We retrieved two cycles of data from 2015 to 2018 in the National Health and Nutrition Examination Survey (NHANES) database for research. A total of 19,225 potential participants were included in the study, and 11,583 participant were excluded for the following reasons: missing bedtime (n = 6,835), missing SBP data (n = 1,053), taking antihypertensive medications (n = 2,941), bedtime between 8:00 and 18:00 o'clock (n = 157), and age less than 18 years (n = 597). Finally, 7,642 participants were included in the study, after all exclusion criteria were met (Figure 1).

Bedtime was defined as the response to the question: “What time do you usually fall asleep on weekdays or workdays?”, the responses were categorized into 13 groups:19:00, 20:00, 21:00, 22:00, 23:00, 0:00, 1:00, 2:00, 3:00, 4:00, 5:00, 6:00, and 7:00. Moreover, it was changed into a continuous variable for data analysis.

Well-trained and certified inspectors used standardized protocols and calibrated equipment to obtain BP readings. Three consecutive BP readings were obtained through auscultation. If the BP measurement was not completed successfully, the fourth measurement would be performed. The SBP was the average of all available measurement data.

In our research, sociodemographic, lifestyle, metabolic, and sleep factors were collected. Sociodemographic factors included race, age, and gender. The race was classified into 4 groups: Mexican American, White, Black, and other race. Age was classified into 3 groups: young (18–44), middle-aged (45–59), and old (≥60). Lifestyle factors included alcohol consumption, smoke, and total physical activity (TPA). Alcohol consumption was defined as the response to the question: “In the past 12 months, how often did you drink any type of alcoholic beverage?”, and the responses were divided into 3 groups: drinking, no drinking, and not recorded. Smoking was defined as the response to the question: “Do you now smoke cigarettes?”, and the responses were divided into 3 groups: smoking, no smoking, and not recorded. According to the physical activity guideline and previous literature (15, 16), the TPA was divided into 5 groups: <150, 150–300, 300–450, ≥450 min, and not recorded. Metabolic factors included glomerular filtration rate (GFR), urea acid (UA), total cholesterol (TC), low-density lipoprotein (LDL), high-density lipoprotein (HDL), body mass index (BMI), and diabetes. GFR was estimated by the simplified Modification of Diet in Renal Disease (MDRD) equation: 186 × SC^−1.154 × Age^−0.203× (0.742 if female) (17). According to the WHO guidelines, BMI was divided into underweight (<18.5 kg/m²), normal weight (18.5–24.99 kg/m²), overweight (25–29.99 kg/m²), and obesity (≥30 kg/m²) (18). Diabetes was defined as the responses to the question: “Have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?”, and the responses were divided into 4 groups: yes, no, borderline, and not recorded. The method of obtaining other covariates, including UA, TC, LDL, HDL, can be found at www.cdc.gov/nchs/nhanes/.

Sleep duration was defined as the responses to the question: “How much sleep do you usually get at night on weekdays or workdays?”, and based on the guideline (19), the responses were categorized into 3 groups: <7 h, ≥7 h, and not recorded. Snore was based on the responses to the question: “In the past 12 months, how often did you snore while you were sleeping?” Trouble sleeping was based on the responses: “Have you ever told a doctor or other health professional that you have trouble sleeping?” Snort or stop breathing was based on the responses to the question: “in the past 12 months, how often did you snort, gasp, or stop breathing while you were asleep?” Overly sleepy was defined as the responses to the question: “In the past month, how often did you feel excessively or overly sleepy during the day?” The above 4 variables, namely, snore, trouble sleeping, snort or stop breathing, and overly sleepy, were divided into 3 groups according to the answers: yes, no, and not recorded.

