ABSTRACT
Background
There is increasing recognition that a morning or evening preference is associated with time of eating, metabolic health, and morbidity. However, few studies have examined the association of time of eating with mortality.
Objectives
To examine the association of time of first recalled ingestive episode with the prospective risk of all-cause mortality.
Methods
We used mortality-linked data from the NHANES conducted in 1988–1994 and 1999–2014 (n = 34,609; age ≥ 40 years). The exposure was quartiles (Q1–Q4) of clock time of first eating episode self-reported in the baseline 24-hour dietary recall. The outcome was follow-up time from the date of NHANES examination to the date of death or end of the follow-up period (31 December 2015). We used proportional hazards regression methods to determine the independent association of time of first eating episode with relative hazard of all-cause mortality, with adjustments for multiple covariates and the complex survey design. Multiple linear regression methods were used to examine the associations of time of first eating episode with baseline cardiometabolic biomarkers and dietary attributes.
Results
In this national cohort, with a median age of ∼55 years (95% CI: 54.6–55.4 years) at baseline and a median follow-up of 8.3 years (IQR, 8.75 years), there were 10,303 deaths. The median times of first eating episodes in Q1–Q4 were 05:45, 07:00, 08:00, and 10:00, respectively. Covariate-adjusted relative hazards of mortality in Q1 to Q3 of the time of the first eating episode were 0.88 (95% CI: 0.81–0.96), 0.88 (95% CI: 0.81–0.95), 0.94 (95% CI: 0.87–1.02), with Q4 as the referent (P = 0.0008). Qualitative dietary attributes were inversely related with the time of the first eating episode; however, BMI and serum concentrations of glycemic biomarkers increased with later times of first eating episode (P ≤ 0.0001).
Conclusions
Recall of an earlier time of the first eating episode by ≥40-year-old US participants was suggestive of a small relative survival advantage in this observational study.
Keywords: NHANES, time of eating, all-cause mortality, temporal eating behaviors, biomarkers
Introduction
Daily feeding and fasting cycles are among the external factors that entrain peripheral tissue clocks and their metabolic activities (1–5). Interactions of central and peripheral clocks and a so-called secondary brain “food clock,” or timing of food intake, contribute to food anticipatory behaviors and secretion of feeding-related hormones (4). However, the circadian timing of peak concentration of glucocorticoids is linked to the central clock in the suprachiasmatic nucleus and may be independent of the anticipated time of food intake (4, 5). Available evidence suggests that feeding early in the wake cycle is associated with greater insulin sensitivity and glucose tolerance, and higher resting and diet-induced energy expenditure (3, 6–11). These observations suggest that misalignment of sleep-wake cycles from feeding-fasting cycles may have the potential to influence energy expenditure and macronutrient metabolism. Thus, eating later in the day may promote energy storage and metabolic dysfunction, with consequent adverse metabolic and health outcomes (12).
Circadian timing has also been described as a morning/evening preference or “chronotype,” and an evening chronotype has been linked to a later time of eating, higher risk of obesity, metabolic syndrome, high blood pressure, poor glycemic control, and breast and prostate cancers (13–20). In 2 European cohorts, a self-reported evening preference was associated with a higher risk of all-cause mortality (21, 22). However, no published studies have examined the associations of a morning preference with biomarkers, diet quality, and risks of mortality in a nationally representative US cohort.
Given the published evidence on timing of eating in relation to chronotype (13, 14), we hypothesized that the reported time of the first eating episode of the day may serve as a proxy for a morning preference, and we examined the prospective association of the recalled time of the first ingestive episode with the risk of all-cause mortality in a nationally representative cohort of US men and women. To understand potential drivers of the association of the time of the first eating episode with mortality, we also examined the cross-sectional association of the time of first eating with cardiometabolic biomarkers and dietary nutrient intakes reported at baseline.
Methods
We used data from the NHANES III, conducted from 1988 to 1994, and the continuous NHANES, conducted from 1999 to 2014, for this prospective cohort study (23, 24). Our use of anonymized extant public domain data was deemed exempt from Human Subjects Review by the City University of New York institutional review board. The NHANES, conducted by the National Center for Health Statistics (NCHS), provide information about general health and nutritional status of a representative sample of the US population. The survey procedures include a household interview and a medical examination of the sample participant in a specially equipped mobile examination center (MEC). The MEC visit also includes anthropometric measurements and collection of blood and urine specimens and dietary information using standardized protocols. Unweighted response rates for the MEC sample in the NHANES III and the continuous NHANES 1999–2010 were >70%; in the 2011–12 and 2013–2014 survey cycles, the rates were ∼68% (25).
