Modifying effects of compensatory sleep on elevated blood pressure related to school-day sleep patterns Authors: Jieyu Liu, Xinli Song, Manman Chen, Yi Zhang, Jianuo Jiang, Ruolin Wang, Ziqi Dong, Li Chen, Yang Qin, Tongjun Guo, Zhiying Song, Wen Yuan, Xinxin Wang, Jun Ma, Yanhui Dong, Yi Song Published: 2025 ABSTRACT AbstractBackgroundTo explore the complex effects of school-day sleep patterns on elevated blood pressure (EBP), with a particular focus on the role of compensatory sleep.MethodsUtilizing propensity score matching, 15,612 children aged 7–12 years were obtained from the Chinese National Survey on Students’ Constitution and Health in 2019. School-day wake-up and bedtimes were gathered by questionnaires, categorizing sleep midpoint into morning, intermediate, and evening phenotype and classifying sleep duration as “inadequate,” “moderate,” or “excessive.” EBP was defined as systolic blood pressure (SBP) and/or diastolic blood pressure (DBP) at or above the 95th percentile for their respective sex, age, and height. The nonlinear relationships between sleep midpoint and sleep duration with EBP were explored by restricted cubic splines. Logistic regressions were utilized to assess the school-day sleep patterns, combining sleep midpoint and duration on EBP. Stratified analyses were conducted to examine the roles of compensatory sleep, and the population attribution fraction was calculated to describe the contributions of compensatory sleep to EBP related to school-day sleep patterns.Results15,612 children aged 7–12 from 504 schools nationwide were included. An early sleep midpoint and short sleep duration significantly increase the odds of EBP, but the risks posed by late sleep midpoint and late bedtimes should not be overlooked either (P for non-linear < 0.05). Both morning-types (odds ratio, OR = 1.10, 95% CI = 1.02–1.18) and evening-types (OR = 1.11, 95% CI = 1.02–1.21), as well as inadequate sleep duration (< 9 h), exhibited positive associations with EBP. Even “morning larks” phenotype with sufficient sleep duration was associated with EBP. Compensatory sleep may influence the odds of EBP in sleep-deprived “morning larks” and “night owls.”ConclusionsPrioritize the assurance of ≥ 9 h of sleep on schooldays, promote the adoption of an intermediate-phenotype, and discourage excessively early wake-up times. Compensatory sleep appears to modify the odds of EBP in sleep-deprived “morning larks” and “night owls.”Supplementary InformationThe online version contains supplementary material available at 10.1186/s12916-025-04439-7.Keywords: Sleep midpoint, Sleep duration, Blood pressure, Compensatory sleep, Children Background Sleep, occupying a substantial one-third of an individual’s lifespan, has long been a strikingly underappreciated aspect within the medical field and its associated disorders [1]. Sleep represents a sophisticated neurobehavioral condition that brings about extensive alterations in both cerebral activities and overall physiological functions [2]. This vital physiological process, governed by the body’s endogenous circadian rhythm, is instrumental in maintaining the intricate balance between sleep and wakefulness. Contemporary societal factors, including economic pressures, lifestyle changes, work-related stressors, and evolving living environments, have profoundly disturbed sleep patterns, duration, and quality in a significant portion of the population [3]. Despite concerted initiatives to promote healthy dietary practices and regular physical activity, the significance of sleep health remains frequently overlooked by the general public. Thus, researching sleep’s impact on human health—particularly in children—has become increasingly urgent, as their developing bodies and brains are uniquely vulnerable to disruptions in sleep homeostasis. Blood pressure (BP) levels established during early life have been observed to persist into adulthood [4], serving as potential predictive markers for various health conditions such as metabolic syndrome [5] and carotid intima-media thickness in later years [6]. Epidemiological reports revealed that the prevalence of elevated blood pressure (EBP) among children hovered around 4.9% during The 2003 to 2016 period [7, 8]. A key modifiable risk factor for the development of EBP is an unhealthy lifestyle [9], which notably includes inadequate sleep quality and duration [10]. Although the link between poor sleep and obesity is well-established in the literature [11, 12], the specific impact of sleep on EBP in children and adolescents remains under-recognized and debate. Previous evidence indicated that sleeping for less than 5 h a day was linked to a higher likelihood of developing hypertension; conversely, adolescent girls who consistently sleep over 9.5 h may also face an elevated risk of hypertension [13]. Research in this domain has predominantly focused on discrete sleep factors, such as sleep duration [14, 15] and bedtime routines [16, 17]. The importance of circadian rhythms in health and disease has garnered increasing recognition, as they exert a fundamental influence on the regulation of physiological and behavioral functions [18]. While circadian rhythms and chronotype research have primarily focused on internal biological mechanisms, the “school-day sleep patterns” imposed by social pressures (e.g., staying up late, inadequate sleep duration) pose distinct health risks that demand equal attention. These patterns, driven by academic obligations, extracurricular pressures, and digital media exposure, disrupt sleep homeostasis independently of an individual’s innate chronotype. Early school start times (e.g., before 8:00 AM) compel students to awaken 2–3 h earlier than their biological preference, leading to cumulative sleep debt. A study of high school students found that each additional hour of sleep loss was associated with a 27% increase in insulin resistance [19]. A common misconception persists that staying up late, as long as it is compensated by an adequate sleep duration, constitutes a healthy practice. Nevertheless, the specific role of sleep duration in modulating the influence of school-day sleep phenotype on children’s health remains poorly understood. To date, studies examining the effects of phenotype in pediatric populations have been constrained with sleep duration often treated merely as an adjustment factor [20, 21], underscoring the need for a comprehensive understanding of how these combined sleep-related patterns influence BP levels in young populations. For instance, despite the continuous efforts by China’s education authorities to optimize and reform basic education recently, which explicitly stipulate that primary school students’ class start time must not be earlier than 8:20, no research has yet systematically proven the benefits of this policy on children’s BP. Furthermore, the current recommendation by China’s Ministry of Education regarding appropriate sleep duration for primary school’s students is presented as a specific time point (at least 10 h), rather than a sleep range as proposed by the American