprivation has also been shown to have adverse effects on leptin and ghrelin levels, leading to increased appetite [7, 8] but we found no evidence that food choice, including increased frequency of high sugar and/or high fat foods, influences relationships between sleep and body composition in this cohort. Sleeping less may also lead to fatigue or changes in thermoregulation resulting in decreased energy expenditure .
A limitation of this study is that sleep data were based on self report and questions used in this study asked about time of going to bed and getting up and therefore our measures of sleep duration may reflect time at rest plus sleep time, rather than sleep per se. However, the questions used in this survey have been validated against both sleep diary and actigraph measurements  and have shown acceptable relative validity for use in large studies of adolescents. All questionnaires used in this study were pilot tested in the sample population before use in this study to ensure they were suitable for use in OSSLS2.
If the errors in self-reported sleep data are non-differential then this would lead to attenuation of results, rather than overestimation of relationships . Similar relationships between adolescent sleep duration and body composition measures have been seen in studies using self or parental reported sleep duration compared to those using objective measures based on actigraphy [6, 35, 36]. Reported sleep times and the shorter sleep times for boys found in this cohort are similar to those found in a recent meta-analyses of adolescents from 23 countries , where adolescents aged fifteen to eighteen years slept for around 9 to 10 hours each day. The amount of catch-up sleep reported at weekends is very similar to the 40 minutes reported in the only other large-scale study of sleep in New Zealand adolescents . We also had no measure of participants’ pubertal status, as it was not feasible to collect accurate information on this within the current study design, but as the age of participants ranged from fifteen to eighteen years of age, it is likely that the majority of participants were pubertal or post-pubertal. We did not collect information on depression, which has been shown to play a role in the relationship between sleep and body composition .
The use of BIA measurements is not without limitations. Estimates of fat and fat-free mass are calculated using sex and aged based equations . Because of the nature of carrying out school-based research, particularly the logistical demands, in samples of this size, BIA measurements were carried out throughout the school day. Therefore BIA estimates may be affected by recent exercise, food consumption or hydration status.
However, it would be impractical to use other measures such as DXA in large population studies and fat mass and fat-free mass estimates obtained from BIA measurements in New Zealand adolescents have been found to be highly correlated with DXA measurements . The use of standardised operating procedures for body composition measuring, intensive training in measurement, and quality control procedures in place during the study would ensure that this error is minimised.
One strength of this study is the use of an FFQ that has been developed and validated for use specifically in this study. This means that the food data collected is robust . However, as the FFQ used is non-quantitative, we can only be sure that frequency of consumption of foods does not play a part in these relationships and the role of particular foods cannot be assessed independently of energy intake. While using more intensive methods such as food diaries would provide data on nutrients and energy intake, compliance with such methods in this age group is low, and results of our pretesting showed that almost 50% of adolescents failed to complete a four day food record . Another limitation of the current study is that we were unable to adjust for physical activity in these analyses. Therefore we cannot rule out the possibility that energy imbalance, rather than food choice itself, is an important moderator in these relationships, as has been found in previous studies [29, 36], or that energy balance may be a mediator of the sleep-obesity pathway. The 78% participation rate in OSSLS2 shows that the current sample is representative of the Otago region but it is not nationally representative due to the low prevalence of those of Māori or Pacific origin, who are also more likely to be of lower socio-economic status and to have higher BMIs .
In conclusion, the relationship between shorter sleep duration in boys and a more detrimental body composition, both in terms of lean and fat mass, indicates that sex specific factors may play a role in this relationship in older adolescents. Despite the fact that the study population may not be nationally representative and that the majority of the data was collected by questionnaire, our results reflect international results, including results from studies using more intensive data collection methods. Further research is needed to identify what sex specific factors are responsible for these findings so that targeted approaches to help reduce rates of overweight and obesity could be identified.