Cross-sectional and longitudinal associations between sleep duration, sleep quality, and bone stiffness in European children and adolescents

The IDEFICS/I.Family cohort is a multicentre population-based study which was conducted in eight European countries: Belgium, Cyprus, Estonia, Germany, Hungary, Italy, Spain, and Sweden. The main aim of this cohort was to investigate and prevent diet- and lifestyle-related diseases and disorders among European children and adolescents. The baseline survey (IDEFICS study) was conducted between September 2007 and May 2008, two or more communities in each country whose socio-demographic profile and infrastructure were similar and typical for their region were selected. In total, 16,229 children aged 2 to 9 years participated and fulfilled the inclusion criteria. Two follow-up examinations were further conducted with similar examination modules after 2 years (IDEFICS study) and 6 years (I.Family study) at 2009/2010 and 2013/2014, respectively. The study was conducted according to the standards of the Declaration of Helsinki. Ethical approval was obtained from the ethics committees by all eight study centers. Before children entered the study, parents provided written informed consent. Additionally, children aged 12 years and older gave simplified written consent and children younger than 12 years gave oral consent before each examination. More details on the study design of the IDEFICS/I.Family cohort were published previously [18].

We included data from who participated in the 2009/2010 (in the following referred to as baseline) and 2013/2014 surveys (in the following referred to as follow-up) of the IDEFICS/I.Family cohort, since the sleep variables of interest were only assessed at these two time points. QUS measurements were conducted as an optional module in each survey, with 6886 participants at baseline (50.6%) and 2892 participants at follow-up (30.3%), respectively. There were no significant differences between participants and non-participants for main demographic characteristics (i.e., sex, age, and family socioeconomic status (SES)). We further excluded children and adolescents who (1) had implausible QUS measurements, i.e., absolute difference of SI between the right and left foot exceeded the 97th percentile (39 units) based on total 8280 measurements; (2) had an indication of impaired bone health at baseline, i.e., with disease or receiving medical treatments affecting the bone; (3) had implausible nocturnal sleep duration, i.e., more than 16 h per night or less than 5 h per night; and (4) had incomplete data of covariates, i.e., PA, SB, and pubertal status. Finally, a total of 4871 children at baseline were eligible for the cross-sectional analysis. Of these, 861 participants with complete information at both baseline and follow-up were included in the longitudinal analysis.

General questionnaires related to lifestyle behaviors were collected in the IDEFICS/I.Family cohort. Parents were asked to complete the questionnaires on behalf of their children younger than 12 years old, while for older children and adolescents, the questionnaire was self-reported.

The duration of nocturnal sleep and daytime napping (hours and min) at weekdays/school days and weekend days/vacations was separately recorded. The average of daily nocturnal sleep duration was calculated for each child as follows: (nocturnal sleep duration on weekdays/school days × 5 + nocturnal sleep duration on weekend days/vacations × 2) / 7. The value of nocturnal sleep duration for each child was further transformed to an age-specific z-score based on the reference population from I.Family study, in order to consider both children and adolescents. The maximum number of observations with plausible information on age and nocturnal sleep duration was included to obtain the best possible estimates; the years of age were used as categories to calculate age-specific means and standard deviations. Analogously, the average of daily napping time was calculated and expressed as 10 min/day in order to better interpret the regression coefficients.

We further calculated the total sleep duration by adding up the nocturnal sleep and daytime napping. According to the sleep recommendation from the National Sleep Foundation (NSF) [19], we used the definition of short and long sleep duration as < 10 h/day and ≥ 13 h/day for pre-school children aged 2 to < 6 years, < 9 h/day and ≥ 11 h/day for primary school children aged 6 to < 12 years, < 8 h/day and ≥ 10 h/day for adolescents aged 12 to 15 years, respectively.

Additionally, participants were asked to report their typical sleeping habits and daytime condition as follows: (1) do you have a regular bedtime routine (yes = 1 and no = 0); (2) do you have trouble getting up in the morning (yes = 0 and no = 1); and (3) do you have difficulty falling asleep (yes = 0 and no = 1). Similar items were used previously in other large population-based study [20]. We calculated a cumulative score as the sum of these three items (range from 0 to 3); individuals scoring 3 were considered as having good sleep quality.