Academy of Sleep Medicine [13]. This necessitates a reevaluation, as previous studies have also indicated that both excessive and inadequate sleep duration are detrimental to population health. While chronotype research illuminates innate circadian influences, the school-day sleep patterns shaped by social pressures represent a parallel public health crisis. Addressing this gap is crucial for developing targeted interventions. Intriguingly, evidence has suggested that increased weekend compensatory sleep may serve as a compensatory mechanism, mitigating the detrimental effects of weekday sleep deprivation on BP abnormalities [22]. Even individuals with excessive sleep duration may experience fragmented sleep, leading to residual debt that compensatory sleep aims to address. To date, the extent to which compensatory sleep modifies risks associated with EBP among individuals with shorter or longer sleep duration—who may also experience sleep fragmentation or circadian misalignment—remains undefined. To fill this knowledge gap, our study leverages a large, nationwide pediatric sample to comprehensively evaluate the effects of school-day sleep patterns on pediatric BP health. Furthermore, we aim to elucidate the potential roles of weekend compensatory sleep in modulating these associations. Methods Study population We obtained data from The 2019 cycle of the Chinese National Survey on Students’ Constitution and Health (CNSSCH), which is the largest nationwide study on the health status of school-aged children and adolescents in China. This survey employed a multistage, stratified cluster sampling method to ensure that the sample was representative of the entire population of school-aged children (Fig. 1). The CNSSCH encompasses 30 provinces, excluding Tibet, Hong Kong, Macau, and Taiwan, utilizing consistent sampling and investigation methodologies across all survey cycles. Participants were categorized as either urban or rural based on their household registration, in accordance with their geographic locale [23]. Initially, there were 212,713 children aged 7 to 18 years in The 2019 CNSSCH survey cycle, and the number of children aged 7 to 12 years was 107,853. To mitigate potential biases and accommodate variations in BP characteristics among diverse demographic groups, propensity score matching (PSM) was employed. Propensity scores for each participant were derived using a logistic regression model incorporating baseline variables such as age, sex, province, urban/rural classification, obesity status, and sleep duration. We implemented 1:1 nearest neighbor matching with a caliper width of 0.1 standard deviations of the propensity score, ensuring non-replaceable matching. Ultimately, our study included a total of 15,612 (14.48%) children aged 7–12 years from 504 primary schools, maintaining a 1:1 ratio of those with EBP versus those without EBP. The CNSSCH was approved by the Medical Research Ethics Committee of Peking University Health Science Center (IRB00001052-18002), and verbal informed consent was obtained from both students and their parents. The study population selection of CNSSCH across China and the distributions of sleep timing Blood pressure measurements In this study, BP was measured on the right arm employing a mercury sphygmomanometer, adhering to the National High Blood Pressure Education Program Working Group in Children and Adolescents. The measurements were all standardized and taken in the morning. Field professionals who had completed a training course in anthropometric measurements carried out the measurements at the survey site. Before taking the initial reading, the child was seated in a comfortable position for a minimum of 10 min. The mean of three consecutive measurements was computed for each participant and utilized. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were determined by documenting Korotkoff sounds at phases 1 and 5, respectively. For participants aged 7–12 years, EBP was defined as SBP and/or DBP at or above the 95th percentile for their respective sex, age, and height [24]. Definition for sleep patterns and compensatory sleep In this study, we defined several key concepts related to sleep patterns and physical activity. Given that our study focused on primary school’s children, who spend the majority of their time in school, their retrospective estimates of sleep duration were marginally more precise for schooldays than for non-schooldays. This discrepancy may stem from the greater structure and fewer disruptions in sleep schedules during schooldays, which are governed by academic obligations such as classes and homework. Consequently, the structured nature of schooldays likely facilitates more reliable recollection of sleep patterns among this demographic. Sleep phenotype was assessed by inquiring about typical bedtimes and wake-up times using the question: “When do you usually go to bed/wake up on schooldays?”. The sleep midpoint, recorded in hours and minutes, was calculated as the midpoint on schooldays between bedtime and wake-up time, with hourly increments (e.g., 2:30 AM recorded as 2.5). Participants with later midpoints were more likely to have an evening phenotype. Based on this definition, the sleep timing is more likely to reflect adherence to social schedules and school-related pressures. Sleep duration was calculated based on the interval between bedtimes and wake-up times. Weekend compensatory sleep was quantified as the difference between average sleep duration on non-schooldays and schooldays, reflecting the compensatory sleep behavior on weekends [25]. For subsequent analysis, it was dichotomized based on its median values. Covariates Anthropometric assessments included measurements of height and weight. Height was measured with students standing upright and barefoot, while weight was recorded to the nearest 100 g using electronic or lever scales, with minimal clothing and no shoes worn. Body mass index (BMI) was calculated as weight (kg) divided by height (m) squared. Obesity was defined using age- and sex-specific BMI cutoffs based on the national standards [26]. Specifically, children were classified as obese if their BMI exceeded the 95th percentile for their age and sex. Lifestyle habits were evaluated based on the frequency of breakfast consumption, egg intake, milk consumption, and sugar-sweetened beverage (SSB) intake over the past week. Social jet lag was defined as the absolute difference between the mean sleep midpoint on weekends and weekdays. Higher values indicated a shift towards later sleep timing (i.e., later bed and wake times) on weekends compared to weekdays [20], which was also dichotomized based on their median values. It captured both earlier and later sleep shifts on weekends compared to schooldays. Within the sample, 80.27% of children exhibited positive social jetlag (going to sleep later on weekends than on schooldays), 9.54% showed negative social jetlag (going to sleep earlier on weekends than on schooldays), and 10.18% maintained the same sleep midpoint on weekends and schooldays. Physical activity was