A multivariate linear regression model was used to assess the correlation between bedtime at night and SBP in adults. The covariates mentioned above were adjusted as potential confounders. The continuous variables were represented by mean ± SD (normal distribution) or median (quartile) (skew distribution), which included GFR, UA, TC, LDL, HDL, and SBP. Categorical variables were represented by percentage or frequency, which included age, race, alcohol consumption, smoke, TPA, BMI, diabetes, sleep duration, snore, trouble sleeping, snort or stop breathing, overly sleepy, and bedtime. To calculate the differences between male and female, a linear regression model was used for continuous variables or chi-square tests for categorical variables. The values of missing continuous covariates were indicated by dummy variables, including GFR, UA, TC, LDL, HDL, and the missing ratios were 6.4, 6.4, 6.2, 57.4, and 6.2%, respectively. The missing categorical variables were included in the analysis as a single group, including alcohol consumption, smoking, TPA, BMI, diabetes, sleep duration, snore, trouble sleeping, snort or stop breathing, and being overly sleepy. We changed the bedtime from categorical variable to continuous variable for data analysis. Stratified analysis and smooth curve fittings were conducted to explore the relationship between bedtime and SBP based on gender, age, race, and BMI. The statistical software packages R (http://www.R-project.org) and EmpowerStats (http://www.empowerstats.com) were used for the data analyses. When the P-value was < 0.05, it was considered statistically significant.

Table 1 shows the description of sociodemographic and baseline characteristics. Our study included 7,642 participants subclassified based on gender. Among the participants, the proportion of male and female was 48.64% (n = 3,717) and 51.36% (n = 3,925), respectively. In ethnicity, the proportion of Mexican American, White, and Black was 17.43, 32.92, and 19.34%, respectively. Overall, the mean (SD) values for GFR, UA, TC, LDL, HDL, and SBP were 100.35 (27.49) ml/min/1.73 m², 313.49 (82.47) μmol/L, 4.90 (1.06) mmol/L, 2.91 (0.90) mmol/L, 1.40 (0.42) mmol/L, and 121.59 (17.21) mmHg, respectively. In the participants, 65.18% were alcohol drinkers, 18.61% were smokers, 55.51% were young, 30.31% were normal weight, and 9.83% were diabetes (yes/borderline). Table 2 shows the description of sleep factors. 22.09% of participants were sleep duration with <7 h, 65.60% were snored, 22.18% have trouble sleeping, 21.08% have snorted or stopped breathing, and 77.81% were overly sleepy. Table 3 shows the results of the univariate analysis of SBP.

The results from multiple linear regression analyses are illuminated in Table 4, which are used to explore the association between bedtime and SBP. In model I, after adjustment for sleep factors: sleep duration, snore, trouble sleeping, snort or stop breathing, and overly sleepy, a significantly negative association was identified (β, −0.74 [95% CI, −0.97, −0.51]). In model II, we adjusted for gender, race, and age in addition to the model I, a significantly negative association still presented (β, −0.26 [95% CI, −0.47, −0.05]). In model III, we adjusted for alcohol consumption, smoke, TPA, GFR, UA, TC, LDL, HDL, BMI, and diabetes in addition to model II, a significantly negative association still existed (β, −0.23 [95% CI, −0.43, −0.02]). We also conducted a smooth curve fitting to explore the non-linear relationship between bedtime and SBP (Figure 2). With the delay of bedtime, the SBP had a gradual decrease trend, which was decreased to the minimum at 0:00, and then increased gradually.

In the subgroup analysis stratified by gender, race, age, and BMI, the association between bedtime and SBP is explored in Figures 2–6. All the potential confounding factors except the subgroup variable were adjusted. In the female group (Figure 3), with the delay of bedtime, the range of SBP was decreased more obviously at 0:00, then was increased gradually, and fell again after 4:00. For the second downward trend, we thought that the sample size at each time point was too small to reflect the true relationship. In different age groups (Figure 4), with the age increasing, BP was gradually increased. In the 18–44 year group, bedtime had little effect on SBP. Among the≥45 years old group, this trend also existed. In the Black group (Figure 5), an obvious downward trend was found at 22:00, and the SBP was the lowest. However, the trend was not found in Mexican Americans. In different BMI groups (Figure 6), the downward trend of SBP was more obvious in groups with 18.5–24.9 BMI. In the group with BMI ≤ 18.5, the sample size was too small to reflect the true relationship.