PA level of the participants was recorded as weekly duration of participation in sports clubs. They were asked to report whether the child was a member of a sports club (yes or no); if the answer was yes, then they had to report how many hours and minutes per week. The weekly time spent at sports clubs was calculated by adding hours and minutes and expressed as hours per week (h/w). SB level of the participants was recorded as daily duration of total screen time. Screen time was calculated from reported usual duration of the child watching TV/videos/DVDs and playing computer/game console on a normal weekday and weekend day. For both questions, six response categories were offered and converted into the following scoring system: not at all = 0, < 30 min = 1, < 1 h = 2, 1– < 2 h = 3, 2–3 h = 4, and > 3 h = 5. Each screen-based SB was calculated separately for weekdays and weekend days by adding up the converted responses of questions as follows: screen-based SB on weekdays/school days × 5 + screen-based SB on weekend days/vacations × 2. The weekly duration of screen time was the total duration of these two screen-based SB and expressed as hours per week (h/w).

The age, sex, and SES of participants were obtained from parental questionnaire. The highest education of parents was obtained as a proxy indicator for SES according to the International Standard Classification of Education (ISCED) (low: ISCED levels 0–2; medium: ISCED levels 3–4; high: ISCED levels 5 and higher) [21]. The pubertal status was self-reported by children aged 8 years and older, and was defined as pre-pubertal or pubertal based on voice change in boys and first menstrual period in girls [16]. Sunlight exposure accounts for the main source of vitamin D synthesis [22]; therefore, mean daylight duration (± 0.1 h) for each examination month in each location was calculated using astronomical tables as a proxy for vitamin D level.

SI was measured on the left and right calcaneus using QUS (Achilles Lunar Insight TM GE Healthcare, Milwaukee, WI, USA); the reliability study and the methodology of QUS device have previously been described in detail [23]. The measurements and quality control were performed by trained nurses based on the standard operating procedures provided by the manufacturer. The foot was positioned using two different sizes of adapters for participants to keep their calcaneus properly. Before recording each measurement, the preview image of the calcaneus and the region of interest were required to display, in order to avoid measurement error caused by incorrect locations, e.g., the growth plate. The SI was calculated automatically as a percentage (units) based on normalized and scaled values of broadband ultrasound attenuation (BUA, dB/MHz) and speed of sound (SOS, m/s) according to: BSI = (0.67*BUA) + (0.28*SOS) − 420. The value of BUA represents the spatial orientation of the bone trabeculae and increases with greater trabecular complexity. The value of SOS represents the velocity of sound traveling through the bone and increases with greater structures density, and the combination of SOS and BUA was slightly better at predicting bone strength than either parameter alone [24]. The mean SI of the left and right calcaneus measurements was calculated and used in the statistical analysis. For each individual, the SI percentile was calculated additionally as outcome according to age, sex, and height based on the IDEFICS/I.Family reference population. The processing method and the first descriptive results of SI percentile values in the IDEFICS study can be found elsewhere [23], and we additionally provided the sex-specific reference curves for SI percentiles by age for average children based on the 50^th height percentile in the IDEFICS/I.Family study in Online Resource (Figure S1).

Physical examinations including body weight and height were measured by trained nurses based on the standard operating procedures. Weight (kg) was measured to the nearest 0.1 kg using the Tanita scale (BC420 MA for children and BC418 MA for adolescents, Tanita Europe GmbH, Sindelfingen, Germany). Height (cm) was measured to the nearest 0.1 cm using the calibrated stadiometer (Seca 225/213 stadiometer, Birmingham, UK). All examinations were in light clothing without shoes. Body mass index (BMI) was calculated as weight (kg) divided by height (m) squared; the values were transformed to age- and sex-specific z-scores based on Cole et al. [25]

Associations between sleep duration, sleep quality, and SI were investigated using linear mixed-effects models, and a random effect for country was added to account for cluster effects. All models were adjusted for age, sex, family SES, daylight duration, and BMI z-scores in each survey. The exposures of sleep duration at baseline and/or follow-up were included as continuous variables (i.e., nocturnal sleep duration z-scores and daytime napping duration) and a dichotomous variable (i.e., fulfilling the sleep recommendation or not) in separate models, to additionally investigate the benefit of meeting sleep recommendation on SI percentiles. The exposure of sleep quality at baseline and/or follow-up was included as a dichotomous variable (i.e., good and poor). The outcome for cross-sectional analyses was baseline SI percentile, with additional adjustments for baseline duration of PA and SB. The outcome for longitudinal analyses was SI percentile at follow-up, with additional adjustments for SI percentile at baseline, pubertal status at follow-up, and average durations of PA and SB at both surveys. Since sleep duration and sleep quality may have interactive effects on SI, we further stratified the whole group by total sleep duration and quality, to investigate the different effects of interests across stratifications.