evaluated as school-based activity using the question: “In the past 6 months, on average, how much time per day did you engage in physical exercise at school (primarily including morning exercises, physical education classes, and extracurricular sports activities)?” Responses were categorized into three groups: < 1 h/day, 1–2 h/day, and ≥ 2 h/day. Given China’s extensive longitudinal span, resulting in a wide range of time zones, sunrise and sunset times in Xinjiang are notably delayed compared with those in eastern China, influencing local wake-up and bedtime patterns. Given that anthropometric measurements were conducted in September 2019, we obtained sunrise and sunset times for September 15, 2019, from the National Time Service Center, Chinese Academy of Sciences (Additional file 1: Fig. S1), to account for regional variations in children’s sleep–wake schedules due to latitudinal and longitudinal differences. This delay in sunrise and sunset in Xinjiang results in correspondingly later sleep and wake times among children. Statistical analysis To examine the differences in group characteristics, One-way Analysis of Variance (ANOVA) was used for continuous variables, while χ2 tests were applied for categorical variables. To elucidate the associations between sleep midpoint/duration and BP levels in the province-level analyses, we first calculate the average sleep midpoint/duration and average BP (both are continuous variables) for each province, and then establish scatter plots with a non-parametric smoothing test. Due to the significantly later sunset times in Xinjiang compared to eastern China, the sleep midpoint in Xinjiang starts later accordingly. Then, at the level of 504 primary schools nationwide, scatter plots with a non-parametric smoothing test were also utilized, depicting the average sleep midpoint/duration versus the mean BP at the school level. Meanwhile, we separately excluded data from primary schools in Xinjiang due to the significant time difference between Xinjiang and other regions of China, to further verify the robustness of the associations between sleep midpoint/duration and BP. P-values for non-linear associations were derived from a generalized additive model with a smooth term, using a likelihood ratio test for the non-linear component. Furthermore, at the individual level, a quadratic linear regression model was employed to analyze the relationships between sleep midpoint/duration and SBP/DBP, with statistical significance defined as the p-value for the quadratic term < 0.05 for non-linear associations. The quadratic regression model is based on the principle that it incorporates a squared term of the sleep midpoint/duration in the regression equation, typically expressed as (y = β0 + β1x + β2x2 + ϵ). This allows for the detection of non-linear relationships. To mitigate the potential influence of outliers, we conducted sensitivity analyses, only including schools or individuals where the average BP and average sleep timing fall within a certain normal range. For the binary variables of EBP status, non-linear associations between sleep midpoint/duration and EBP were evaluated using dose–response curves fitted with restricted cubic spline (RCS) methods. RCS can capture complex non-linear relationships without assuming a specific pre-defined functional form. It divides the range of the independent variable into several segments (Knots) and fits a separate cubic polynomial within each segment, while ensuring smoothness at the knots through specific constraints. To determine the optimal number of segments (Knots) for RCS curve piecewise fitting, Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were calculated across K = 3 to 7. The 3-knot model yielded the lowest AIC and BIC compared to other alternatives (Additional file 1: Table S1). To avoid overfitting, we selected K = 3 as the optimal segment number, corresponding to two key inflection nodes in the RCS curve. Based on Harrell’s defaults, three nodes (placed at the 15%, 50%, and 85% quantiles of the sleep variable) were aligned with the study context. To assess the fit of the RCS model, we also employed a likelihood ratio test (LRT) (Additional file 1: Table S1). To determine the “beneficial” range of sleep timing derived from the RCS curve, we first used the R package “mgcv” to generate point-by-point predicted values of the RCS fitted curve, thereby identifying the specific sleep timing associated with the minimum risk of EBP. We then estimated the non-significant confidence interval (CI) for this lowest point, specifically by quantifying the variability of the minimum point through 1,000 bootstrap resamplings. Finally, this interval was expanded into a clinically meaningful range based on public health relevance and clinical practicality to define the “beneficial” range (Additional file 1: Table S2). Specifically, the non-significant range for sleep midpoint was 1.81–3.37 (1:49–3:22 AM), and the school-day sleep phenotype was classified into three groups based on the tertiles of sleep midpoint: morning-type (“morning larks,” before 2:00 AM), intermediate-type (2:00–2:30 AM), and evening-type (“night owls,” after 2:30 AM), consistent with previous methodology [27]. The non-significant range for sleep duration was 8.37–14.83 h/night. In 2015, the National Sleep Foundation convened a multidisciplinary expert panel (Panel) with broad representation from leading stakeholder organizations; consensus for sleep duration recommendations was reached for specific age groupings: 9 to 11 h is recommended for school-aged children. Sleep duration was then classified into three categories: inadequate (< 9 h/night), moderate (9 ~ 11 h/night), and excessive (≥ 11 h/night) [28] (Additional file 1: Table S2). Sensitivity analyses were also conducted for the sleep duration of 8.5 ~ 11.5 h/night. P for non-linear was derived from a likelihood ratio test comparing the RCS model with 3 knots (df = 2) to a linear model, where a P < 0.05 indicates significant non-linear associations. Additionally, a percentage stacked bar chart was employed to depict the distribution of sleep phenotype (morning-, intermediate-, and evening-type) and sleep duration groups (inadequate, moderate, excessive) at the province level in China. By combining sleep phenotype with sleep duration, we formed nine groups of school-day sleep patterns (morning-type + inadequate duration; morning-type + moderate duration; morning-type + excessive duration; intermediate-type + inadequate duration; intermediate-type + moderate duration; intermediate-type + excessive duration; evening-type + inadequate duration; evening-type + moderate duration; evening-type + excessive duration), which were presented in a chordal graph. Furthermore, multi-variable linear regression and logistic regression models were utilized to examine the effects of each sleep pattern on SBP/DBP levels (β, 95% CI) and EBP (odds ratio, ORs), with the group of intermediate-type + moderate duration as a reference. All models were adjusted for various factors including age, sex, province, obesity status, and