Table 1. Observational studies on the association between sleep and glucose homeostasis in children and adolescents.
| Reference | Study design | Age (years) | n | Sleep assessment | Outcome | Covariates | Main findings |
|---|---|---|---|---|---|---|---|
| Androutsos et al.27 | Cross-sectional | 9–13 | 2026 | Parent reported | HOMA-IR | Age, sex, Tanner stage, WC, parental BMI, SES index and birth weight | A lifestyle characterized by short sleep duration (⩽8.35±0.73 h per day), more screen time (⩾3.61±1.68 h per day) and higher consumption of sugared-sweetened beverages (⩾222.96±222.81 g per day) was associated with increased HOMA-IR (β=0.043, P=0.04). Note that the association between sleep duration alone and HOMA-IR was not reported |
| Armitage et al.15 | Cross-sectional | 13–18 | 18 | 1 night of PSG and prior to PSG 5 nights of sleep diary and actigraphy | HOMA-IR and WBISI | Age, BMI and Tanner stage | WBISI was not significantly associated with sleep characteristics after controlling for Tanner stage as a covariate. Those with highest HOMA-IR (13.1±6.2; n=4) had a significantly higher proportion of NREM1 and lower NREM2–4 than those with moderate HOMA-IR (5.6±1.0; n=8) and low HOMA-IR (3.0±2.2; n=6). Sleep duration did not significantly differ by HOMA-IR category |
| Azadbakht et al.17 | Cross-sectional | 10–18 | 5528 | Parent reported | Fasting glucose | Age, SES index, parents’ education, family history of chronic disease, sedentary lifestyle and BMI | No association was found between sleep duration and fasting glucose in both boys and girls |
| Berentzen et al.18 | Cross-sectional | 11–12 | 1481 | Self-reported questionnaire | HbA1c | Child’s age at the completion of questionnaire and medical examination, height, Tanner stage, screen time, storage time for blood sample and maternal education | No associations were found between sleep quality or duration and HbA1c in both boys and girls |
| Cespedes et al.25 | Longitudinal | 6 months–7 years | 652 | Parent reported | HOMA-IR, fasting glucose and fasting insulin | Age, sex, maternal education, prepregnancy BMI, number of previous pregnancy, age at enrollment, ethnicity, SES, and BMI z-score | After adding BMI z-score to the model, the association between sleep curtailment score and HOMA-IR and insulin fell short of significance. No association was found between sleep and fasting glucose |
| De Bernardi Rodrigues et al.28 | Cross-sectional | 10–19 | 615 (subsample for ISI 81) | Self-reported questionnaire | Fasting glucose, fasting insulin, ISI (via hyperglycemic clamp) and HOMA-IR | Age and sex | In the subsample (n=81), youth with short sleep duration (<8 h per night) had a lower median (IQR) ISI (assessed by hyperglycemic clamp) than those who slept an adequate duration per night (⩾8 h per night) (β=−0.01, 95% CI=−0.01; −0.00, P=0.02). In the large sample, no significant association were found between sleep duration and fasting glucose, fasting insulin and HOMA-IR |
| Flint et al.29 | Cross-sectional | 3–18 | 39 | 1 night of PSG | Fasting glucose, fasting insulin, peak insulin, IGI, HOMA-IR and WBISI | Age, BMI z-score, OSAS and Tanner stage | Compared with children and adolescents with a sleep duration of >6 h, those with ⩽6 h (n=14) had significantly higher fasting insulin (25.7±12.6 μU ml−1 vs 16.0±11.4 μU ml−1, P=0.02), higher peak insulin (226±142.3 vs 113.6±93.5 μU ml−1, P=0.02), higher HOMA-IR (3.3±2.4 vs 5.5±2.9, P=0.01) and lower WBISI (2.2±1.1 vs 7.0±6.2, P=0.01). No significant differences were observed between the two sleep duration groups for fasting glucose, IGI and glucose level 2 h after OGTT. The %REM sleep was significantly lower for the short sleepers (13.5±5.8%) compared with the longer sleeper group (18.6±5.7%) |