Exposure assessment
The NHANES protocol includes collection of an in-person, computer-assisted, 24-hour recall of dietary intake, conducted by a trained dietary interviewer (26). Surveys conducted from 2003 onwards also included a second recall, administered via telephone, 3–10 days after the MEC visit. However, exposure for the current study was determined from the first recall. The dietary interview queried about details of food/beverage intake and the clock time of the start of each recalled ingestive episode. We considered the recalled clock time of the first eating/drinking event of the 24-hour recall as the time of the first eating episode. Events where the only reported item was plain water were not considered ingestive episodes. We have previously used these methods to examine trends in eating patterns of US adults and children (27–29). The primary exposure variable, the reported time of the first ingestive episode in the 24-hour dietary recall, was operationalized as survey-weighted, sex-specific quartiles.
Primary study outcome
The primary study outcome was mortality from all causes. Mortality outcome data of the NHANES III and the continuous NHANES 1999–2014 respondents, with follow-up to 31 December 2015, are available in the public domain (30). Briefly, the NCHS used standardized protocols to determine the mortality status of each survey participant by linkage to the National Death Index and databases of the Social Security Administration and Centers for Medicare and Medicaid Services (31). The public domain Linked Mortality File includes information on person-months of follow-up from the date of MEC examination to the date of death or the end of the mortality follow-up period, 31 December 2015.
Secondary outcomes
Dietary outcomes included intakes of energy, key nutrients [percentages of energy from dietary fat, dietary fiber (g), vitamin C (mg), potassium (mg), magnesium (mg), and sodium (mg)], food groups [whole grains (oz), fruits (cup servings), vegetables (cup servings), and added sugar (teaspoons)], overall diet quality, and other temporal variables (clock time of last eating episode, duration of the 24-hour ingestion period or the eating window, and intervals between eating episodes) (27–29), reported in the 24-hour diet recall. The public domain data for each NHANES respondent include energy and nutrient intake information computed by the NCHS using the USDA's Food and Nutrient Database for Dietary Studies (26). The NHANES dietary intake data have also been linked to the Food Patterns Equivalents Database (FPED) and MyPyramid Equivalents Database (MPED) to allow disaggregation into serving equivalents of food groups (32). We also computed a measure of overall diet quality, the Healthy Eating Index–2015 (HEI-2015), using the information from the FPED/MPED linked database (33). An additional dietary outcome was a second measure of diet quality: the energy density of reported foods and beverages (kcal/g) (34).
The examined cardiometabolic biomarkers included measured BMI (kg/m2), serum total and HDL cholesterol, C-reactive protein, glycated hemoglobin, fasting triglycerides, fasting glucose, fasting insulin, and fasting C-peptide. The biomarkers were assayed using standardized assay procedures in blood samples collected during the MEC visit (35).
Analytic cohort
All men and women aged ≥40 years with an in-person dietary recall at baseline in NHANES 1988–1994 and 1999–2014 were eligible for inclusion in the study cohort (n = 36,854). We excluded respondents with dietary recalls considered unreliable by the NCHS (n = 2170), reporters of no energy intake in a recall considered reliable by the NCHS (n = 2), women who were pregnant or lactating at baseline (n = 39), and those missing a National Death Index match at follow-up (n = 34). The final analytic cohort included 16,884 men and 17,725 women (n = 34609; Supplemental Figure 1).
Covariate information
The available information on potential confounders of the association between the time of eating and mortality was collected in each survey and included age, race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican-American, and all other race/ethnicities), family income (as a percentage of the poverty threshold), years of education, smoking status, alcohol intake, any leisure-time physical activity, a self-report of doctor-diagnosed chronic disease (high blood pressure, diabetes, coronary heart disease, and cerebrovascular disease), and BMI (kg/m2; computed from measured height and weight).
Statistical methods
Descriptive methods
We describe characteristics of the analytic cohort by survey-weighted, sex-specific quartiles of recalled time of the first eating episode of the day. For each variable examined, we used the chi-square test of independence to indicate significance of the observed association with the time of the first eating episode.
Hypothesis testing
Prospective association of time of first eating episode with primary outcome of all-cause mortality
We used Cox proportional hazards regression methods to examine the relative hazard of mortality from all causes in relation to quartiles of the recalled time of the first eating episode. These regression models adjusted for established confounders of the association and included sex, age, age2, race/ethnicity, family income, level of education, BMI, leisure time physical activity, smoking status, alcohol use status, and self-report of a doctor-diagnosed history of chronic disease. In other iterations of this basic model, we also adjusted for the time of the last eating episode of the recall, diet quantity (total energy intake), and diet quality (HEI-2015).
Respondents missing information on education (n = 101), smoking status (n = 23), and any leisure-time physical activity (n = 6) were excluded from these models. Larger numbers were missing the family income (n = 3064), BMI (n = 560), and alcohol use (n = 1574) information; for each of these variables a missing category was included in regression models. Hazard rates of mortality associated with the first 3 quartiles of the time of the first eating episode were compared to the referent, fourth quartile. To examine linear trends, we operationalized the median reported time of the first eating episode for each quartile as a trend variable with scores equal to medians. We also examined the time of the first eating episode as a continuous exposure variable.