the frequency of egg, milk, breakfast, and SSB consumption, sunrise/sunset times, social jet lag, and physical activity. For subgroup analyses, interaction effects were evaluated by including a product term of the subgroup variable (e.g., sex, obesity status, physical activities, and social jet lag) and school-day sleep patterns in the multi-variable logistic regression models. The significance of the interaction term was assessed using a likelihood ratio test, with a P-value < 0.05 considered statistically significant for indicating heterogeneity of the association across subgroups. To further investigate the moderating effects of compensatory sleep, a series of stratified analyses using logistic regression models were performed, adjusting for the same covariates. Then, the population attributable fraction (PAF%) was calculated to quantify the proportion of EBP cases in the population associated with school-day sleep patterns that could be theoretically avoided if the detrimental effects of compensatory sleep were eliminated. Mathematically, it is computed using the formula: where P represents the prevalence of compensatory sleep in the study population, and RR denotes the risk of EBP associated with sleep patterns combined with less or more compensatory sleep, derived from the logistic regression models. We aimed to determine whether the “compensatory sleep” (less vs. more groups) can significantly modify the PAFs of school-day sleep patterns for EBP. Collinearity diagnostics for these sleep parameters are incorporated, and results are presented in the Additional file 1: Table S3, which indicate that collinearity among these parameters is not severe. These analyses were then conducted using STATA/SE 15 (StataCorp LLC, College Station, TX) and R software (version 4.0.3; created by John Chambers and colleagues; located in Jersey City, NJ, USA), with data preprocessing (including handling missing values and factor conversion) conducted to ensure model validity. Statistical significance was set at two-sided P-values < 0.05. Results Study population characteristics The study ultimately enrolled 15,612 children (Table 1). The gender distribution was nearly balanced, with boys accounting for 48.6% of the sample, and the mean age was 10.59 years (Standard deviation, SD = 1.18). Among the total population, 37.2% of children and adolescents were categorized as overweight or obese. Regarding dietary habits, 80.5% of children consumed breakfast daily, but only 22.1% ate an egg daily and 37.1% consumed dairy products daily. Furthermore, 68.3% of participants reported not consuming SSBs in the past 7 days. Detailed sleep characteristics were shown in Additional file 1: Table S4. The mean sleep midpoint on schooldays is 2.26 (approximately 2:16), while on non-schooldays it is 3.10 (approximately 3:06), showing a shift. Table 1The basic characteristics of study population by nine groups of school-day sleep patternsCharacteristicsTotal populationMorning-typeIntermediate-typeEvening-typeP-valueInadequate durationModerate durationExcessive durationInadequate durationModerate durationExcessive durationInadequate durationModerate durationExcessive durationN15,6122132 (13.7)4083 (26.2)154 (1.0)3040 (19.5)2469 (15.8)32 (0.2)2748 (17.6)614 (3.9)340 (2.2)Boys, n (%)7582 (48.6)996 (46.7)2001 (49.0)84 (54.6)1413 (46.5)1301 (52.7)18 (56.3)1278 (46.5)283 (46.1)208 (61.2) < 0.001Age (mean ± SD)10.59 ± 1.1811.00 ± 1.0910.33 ± 1.089.99 ± 1.0310.79 ± 1.1210.35 ± 1.1510.44 ± 1.0510.86 ± 1.1710.07 ± 1.5010.35 ± 1.32 < 0.001BMI, kg/m2 (mean ± SD)19.50 ± 4.0020.14 ± 4.2519.02 ± 3.8419.14 ± 4.0519.64 ± 4.0619.15 ± 3.7719.59 ± 3.6919.87 ± 4.0319.50 ± 4.1319.61 ± 4.07 < 0.001Overweight/obesity, n (%)5809 (37.2)857 (40.2)1392 (34.1)58 (37.7)1120 (36.8)887 (35.9)14 (43.8)1089 (39.6)255 (41.5)137 (40.3) < 0.001Urban residence, n (%)7336 (47.0)702 (32.9)1342 (32.9)48 (31.2)1402 (46.1)1402 (56.8)18 (56.3)1872 (68.1)402 (65.5)148 (43.5) < 0.001Sunrise time (mean ± SD)6.23 ± 0.495.94 ± 0.376.03 ± 0.436.16 ± 0.465.66 ± 0.366.09 ± 0.456.28 ± 0.466.28 ± 0.466.35 ± 0.476.41 ± 0.62 < 0.001Sunset time (mean ± SD)6.34 ± 0.496.04 ± 0.376.14 ± 0.436.27 ± 0.455.77 ± 0.366.19 ± 0.456.38 ± 0.466.38 ± 0.456.45 ± 0.476.51 ± 0.61 < 0.001Breakfast every day, n (%)12,561 (80.5)1691 (79.3)3445 (84.4)97 (62.9)2449 (80.6)2074 (84.0)22 (68.8)2178 (79.3)486 (79.2)119 (35.0) < 0.001Egg intake every day, n (%)3447 (22.1)454 (21.3)1062 (26.0)35 (22.7)628 (20.7)563 (22.8)8 (25.0)530 (19.3)141 (22.9)26 (7.7) < 0.001Diary intake every day, n (%)5798 (37.1)790 (37.1)1523 (37.3)44 (28.6)1115 (36.7)986 (39.9)15 (46.9)1033 (37.6)243 (39.6)49 (14.4) < 0.001No SSBs intake, n (%)10,662 (68.3)1509 (70.8)2717 (66.5)98 (63.6)2158 (70.9)1689 (68.4)17 (53.1)1982 (72.1)383 (62.4)109 (32.1) < 0.001Physical activity ≥ 2 h/day, n (%)1892 (12.1)272 (12.8)519 (12.7)19 (12.3)347 (11.4)303 (12.3)4 (12.5)328 (11.9)79 (12.9)23 (6.8) < 0.001Wake-up times on schooldays (AM, mean ± SD)6.64 ± 0.835.90 ± 0.326.44 ± 0.376.90 ± 0.526.35 ± 0.327.00 ± 0.197.96 ± 0.246.71 ± 0.457.93 ± 0.5510.62 ± 1.05 < 0.001Wake-up times on non-schooldays, (AM., mean ± SD)7.84 ± 7.727.40 ± 1.117.65 ± 1.047.88 ± 1.047.73 ± 1.078.01 ± 0.958.08 ± 0.707.98 ± 1.138.33 ± 1.0010.51 ± 1.26 < 0.001Bedtimes on schooldays (PM, mean ± SD)9.40 ± 0.879.26 ± 0.448.63 ± 0.647.63 ± 0.739.56 ± 0.679.34 ± 0.298.37 ± 0.3310.44 ± 0.599.68 ± 0.6710.58 ± 1.02 < 0.001Bedtimes on non-schooldays, (PM, mean ± SD)9.82 ± 0.959.72 ± 0.859.32 ± 0.878.66 ± 1.269.92 ± 0.909.86 ± 0.729.05 ± 0.8310.43 ± 0.849.96 ± 0.9310.81 ± 0.96 < 0.001Sleep midpoint on schooldays (mean ± SD)2.26 ± 0.701.81 ± 0.241.65 ± 0.321.25 ± 0.462.32 ± 0.132.31 ± 0.162.26 ± 0.162.95 ± 0.323.08 ± 0.484.88 ± 1.09 < 0.001Sleep midpoint on non-schooldays (mean ± SD)3.10 ± 0.912.85 ± 0.832.73 ± 0.812.47 ± 1.043.09 ± 0.833.19 ± 0.752.80 ± 0.473.50 ± 0.813.42 ± 0.864.98 ± 1.05 < 0.001Sleep phenotype on non-schooldays < 0.001 Morning type1604 (10.3)410 (19.2)784 (19.2)60 (39.0)201 (6.6)80 (3.2)1 (3.1)49 (1.8)16 (2.6)3 (0.9) Intermediate type3385 (21.7)524 (24.6)1159 (28.4)32 (20.8)816 (26.8)510 (20.7)10 (31.3)256 (9.3)72 (11.7)6 (1.8) Evening type10,623 (68.0)1198 (56.2)2140 (52.4)62 (40.3)2023 (66.6)1879 (76.1)21 (65.6)2443 (88.9)526 (85.7)331 (97.4) Social jet lag (median ± IQR)0.92 ± 0.791.07 ± 0.801.10 ± 0.751.24 ± 0.920.85 ± 0.740.93 ± 0.720.59 ± 0.450.73 ± 0.660.58 ± 0.590.20 ± 0.38 < 0.001 Social jet lag < median, n (%)6985 (44.7)800 (37.5)1201 (29.4)49 (31.8)1522 (50.1)1077 (40.8)22 (68.8)1655 (60.2)422 (68.7)307 (90.3) < 0.001 Sleep duration on schooldays, hours (mean ± SD)8.76 ± 1.138.17 ± 0.519.58 ± 0.5311.29 ± 0.478.07 ± 0.699.37 ± 0.3311.39 ± 0.367.52 ± 0.809.69 ± 0.5511.48 ± 0.58 < 0.001 Sleep duration on non-schooldays, hours (mean ± SD)9.48 ± 1.309.11 ± 1.259.83 ± 1.2210.83 ± 1.349.28 ± 1.269.63 ± 1.0710.57 ± 1.248.95 ± 1.349.81 ± 1.1211.05 ± 0.98 < 0.001Sleep duration on non-schooldays < 0.001 Inadequate4185 (26.8)812 (38.1)611 (15.0)4 (2.6)986 (32.4)447 (18.1)2 (6.3)1236 (45.0)78 (12.7)9 (2.7) Moderate9115 (58.4)1136 (53.3)2637 (64.6)63 (40.9)1749 (57.5)1739 (70.4)14 (43.8)1289 (46.9)428 (69.7)60 (17.7) Excessive2312 (14.8)184 (8.6)835 (20.5)87 (56.5)305 (10.0)283 (11.5)16 (50.0)223 (8.1)108 (17.6)271 (79.7) Compensatory sleep, hours, (median ± IQR)1.10 ± 1.151.00 ± 1.330.67 ± 1.000.83 ± 1.331.14 ± 1.500.67 ± 0.921.00 ± 0.751.33 ± 1.500.75 ± 1.080.00 ± 1.00 < 0.001 Compensatory sleep < median, n (%)7381 (47.3)906 (42.5)2318 (56.8)82 (53.3)1151 (37.9)1463 (59.3)15 (46.9)847 (30.8)348 (56.7)251 (73.8) < 0.001Mean BP levels, mmHg (mean ± SD) Mean SBP111.98 ± 13.09113.61 ± 12.93111.62 ± 12.72110.99 ± 13.04112.37 ± 13.29110.19 ± 13.05113.06 ± 11.38112.54 ± 13.61111.29 ± 11.84112.51 ± 13.34 < 0.001 Mean DBP71.71 ± 10.6072.14 ± 10.2271.78 ± 10.6770.50 ± 11.1971.63 ± 10.4171.33 ± 10.9473.00 ± 9.8771.58 ± 10.7772.69 ± 9.6271.48 ± 11.230.012BMI body mass index, SSB sugar-sweetened beverage, SBP systolic blood pressure, DBP diastolic blood pressureSleep midpoint of 2:30 AM is recorded as 2.50, therefore 2.26 corresponds to approximately 2:16 AM The basic characteristics of study population by nine groups of school-day sleep patterns BMI body mass index, SSB sugar-sweetened beverage, SBP systolic blood pressure, DBP diastolic blood pressure Sleep midpoint of 2:30 AM is recorded as 2.50, therefore 2.26 corresponds to approximately 2:16 AM Associations between school-day sleep patterns and EBP We firstly investigated the relationships of sleep midpoint and duration with BP at the province-level (Additional file 1: Fig. S2), then, at the level of 504 schools in China, particularly sleep duration exhibited a non-linear relationship with BP (P for non-linear < 0.05, Additional file 1: Fig. S3). Recognizing the substantial difference in sunrise and sunset times between Xinjiang and eastern China, we conducted a sensitivity analysis excluding schools from Xinjiang. The results remained consistent, indicating non-linear associations between sleep duration with BP, based on data from 493 schools nationwide (P for non-linear < 0.05, Additional file 1: Fig. S4). When excluding the potential outliers of sleep timing and BP levels at school-level, the results did not change essentially (Additional file 1: Fig. S5). At the population level, there was a non-linear association between sleep midpoint and SBP (P for non-linear = 0.009), as well as between sleep duration and both systolic and diastolic BP (P for non-linear < 0.001, Additional file 1: Fig. S6). When EBP status was examined as a categorical variable using RCS model, an early sleep midpoint and short sleep duration significantly increase the odds of EBP, but the risks posed by late sleep midpoint and late bedtimes should not be overlooked either (P for non-linear < 0.05, Fig. 2). Meanwhile, compared with the OR value at the minimum point, the OR values at the 10th and 90th percentiles of each sleep parameter were greater than 1, but a two-sided rise was particularly evident for sleep midpoint (Additional file 1: Table S5). Based on the sleep midpoint, “morning larks” (OR = 1.10, 95%CI = 1.02, 1.18) or “night owls” (OR = 1.11, 95%CI = 1.02, 1.21) were both associated with EBP (P < 0.05). What’s more, inadequate (OR = 1.11, 95%CI = 1.04, 1.19) sleep duration also contributed to EBP, while excessive duration was not a significant risk (OR = 1.04, 95%CI = 0.87, 1.25) (Additional file 1: Table S6). Sensitivity analyses for the sleep duration of 8.5 ~ 11.5 h/night showed similar trends (Additional file 1: Table S7). Relationships between schoolday’s wake-up times, bedtimes, sleep midpoint, duration, and EBP (P for non-linear was derived from a likelihood ratio test comparing the restricted cubic spline model with 3 knots to a linear model, where a P < 0.05 indicates significant non-linear association; the red interval refers to the range with no statistical significance) By combining sleep phenotype with sleep duration, we have formed nine groups of school-day sleep patterns. Notably, the majority of children were classified as “morning” (40.80%) and “intermediate” (35.48%) type on schooldays (Fig. 3, Additional file 1: Table S8). At the province-level in China, children from Liaoning and Jilin provinces exhibited a high tendency to be “morning larks” during schooldays, whereas those from Hebei and Guangxi provinces had a higher proportion of “night owls” (Additional file 1: Fig. S7). Associations of school-day sleep patterns with SBP/DBP levels and EBP (*P < 0.05; a the prevalence of sleep patterns combining sleep phenotype and sleep duration on schooldays and non-schooldays; b associations of school-day sleep patterns with SBP/DBP levels using multivariable linear regression (β, 95% CI) and EBP using multivariable logistic regression (ORs, 95% CI)) Even “morning larks” or “night owls” who obtained sufficient sleep were associated with increased SBP (morning-type + moderate duration: β = 1.77, 95% CI = 1.12, 2.41; evening-type + moderate duration: β = 1.20, 95% CI = 0.08, 2.33) and increased DBP levels (morning-type + moderate duration: β = 0.58, 95% CI = 0.04, 1.12; evening-type + moderate duration: β = 1.49, 95% CI = 0.55, 2.43). All morning-, intermediate- and evening-phenotype, along with inadequate sleep duration, were associated with higher odds of EBP, compared to the groups of intermediate-phenotype with moderate sleep duration (P < 0.05) (Fig. 3), with the prevalence of EBP shown in Additional file 1: Table S9. Sensitivity analyses for sleep duration of 8.5 ~ 11.5 h/night revealed trends that mirrored those of the primary analysis (Additional file 1: Fig. S8). subgroup analyses found that “morning larks” and “night owls” with unhealthy sleep duration tended to have increased odds of EBP significantly in girls and children with low physical activities, but sleep-deprived “night owls” had a higher risk of EBP among obese individuals (P for interaction < 0.05, Fig. 4). Subgroup analysis for associations of school-day sleep patterns and EBP (a Morning-type + inadequate duration; a Morning-type + moderate duration; c Morning-type + excessive duration; d Intermediate-type + inadequate duration; e Intermediate-type + moderate duration; f Intermediate-type + excessive duration; g Evening-type + inadequate duration; h Evening-type + moderate duration; i Evening-type + excessive duration. Adjusted for age, sex, province, obesity status, and the frequency of egg, milk, breakfast, and SSB consumption, sunrise/sunset times, social jet lag, and physical activity) The impacts of compensatory sleep on EBP associated with school-day sleep patterns The positive association between compensatory sleep and EBP remained essentially unchanged before (OR = 1.04, 95% CI = 0.98, 1.11) and after (OR = 1.03, 95% CI = 0.96, 1.10) adjustment for weekday sleep duration (Additional file 1: Table S10). Considering weekend compensatory sleep as a stratifying factor, when more compensatory sleep was attained, the prevalence of EBP decreased, especially in children with inadequate school-day sleep duration (Fig. 5a). The increased compensatory sleep may influence the odds of EBP in sleep-deprived “morning larks” (ORs decreased from 1.94 to 1.12) and “night owls” (ORs decreased from 1.66 to 1.15) (Fig. 5b). Within the group that experienced more compensatory sleep, the population attribution fraction of EBP declined, especially in “morning larks” (Fig. 5c). In the subgroup of more compensatory sleep, the associations between school-day sleep patterns and EBP significantly decreased in girls, while there were no significant changes in children with obesity; these subgroup findings should be interpreted as exploratory (Additional file 1: Table S11). The potential moderating impacts and the modified population attribution fraction (PAF%) of compensatory sleep in children (a subgroup prevalence of EBP (difference%) in each group of school-day sleep patterns among children according to compensatory sleep; b changes of ORs; c changes of population attribution fraction) Discussion Principal findings Utilizing a meticulously matched national sample of children spanning across China, our study presents novel and groundbreaking insights. New evidence has emerged regarding the impact of sleep on EBP, suggesting the existence of potential optimal mid-sleep points and sleep duration. An early sleep midpoint and short sleep duration significantly increase the risk of EBP, but the risks posed by the late sleep midpoint and late bedtimes should not be overlooked either. Even the “morning larks” phenotype with sufficient sleep duration is associated with an increased risk of EBP. Furthermore, our findings reveal that engaging in compensatory sleep has a potential modifying effect on EBP, particularly among sleep-deprived individuals identified as “morning larks” and “night owls.” For practical guidance, prioritize ensuring ≥ 9 h of sleep on schooldays, recommend adopting an Intermediate-type sleep phenotype, and avoid excessively early wake-up times. In light of recent educational reforms in China, which mandate that primary school classes commence no earlier than 8:20 AM, our study suggests that appropriately delaying school start times may be a beneficial strategy for managing pediatric BP. Our findings underscore the necessity of considering sleep phenotype, sleep duration, and compensatory sleep when formulating recommendations for sleep to enhance pediatric cardiovascular health with both policy and family engagement perspectives. However, these findings are reported in a tentative manner to refrain from suggesting causal links or immediate mitigating effects, underscoring the necessity for additional research to verify these associations. School-day sleep patterns and blood pressure In China, the majority of children adhere to a “morning lark” or “intermediate type” pattern on schooldays, which may be presumably influenced by the early and rigid schedules imposed by the Chinese education system. Previously, the elementary schools in China required students to arrive at school before 8:00 AM to start classes, which necessitates children developing the habit of waking up early. According to our results, getting up between 6:40 and 8:00 in the morning may be beneficial to children’s BP, while getting up earlier than 6:40 significantly increased the risk of EBP. “Morning larks” were more likely to develop EBP; policy on appropriately delaying the start time of school may be more conducive to the BP health of children. This finding aligns with earlier evidence indicating that morning types may be more strongly associated with EBP relative to those with an intermediate type [29]; the early rising habits can be a double-edged sword, as waking up before 6:00 AM may result in sleep deprivation [30]. Recently, China’s education authorities have optimized and reformed basic education by clearly stipulating that primary school students must not start classes earlier than 8:20, which can ensure to a certain extent that children do not have to wake up too early. Besides, the interplay between school start times and sunrise is vital for comprehending how school start times impact adolescents’ sleep patterns [31]; therefore, we have adjusted the sunrise/sunset time as crucial factors in the models. In addition to the morning phenotype, preliminary studies have suggested a significant correlation between evening-type and increased cardio-metabolic risk among young adults [21] and adults [32, 33]. What’s more, children and adolescents with significantly earlier or later bedtime routines were reported to have higher risks of EBP, in contrast to those with bedtimes between 9:00 and 10:00 PM [34]. Similar to our results, an early sleep midpoint significantly increases the odds of EBP, but the risks posed by a late sleep midpoint and late bedtimes should not be overlooked either. This reinforces the importance of establishing healthy school-day sleep patterns to promote optimal BP health. It is worth noting that we defined school-day sleep patterns, which more reflect the impact of students’ socially stress-adaptive sleep patterns on their BP health, rather than the influence of traditional internal circadian rhythms. Therefore, this study emphasizes the need to call on multiple organizations such as governments, schools, and social institutions to actively participate in regulating students’ social stress, ensuring that students have healthy sleep patterns to reduce their risk of EBP. Moderate sleep duration range and blood pressure The literature on the associations between sleep duration and children’s BP was extensive yet yields inconsistent findings [35, 36]; both insufficient and excessive sleep duration had been implicated in hypertension [15], which may be underpinned by physiological mechanisms such as dysregulation of the hypothalamic–pituitary–adrenal axis, autonomic nervous system imbalance, and perturbations in inflammatory and metabolic pathways—processes that are disrupted by either inadequate or excessive sleep, thereby contributing to EBP. To be noted, the National Sleep Foundation consensus for sleep duration recommendations was reached for specific age groupings: 9 to 11 h is recommended for school-aged children [28]. Whereas, according to the sleep guidelines proposed by China’s Ministry of Education, elementary school students should aim to get 10 h of sleep per night. Based on the current data, this study posits that among primary school students in China, short sleep duration is significantly associated with an increased risk of EBP; by contrast, the risk implications of long sleep duration demand more rigorous evaluation. Specifically, our data reveal that the group with a sleep duration of ≥ 11 h does not exhibit a statistically significant overall risk of EBP. Furthermore, within the 9 ~ 11 h sleep duration, variations in sleep phenotype exert an additional influence on BP: relative to children with an intermediate-type and moderate sleep duration, those with morning- or evening-type may demonstrate higher levels of BP and an elevated risk of EBP. These results suggest that using the single value of 10 h as the sole criterion to judge whether a child’s sleep is adequate is insufficiently comprehensive; instead, simultaneous attention should