| Hitze et al.19 | Cross-sectional | 6–19 | 250 | Self-reported questionnaire (children <11 years were also helped by parents) | HOMA-IR, fasting glucose and fasting insulin | Age and WC z-score | In girls (n=122), sleep duration was negatively correlated with both fasting insulin and HOMA-IR (both, r=−0.20, P=0.05); however, after controlling for WC z-scores the relationship was no longer significant. In boys, no correlations were found between sleep duration and all of the outcome measurements |
| Hjorth et al.26 | Cross-sectional and longitudinal | 8–11 | 723 (subsample for longitudinal sleep data 486) | 8 nights of actigraphy (waist), sleep log (both self-reported and parent reported) and CSHQ (parent reported) | HOMA-IR | Cross-sectional: age, sex, Tanner stage, sex–pubertal status interaction, MVPA, sedentary time, and total physical activity Longitudinal: age, sex, Tanner stage, sex–pubertal status interaction, MVPA, sedentary time, total physical activity, and changes in fat mass index | Cross-sectional data (n=719) revealed that sleep problems noted by parents in the CSHQ were positively associated with HOMA-IR (β=0.007, 95% CI=0.002; 0.013) Sleep duration (n=473) was negatively associated with HOMA-IR (β=−0.080, 95% CI=−0.174; 0.014). Longitudinal data (n=486) showed that changes in sleep duration were negatively associated with changes in HOMA-IR (β=−0.18, 95% CI=−0.36; 0.01) |
| Javaheri et al.30 | Cross-sectional | 15.7±2.1 | 471 | 5–7 nights of actigraphy (wrist) | HOMA-IR and fasting insulin | Age, sex, ethnicity, preterm status, MVPA and WC | Adolescents who slept 10.5 h had the highest predicted HOMA-IR (2.33; 95% CI=1.97; 2.76) while not statistically significant HOMA-IR levels were approximately 30% lower in adolescents who slept 7.75 h and 22% lower in adolescents who slept 5 h (1.78; 95% CI=1.67; 1.91 and 1.93; 95% CI=1.62; 2.30, respectively) |
| Koren et al.31 | Cross-sectional | 8–17 | 62 | 1 night of PSG | OGTT, HbA1c, FSIGT, insulin levels, glucose levels, HOMA-IR,WBISI, IGI and AIRg | Age, sex, Tanner stage, OSAS and BMI z-score (degree of obesity) | In adolescents with obesity, data displayed a U-shaped association between sleep duration, fasting glucose (R2 quadratic=0.201, P=0.002), 2-h glucose (R2 quadratic=0.442, P<0.001) and HbA1c (R2 quadratic=0.200, P=0.002). NREM3 sleep duration was a strong predictor of insulin level as indicated by IGI (R2 quadratic=0.161, P=0.002) and AIRg (R2 quadratic=0.383, P<0.001). A positive correlation was shown between NREM3% of total sleep and 2-h insulin plasma level (r=0.348, P<0.01). A negative correlation was found not only between NREM2 duration and fasting insulin level (r=−0.267, P<0.05) and HOMA-IR (r=−0.282, P <0.05) but also between NREM2% of total sleep and 2-h insulin plasma level (r=−0.280, P<0.05) |
| Lee and Park16 | Cross-sectional | 12–18 | 1187 | Self-reported questionnaire | Fasting glucose | Age, sex, SES, caloric intake and physical activity | No significant association was found between sleep duration and fasting glucose |
| Matthews et al.32 | Cross-sectional | 14–19 | 245 | 7 nights of actigraphy (wrist) and sleep diary | HOMA-IR, fasting glucose and insulin | Age, sex, ethnicity, WC z-score and BMI residual | The HOMA-IR was negatively associated with weekday sleep duration measured by both actigraphy and sleep diary (β=−0.211, 95% CI −0.314; −0.107, P<0.001, β=−0.147, 95% CI −0.249; −0.046, P=0.005, respectively) and total sleep duration measured by both actigraphy and sleep diary (β=−0.202, 95% CI −0.307; −0.096, P<0.001, β=−0.145, 95% CI −0.248; −0.043, P=0.006, respectively) However, HOMA-IR was not associated with weekend sleep duration (β=−0.054, 95% CI −0.158; 0.049, P=0.306). No associations were found between sleep duration and fasting glucose. However, sleep fragmentation was positively associated with fasting glucose (β=0.140 mg dl−1, P=0.035) but not associated with HOMA-IR |
| Navarro-Solera et al.20 | Cross-sectional | 7–16 | 90 | Self-reported questionnaire | Fasting glucose, fasting insulin and HOMA-IR | Age, sex, BMI, physical activity and KIDMED index | No significant association was found between sleep duration and HOMA-IR, fasting glucose or insulin |