All the statistical analyses were carried out with the Statistical Analysis System (SAS) software package (Version 9.4; SAS Institute, Cary, NC). Regression coefficients (β), 95% confidence intervals (95%CIs), and p values were calculated. The significance level was set at α = 0.05 (2-sided tests); multiple testing was further adjusted and cited in footnotes if p values exceed 0.05 according to Holm's sequential Bonferroni procedure [26].

At baseline, 4871 participants consist of 742 pre-school children and 4129 primary school children were included in cross-sectional analyses. Of these, 861 participants consist of 352 primary school children and 509 adolescents from 5 participating centers were further included in longitudinal analyses after a 4-year follow-up. The proportion of boys (47.0% vs. 50.5%, p = 0.06) and mean age (8.28 years vs. 8.20 years, p = 0.19) of longitudinal analytic sample were comparable to children who did not provide follow-up data, but less participants were classified as low (8.4% vs. 9.1%) and high SES (30.0% vs. 38.1%), p < 0.001. More details of demographic characteristics are shown in Table 1; additional results of Pearson’s correlations and 95%CIs between co-variables, sleep exposures, and bone stiffness index can be found in Online Resource (Table S1).

At baseline, the average daily nocturnal sleep duration was 9.89 h; the median value of reported daily daytime napping was 60 min. There were 11.2% participants defined as short sleep duration and 6.0% participants defined as long sleep duration. Overall, 44.6% of the participants were defined as poor sleep quality. At follow-up, the average daily nocturnal sleep duration was 9.10 h; the median value of reported daily daytime napping was 31.4 min. The proportions of short and long sleep duration were 16.7% and 8.8%, respectively; 58.8% of participants were reported having poor sleep quality. Furthermore, the average bone stiffness index was 81.91 units at baseline and 91.22 units at follow-up. More details regarding exposures, outcomes, and covariates among pre-school children, primary school children, and adolescents in each survey were shown in Table 2.

In the whole group, no cross-sectional associations between nocturnal sleep duration z-scores, sleep quality, and SI percentiles were observed. Daytime napping duration was positively associated with SI percentiles (β = 0.78, 95%CI: 0.43, 1.14, p < 0.001), and this association was independent of negative effect of screen time and positive effect of sports club on SI percentiles as shown in Online Resource (Table S2). After stratifying by total sleep duration categories, we found that nocturnal sleep duration z-scores were positively associated with SI percentiles in participants with adequate sleep duration (β = 1.81, 95%CI: 0.55, 3.07, p = 0.005), and effect size was even larger, however not statistically significant, in participants with long sleep duration (β = 3.71, 95%CI: − 2.28, 9.70, p = 0.22). Nevertheless, the positive association between daytime napping and SI percentiles was statistically significant only in participants with adequate sleep duration (β = 1.19, 95%CI: 0.76, 1.61, p < 0.001), and was more pronounced but not statistically significant in participants with short sleep duration (β = 1.27, 95%CI: − 1.14, 3.69, p = 0.30) (Table 3).

After 4 years, we found that only follow-up daytime napping was positively associated with follow-up SI percentiles (β = 0.84, 95%CI: 0.14, 1.53, p = 0.02), and it was not influenced by the positive effect of sports club as shown in Online Resource (Table S3), but only existed in participants with short sleep duration after stratification (β = 2.42, 95%CI: 0.98, 3.85, p = 0.001). Even though there were no statistically significant associations between sleep quality and SI percentiles in all groups after 4 years, we observed that participants with poor sleep quality tended to have lower SI percentiles compared to their counterparts in short (β = − 10.17, 95%CI: − 21.94, 1.61, p = 0.09) and long sleep duration group (β = − 4.99, 95%CI: − 22.75, 12.78, p = 0.56), respectively (Table 4). The combined effect size was − 11.61 (95%CI: − 21.09, − 2.12, p = 0.02) if we merged short and long sleep duration into one group with participants who did not fulfill the sleep recommendation.

At baseline, there was no statistically significant difference on SI percentile between children who fulfilled the sleep recommendation and who did not. However, extreme total sleep duration at baseline predicted lower SI percentiles after 4 years compared to their counterparts when participants simultaneously had poor sleep quality at baseline (β = − 8.09, 95%CI: − 13.39, − 2.79, p = 0.003) (Table 5). As shown in Online Resource (Table S4), four interaction terms in the cross-sectional model and eight in the longitudinal model were separately introduced to the main effects in the whole group, one of which was statistically significant, i.e., the interaction between extreme total sleep duration at baseline and average screen time (β = − 0.52, 95%CI: − 1.02, − 0.03, p = 0.04). Specifically, screen time has a stronger detrimental effect on SI percentile in children with extreme total sleep duration.