be paid to the synergistic effects of sleep duration range and sleep phenotype. Prioritize ensuring a sleep duration of ≥ 9 h on schooldays, recommend the adoption of an intermediate phenotype, and avoid excessively early wake-up times. No consistent harmful association is observed regarding sleep duration of 11 h or more. By emphasizing the importance of “appropriate sleep duration range” and “intermediate-phenotype,” this study provides data support for a more comprehensive understanding of the multidimensional nature of sleep health and the optimization of sleep health guidance programs for children. The impacts of weekend compensatory sleep and subgroup population It was noted that children with insufficient sleep duration were more prone to engage in compensatory sleep on weekends. The traditional view suggests that catching up on sleep over the weekend may enhance subjective well-being. Several studies have also confirmed that compensatory sleep is primarily beneficial for those with the greatest sleep deficits [37, 38], which aligns with our findings that compensatory sleep has a potential modifying effect on its efficacy in reducing the odds of EBP, especially in children with inadequate sleep duration on schooldays. Interestingly, among children with an early-phenotype and excessive sleep duration, those who engaged in more compensatory sleep exhibited a slight increase in EBP prevalence compared to their counterparts with less compensatory sleep. These results underscore the importance of considering an individual’s baseline sleep characteristics when recommending compensatory sleep as an intervention for EBP. Specifically, the findings indicate associations where compensatory sleep might be detrimentally associated among those with already enough sleep duration on schooldays. However, previous evidence suggests that weekday sleep loss is irrecoverable, and model-based simulations of sleep duration inform individuals of the amount of irrecoverable sleep loss caused by advancing wake-up times on weekdays [39–41]. Therefore, the findings of this study, based solely on inferences from a cross-sectional epidemiological survey, indicated that the association between school-day sleep patterns and children’s BP health may differ between the two subgroups with shorter and longer weekend compensatory sleep. In addition, one study found that the contribution of compensatory sleep became non-significant after accounting for the negative health effect of short weekday sleep duration [42]. However, in the current study, the associations between compensatory sleep and EBP did not change essentially after adjusting for the negative health effect of inadequate weekday sleep duration (Additional file 1: Table S10). Though more compensatory sleep was positively related to EBP, the findings of this study indicated that weekend compensatory sleep may modify the associations between school-day sleep patterns and children’s BP health. Nevertheless, these observations are presented tentatively to avoid implying causal relationships or direct mitigating effects, emphasizing the need for further research to validate these associations. Public health significance, strength and limitations Given the alarming global surge in childhood hypertension [43], our findings provide novel insights into the complex interplay between sleep patterns and pediatric BP health, highlighting several key implications. An early sleep midpoint and short sleep duration significantly increase the risk of EBP, but the risks posed by late sleep midpoint and late bedtimes should not be overlooked either. Within the 9 ~ 11 h sleep duration, variations in sleep phenotype exert an additional influence on BP; even “morning larks” who obtain sufficient sleep are associated with an increased risk of EBP. Additionally, compensatory sleep is shown to influence the associations between sleep-deprived “morning larks” or “night owls” with EBP. Therefore, educational policies should give priority to maintaining a schoolday sleep duration of ≥ 9 h, recommend embracing an intermediate phenotype, and avoid excessively early school start times, thereby promoting healthier BP outcomes. Through highlighting the significance of both an “appropriate sleep duration range” and the “intermediate-phenotype,” this study offers data-driven evidence to foster a more holistic comprehension of sleep health’s multidimensional characteristics and to refine sleep health guidance initiatives for children. Before delving into the significance and implications of our findings, it is essential to acknowledge the inherent limitations of this study. First, a primary constraint, stemming from its exploratory nature, is the cross-sectional design, which prevents the inference of causal relationships between sleep patterns and EBP. Secondly, given that our study focused on primary school’s children, their retrospective estimates of sleep timing were marginally more precise for schooldays; we defined sleep phenotype using the sleep midpoint on schooldays in this large pediatric population, which might represent the school-related pressures and socio-behavioral adaptations but not the intrinsic circadian preference. Thirdly, we defined sleep phenotype using self-reported sleep midpoint; future research should adopt the validated reduced Morningness-Eveningness Scale [44] or rely on objective sleep data (e.g., from actigraphy) and assess children’s sleep quality, as these factors could influence the observed associations. Including these aspects would offer a more comprehensive understanding of the topic. Fourthly, since alarm clock use can significantly impact the assessment of socio-behavioral adaptations, future large-scale epidemiological studies on sleep should focus on alarm clock use behaviors. Fifthly, as non-school physical activities such as after-school sports may independently influence sleep and health outcomes, future studies should incorporate comprehensive assessments of both school and non-school based physical activities. Sixthly, the sparsity of data on excessive sleep duration limits our ability to draw definitive conclusions regarding its independent contribution to the observed non-linear associations. Last but not least, this study did not include children from Tibet, Hong Kong, Macau, or Taiwan. Although their exclusion is unlikely to significantly alter the primary conclusions, incorporating data from these regions could enrich the geographical analysis and provide a more comprehensive picture of the relationships under investigation. Conclusions In contemporary China, where children predominantly exhibit a “morning lark” phenotype on schooldays, our study underscores the critical role of school-day sleep patterns in pediatric BP management, suggesting the existence of optimal mid-sleep points and sleep duration on schooldays. An early sleep midpoint and short sleep duration significantly increase the risk of EBP, but the risks posed by late sleep midpoint and late bedtimes should not be overlooked either. For sleep durations between 9 and 11 h, differences