| Prats-Puig et al.33 | Cross-sectional | 5–9 | 297 | Self-reported questionnaire with parental help | HOMA-IR | Age, sex, nutrition, physical activity and family history of obesity | No association was found between sleep duration and HOMA-IR in the overall sample. Sleep duration was negatively associated with HOMA-IR in children of a specific phenotype (that is, NRXN3 rs10146997 G) (β=−0.171; 95% CI=−0.276; −0.066) |
| Rey-López et al.21 | Cross-sectional | 12–17 | 699 | Self-reported questionnaire | HOMA-IR | Age, sex, SES and MVPA | No association was found between sleep duration and HOMA-IR |
| Spruyt et al.22 | Cross-sectional | 4–10 | 107 | 7 nights of actigraphy (wrist) | Glucose and insulin | Age, sex, ethnicity and BMI z-score | No associations were found between sleep duration and glucose or insulin concentrations |
| Sung et al.23 | Cross-sectional | 10–16 | 133 | 7 nights of actigraphy (wrist) accompanied by sleep log and questionnaires (parent reported and self-reported) | Fasting glucose and HOMA-IR | Age, sex, ethnicity, SES, BMI z-score and OSAS | No associations were found between sleep duration and HOMA-IR or fasting glucose |
| Tian et al.34 | Cross-sectional | 3–6 | 1236 | Parent reported | Fasting glucose | Age, sex, birth weight, gestational age, SBP, parent’s education, BMI z-score, WC, diseases in the past month, breastfeeding at 6 months, diet and nutrition, screen time and physical activity | A negative association between sleep duration and fasting glucose was found (β=−0.043, s.e.=0.021, P=0.04). An increased risk of hyperglycemia (⩾100 mg dl−1) for those sleeping ⩽8 h compared with those sleeping 9–10 h was observed (OR=1.64, 95% CI=1.09; 2.46). When stratified by weight status, the association was only present in obese children (OR 2.15, 95% CI=1.20; 3.84) |
| Turel et al.24 | Cross-sectional | 10–17 | 94 | Fitbit activity watch accompanied by a sleep log | Fasting glucose, fasting insulin and HOMA-IR | Age, sex, SES, BMI z-score, WC and medications | No association was found between sleep duration and HOMA-IR, fasting glucose or fasting insulin |
| Zhu et al.35 | Cross-sectional | 13.1±3.3 | 118 | 1 night of PSG | OGTT, insulin, glucose, ISOGTT and ISSI-2 | Age, sex, BMI z-score, Tanner stage and OSAS | The 2-h glucose was negatively associated with total sleep time and sleep efficiency (β=−9.96 × 10 −4, s.e.=3.23 × 10−4, P<0.001 and β=−0.005, s.e.=0.002, P=0.011, respectively). A positive association was observed between ISOGTT and sleep efficiency and NREM3% of total sleep time (β=0.013, s.e.=0.005, P=0.016, and β=0.024, s.e.=0.009, P=0.012, respectively). ISOGTT was also negatively associated with NREM1% of total sleep time (β=−0.058, s.e.=0.025, P=0.021) ISSI-2 was positively associated with both total sleep time and sleep efficiency (β=0.002, s.e.=0.001, P=0.008, and β=0.010, s.e.=0.004, P=0.014, respectively) |
Abbreviations: AIRg, acute insulin response to glucose; BMI, body mass index; CI, confidence interval; CSHQ, children’s sleep habits questionnaire; FSIGT, frequently sampled intravenous glucose tolerance test; HbA1c, glycated hemoglobin; HOMA-IR, homeostasis model assessment of insulin resistance; IGI, insulinogenic index; IQR, interquartile range; ISI, insulin sensitivity index; ISOGTT, insulin sensitivity index for oral glucose tolerance test; ISSI-2, insulin secretion sensitivity index 2; KIDMED index, Mediterranean Diet Quality Index for Children and Adolescents; MVPA, moderate-to-vigorous physical activity; NREM, non-rapid eye movement; OGTT, oral glucose tolerance test; OR, odds ratio; OSAS, obstructive sleep apnea syndrome; PSG, polysomnography; REM, rapid eye movement sleep; SBP, systolic blood pressure; SES, socioeconomic status; WBISI, whole-body insulin sensitivity index; WC, waist circumference. Note: Main findings from analyses of sleep duration are treated as categorical variables and presented as mean±s.d. unless stated otherwise. Main findings represent the most adjusted models unless stated otherwise.