Chronobiological perspectives: Association between meal timing and sleep quality

The National Health and Nutrition Examination Survey (NHANES), chestrated by the National Center for Health Statistics (NCHS) as part of the Centers for Disease Control and Prevention (CDC), serves as a comprehensive survey encapsulating the diverse demographic of the United States. This survey specifically focuses on noninstitutionalized U.S. residents, endeavoring to amass a vast array of health and nutritional data. Participation in NHANES is contingent upon obtaining informed written consent from the participants, in strict compliance with the ethical standards sanctioned by the National Center for Health Statistics Research Ethics Review Board (NCHS ERB), pursuant to Protocol #2005–06 and its subsequent extension covering the 2005–2006 and 2007–2008 cycles.

In the 2005–2006 and 2007–2008 NHANES cycles, an initial cohort of 20,497 participants was established. From this group, exclusions were made for 8,706 minors (below 18 years), 422 pregnant individuals, and cases with incomplete data in critical demographics and health metrics such as sex, age, race, the Pittsburgh Sleep Quality Index (PSQI), and sleep duration. Additionally, profiles with fewer than two meal occasions and missing data on covariates were also excluded. Consequently, the analysis was conducted on a refined dataset comprising 7,023 participants. This analysis included the computation of various health indices, notably the PSQI, sleep duration, timing of meals (including first and last), the midpoint of meal, meal duration, and frequency of meal occasions.

PSQI has been meticulously developed to evaluate sleep quality and patterns among adolescents and adults. During the NHANES 2005–2008 cycles, participants were required to complete eight self-reported items for the assessment of PSQI. Sleep latency was quantified via two specific items: (a) “How long does it usually take you to fall asleep at bedtime (minutes)?”, and (b) “In the past month, how often did you have trouble falling asleep?”. Similarly, sleep disturbances were gauged using two distinct items: (a) “In the past month, how often did you wake up during the night and have trouble getting back to sleep?” and (b) “In the past month, how often did you wake up early in the morning and was unable to get back to sleep?”. Concurrently, daytime dysfunction was assessed through another set of two items: (a) “In the past month, how often did you feel unrested during the day, no matter how many hours of sleep you had?” and (b) “In the past month, how often did you feel excessively or overly sleepy during the day?”. Typically, the time taken to fall asleep is categorized into four brackets: ≤15 minutes (score of 0), 16–30 minutes (score of 1), 31–60 minutes (score of 2), and ≥60 minutes (score of 3); And possible answers to the other questions were as follows: never-0 time a month (score of 0), rarely-1 time a month (score of 1), sometimes-2 to 4 times a month (score of 2), often-5 to 15 times a month (score of 3), and almost always-16 to 30 times a month (score of 4). The cumulative PSQI score, ranging from 0 to 23, is derived by summing the individual scores for each component: latency, disturbances, and daytime dysfunction. A PSQI score below 5 signifies good sleep quality, whereas scores ranging from 5 to 10 denote moderate sleep quality. Scores exceeding 10 are indicative of poor sleep quality, with escalating scores reflecting progressively deteriorating sleep quality [16].

Meal timing data were meticulously gathered via individual two 24-hour dietary recalls conducted on non-consecutive days during both the 2005–2006 and 2007–2008 NHANES cycles. The initial dietary recall interviews were executed in-person by professionally trained interviewers specializing in dietary assessments. Subsequently, the second dietary recall was conducted telephonically, typically scheduled between 3 to 10 days following the first. In both NHANES cycles, the U.S. Department of Agriculture (USDA) Automated Multiple-Pass Method was employed which uses a 5-step procedure to quantify 24-h food and beverage intake [17]. Meal timing variables incorporated in this study included meal duration, timing of the first and last meals, midpoint of meal, and frequency of meal occasions. Meal duration was determined by calculating the interval in hours between the first and last meal of each participant, and the first meal episode was regarded as having occurred from 05:00. The midpoint of meal was ascertained by identifying the central time point between the first and last meal, i.e., ([Timing of the last meal—Timing of the first meal]/ 2) + Timing of the first meal. Meal occasions were quantified as the count of daily episodes involving the intake of food or beverages totaling 1 kcal or more at a single time point [18]. All analyzed variables represent the mean values derived from the two 24-hour dietary recalls.

Data extracted from the NHANES database encompassed an array of sociodemographic variables: age, gender, ethnicity, education level, poverty ratio, and marital status (classified as Married or living with partner; Widowed, Divorced or never married). Body Mass Index (BMI) data (kg/m²) were categorized as follows: Underweight (<18.5), Normal (18.5 to <25), Overweight (25 to <30), Obese (≥30). Smoking status was delineated into three groups: Never Smokers, Former Smokers, and Current Smokers; Drinking status was bifurcated into two categories: “Yes” for drinkers and “No” for non-drinkers; Sleep duration was recorded as a continuous variable, representing the total hours of sleep. The healthy eating index 2015 (HEI-2015) scores and total energy intake were calculated form two 24-hour dietary recalls and HEI-2015 was categorized as inadequate (less than 50), average (50 and 70) and optimal (more than 70) [19].