in sleep phenotype bring about an additional influence on BP; even those “morning larks” with sufficient sleep are associated with a greater risk of EBP. No consistent harmful associations have been observed when it comes to sleep duration of 11 h or more. Viewed from this angle, policies should give priority to securing ≥ 9 h of sleep for students on schooldays, advocate for the adoption of an intermediate sleep phenotype, and deter excessively early school start times. Compensatory sleep on weekends may effectively affect the relationships between sleep-deprived “morning larks” and “night owls” with EBP risk. The observed influence of school-day sleep patterns and weekend compensatory sleep highlights potential modifiable risk factors, warranting further research to inform evidence-based interventions. Supplementary Information Additional file 1: Table S1. The Akaike Information Criterionand Bayesian Information Criterion, and likelihood ratio testfor RCS models. Table S2. the “beneficial” range of sleep timing derived from the RCS curve. Table S3. Collinearity diagnostics for included confounders and the sleep parameters. Table S4. Detailed sleep characteristics in the total study samples. Table S5. The effect estimates at specified 10th and 90th percentilesof each sleep parameter for EBP in children. Table S6. Associations between sleep parameters and EBP among children. Table S7. Sensitivity analysis for the sleep duration and EBP among children. Table S8. Prevalence of the sleep patterns combining sleep phenotype with sleep duration in children. Table S9. Prevalence of EBP in groups of school-day sleep patterns among children. Table S10. Associations between sleep duration on schooldays and non-schooldays, and the compensatory sleep with EBP. Table S11. Moderating effects of compensatory sleep in subgroup analysis for the associations between school-day sleep patterns and EBP. Fig. S1. The average sunrise and sunset time, wake-up times and bedtimes at all school stages in each province. Fig. S2. Province-level associations between sleep midpoint, duration with SBP/DBP nationwide. Fig. S3. School-level associations between sleep midpoint, duration with SBP/DBP at the level of 504 schools nationwide. Fig. S4. School-level associations between sleep midpoint, duration with SBP/DBP at the level of 493 schools excluding schools from Xinjiang. Fig. S5. School-level associations between sleep midpoint, duration with SBP/DBP at the level of schools nationwide excluding the potential influence of outliers, such as sleep midpoint > 3.00, sleep duration < 7 h/night, SBP > 140 mmHg or SBP < 90 mmHg, or DBP > 80 mmHg or DBP < 45 mmHg. Fig. S6. Quadratic linear associations between sleep midpoint, duration and SBP/DBP at the population level. Fig. S7. The distribution of sleep phenotype, sleep duration across China’s 30 provinces. Fig. S8. Sensitivity ssociations of school-day sleep patterns with SBP/DBP levels using multivariable linear regressionand EBP using multivariable logistic regression Acknowledgements The authors would like to acknowledge the support from all the team members and the participating students, teachers, parents and local education and health staff in the programs. Abbreviations Blood pressure Elevated blood pressure Propensity score matching Systolic blood pressure Diastolic blood pressure Body mass index Sugar-sweetened beverage Analysis of variance Restricted cubic spline Akaike information criterion Likelihood ratio test Confidence interval Odds ratio Population attributable fraction Standard deviation Authors’ contributions JL: Conceptualization, Methodology, Formal analysis, Writing-Original Draft, Writing-Review & Editing; XS, MC, YZ, JJ: Software, Validation, Data curation; RW, ZD, LC, YQ: Investigation, Visualization; TG, ZS, WY, XW, JM: Supervision; YD, YS: Resources, Writing-Review & Editing, Project administration, Funding acquisition. All authors read and approved the final manuscript. Funding The present study is supported by the National Key R&D Program of China (2024YFC3308302 to YD, 2024YFC2707901 to YS), National Natural Science Foundation of China (82103865 and 82373593 to YD; 82273654 to YS), Natural Science Foundation of Beijing Municipality (7222244 to YD; 7222247 to YS), and the Innovation Fund for Outstanding PhD Candidates of Peking University Health Science Centre (BMU2024BSS001 to JL). Data availability The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The CNSSCH was approved by the Medical Research Ethics Committee of Peking University Health Science Center (IRB00001052-18002), and verbal informed consent was obtained from both students and their parents. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Yanhui Dong, Email: dongyanhui@bjmu.edu.cn. Yi Song, Email: songyi@bjmu.edu.cn. References Associated Data Supplementary Materials Additional file 1: Table S1. The Akaike Information Criterionand Bayesian Information Criterion, and likelihood ratio testfor RCS models. Table S2. the “beneficial” range of sleep timing derived from the RCS curve. Table S3. Collinearity diagnostics for included confounders and the sleep parameters. Table S4. Detailed sleep characteristics in the total study samples. Table S5. The effect estimates at specified 10th and 90th percentilesof each sleep parameter for EBP in children. Table S6. Associations between sleep parameters and EBP among children. Table S7. Sensitivity analysis for the sleep duration and EBP among children. Table S8. Prevalence of the sleep patterns combining sleep phenotype with sleep duration in children. Table S9. Prevalence of EBP in groups of school-day sleep patterns among children. Table S10. Associations between sleep duration on schooldays and non-schooldays, and the compensatory sleep with EBP. Table S11. Moderating effects of compensatory sleep in subgroup analysis for the associations between school-day sleep patterns and EBP. Fig. S1. The average sunrise and sunset time, wake-up times and bedtimes at all school stages in each province. Fig. S2. Province-level associations between sleep midpoint, duration with SBP/DBP nationwide. Fig. S3. School-level associations between sleep midpoint, duration with SBP/DBP at the level of 504 schools nationwide. Fig. S4. School-level associations between sleep midpoint, duration with SBP/DBP at the level of 493 schools excluding schools from Xinjiang. Fig. S5. School-level associations between sleep midpoint, duration with SBP/DBP at the level of schools nationwide excluding the potential influence of outliers, such as sleep midpoint > 3.00, sleep duration < 7 h/night, SBP > 140 mmHg or SBP < 90 mmHg, or DBP > 80 mmHg or DBP < 45 mmHg. Fig. S6. Quadratic linear associations between sleep midpoint, duration and SBP/DBP at the population level. Fig. S7. The distribution of sleep phenotype, sleep duration across China’s 30 provinces. Fig. S8. Sensitivity ssociations of school-day sleep patterns with SBP/DBP levels using multivariable linear regressionand EBP using multivariable logistic regression Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.