Data from various sources were amalgamated into a unified dataset in strict adherence to the NHANES protocol. Given our analysis’s emphasis on individual meal timing and considering the substantial individual variability and the multitude of influencing factors in sleep quality, our statistical approaches intentionally excluded weighted factors. Continuous variables were reported as mean ± standard deviation (SD), whereas categorical variables were delineated in terms of frequencies and percentages. Initially, participants were stratified into three distinct groups based on their PSQI scores: Good (<5), Moderate (5–10), and Poor (>10) [20]. Baseline characteristics were systematically summarized in accordance with the classification of PSQI scores. Group differences were evaluated using the F test for continuous variables and the chi-square test for categorical variables. Normality was assessed using the residual plot (Q-Q plot), homogeneity of variance was tested via the Breusch-Pagan test, and collinearity was examined by calculating the variance inflation factor (VIF). Identifying data that perfectly satisfy all the assumptions of linear regression is inherently challenging. Several studies [21–23] have encountered and addressed similar challenges using linear regression as their primary analytical method. Subsequently, a multiple linear regression model was employed to analyze the relationship between meal timing and PSQI scores. Multiple linear regression models were applied to adjust for various covariates, with results being presented across three analytical spectrums: unadjusted, minimally adjusted, and fully adjusted. In the unadjusted model, no covariates were considered. The minimally adjusted model (Model 1) incorporated age, gender, and race as its primary covariates. The fully adjusted model (Model 2) further included additional covariates: Education, Poverty ratio, Marital status, BMI, smoking status, Drinking habits, HEI-2015, and total energy intake. The selection of these covariates was grounded in findings from previous research investigations.

All statistical analyzes were performed using the R statistical software package (http://www.R-project.org↗, The R Foundation) P values less than 0.05 (two-sided) were considered statistically significant.

Table 1 provides a comprehensive overview of the demographic characteristics of the participants. The study included a total of 7,023 participants, with a mean age of 50.54 ± 17.73 years. The study included a total of 7,023 participants, with a mean age of 50.54 ± 17.73 years. The majority of participants were Mexican American, comprising 52.5% of the total sample. The average initiation time for the first meal was recorded as 08:25. The median time for the second meal was observed to be 14:17. The average time for the final meal was recorded at 20:09. The average total duration of meal consumption was 12.16 hours. The average frequency of meals consumed was 4.70 times per day.

In this study, 2,965 participants (42.22%) were categorized as having good sleep quality, 2,339 (33.30%) as having moderate sleep quality, and 1,719 (24.48%) as having poor sleep quality. In the group with good sleep quality, several notable characteristics were observed. This group tended to be older, had more males, fewer smokers, and exhibited higher dietary quality. Among those with good sleep quality, the average time for the first meal was 08:23, the midpoint of meal was at 14:14, and the last meal was at 20:06. The total duration of dietary intake was approximately 12.09 hours, with an average of 4.63 meals per day. Compared to participants with poor sleep quality, those with good sleep quality tended to have earlier first, middle, and last meal timings and fewer meal occasions.

Fig 1 illustrates the relationship between the three-category variables of meal timing, meal durations, meal occasions, and sleep quality using bar graphs and error bars. We refined our categorization approach, segregating the population into three distinct groups based on individual parameters: first meal timing, meal midpoint, last meal timing, meal duration, and meal occasions. The results indicated that, compared to participants in the first tertile of meal occasions and first, midpoint, and last meal timings, those in the last tertile of these factors had higher PSQI scores.

Table 2 presents the relationship between meal timing and PSQI scores. Across all models, later meal timings and more meal occasions are associated with higher PSQI scores. No significant associations were found between meal durations and PSQI scores. After adjusting for all covariates, compared to the first tertile, the β (95% CI) of the third tertile is 0.545 (0.226, 0.864) for first meal timing, 0.586 (0.277, 0.896) for midpoint meal timing, and 0.385 (0.090, 0.680) for last meal timing, respectively. Additionally, more meal occasions are associated with higher PSQI scores. After adjusting for all covariates, compared to the first tertile of meal occasions, the β (95% CI) of the third tertile is 0.332 (0.021, 0.642).

The data central to this research is drawn from open-access sources and can be retrieved from the NHANES official portal at: https://wwwn.cdc.gov/nchs/nhanes/search/default.aspx↗.