We examined the interactions of the time of the first eating episode with sex, age, and BMI at baseline to examine whether the associations of the time of the first eating episode and mortality differed by sex, age, or BMI.
Sensitivity analysis
To examine the possibility that the association between the time of the first eating episode and mortality may reflect reverse causation or baseline morbidity, we examined these associations after excluding the first 2 years of follow-up and in models stratified by self-reported morbidity at the baseline interview. Similarly, analyses were repeated with the exclusion of accidental and unknown causes of death, and the exclusion of respondents reporting low or high energy intakes in the 24-hour dietary recall (<800 or >4200 kcal in men; <600 or >3500 kcal in women). Because eating times may differ on weekdays and weekends we also examined the association of the time of the first eating episode and mortality using survey-weighted, gender- and weekday-/weekend-specific quartiles of the time of the first eating event.
We operationalized a test of the proportional hazards assumption by: 1) dividing the follow-up time into 3 disjointed intervals that covered the length of follow-up, where each interval had about one-third of the deaths; 2) defining 2 dummy variables that were dependent on the follow-up time, indicating which interval included a survey respondent's follow-up time; and 3) testing the statistical significance of the interaction between these 2 dummy variables and the terms indicating the quartiles of the clock time of the first eating event.
Cross-sectional associations of time of first eating episode exposure with secondary outcomes of dietary variables and cardiometabolic biomarkers
We used multiple linear regression models to examine the associations of the time of the first eating episode with key cardiometabolic biomarkers and dietary outcomes adjusted for covariates. The covariates common to models for both biomarker and dietary outcomes included sex; age; race/ethnicity; family income, assessed as poverty income ratio; years of education; BMI; any leisure time physical activity; and self-report of a history of chronic disease. For biomarker outcomes, smoking status, alcohol use status, and hours of fasting before phlebotomy were additional covariates. For dietary outcomes, the models also included the month of dietary recall, day of week of recalled intake, and energy intake. Results present the covariate-adjusted predicted margins (or adjusted means) (36) and their 95% CIs for each biomarker or dietary variable by survey-weighted, sex-specific quartiles of recalled time of first eating episode. For serum total cholesterol and HDL cholesterol, the distributions were approximately normal; for these biomarkers, we present adjusted means with 95% CIs. Linear regression analyses for all other cardiometabolic biomarkers used log-transformed values, and estimates were back-transformed to obtain adjusted geometric means and 95% CIs. Results with 2-sided P values < 0.05 without adjustment for multiple comparisons were considered statistically significant.
We used SAS (version 9.4; SAS Institute, Inc), and SAS callable SUDAAN software (version 11.0.1; RTI International) for statistical analyses. We followed the NHANES analytic guidelines for all procedures and used appropriate survey sample weights to adjust for survey nonresponse and differential probabilities of selection of some oversampled population groups (26).
Results
Descriptive
In this national cohort, with a median age of ∼55 years (95% CI: 54.6–55.4) at baseline and over a median follow-up of 8.3 years (IQR, 8.75 years), there were 10,303 deaths due to all causes (Table 1). The median times of the first eating episode were 07:30 in men and 08:00 in women. The median times of the first eating episode in Q1 to Q4 were 05:30, 07:00, 08:00, and 10:00, respectively, in men and 06:00, 07:15, 08:00, and 09:45, respectively, in women. In both men and women, the variability (IQR) in the first and the fourth quartiles was higher relative to the second and third quartiles.
TABLE 1.
Characteristics of the study cohort at baseline (survey weighted percentage and 95% CI) for all participants and by survey-weighted, sex-specific quartiles of recalled time of first eating episode in a 24-hour recall
| Survey-weighted, sex-specific approximate quartiles of reported time of first eating episode in the 24-hour recall1 | |||||
|---|---|---|---|---|---|
| All n = 34,609 | Quartile 1, n = 6702; 3352 men; 3350 women | Quartile 2, n = 7434; 3447 men; 3987 women | Quartile 3, n = 9740; 5223 men; 4517 women | Quartile 4, n = 10,733; 4862 men; 5871 women | |
| Median time and interquartile range (minutes) of first eating episode in the 24-h dietary recall | |||||
| All | 07:45 (150) | 05:45 (90) | 07:00 (30) | 08:00 (30) | 10:00 (120) |
| Men | 07:30 (135) | 05:30 (120) | 07:00 (30) | 08:00 (30) | 10:00 (120) |
| Women | 08:00 (120) | 06:00 (90) | 07:15 (30) | 08:00 (30) | 09:45 (120) |
| Number of events | 10,303 | 1767 | 2361 | 3256 | 2919 |
| Follow-up time, years, median (IQR) | 8.33 (8.75) | 9.08 (8.58) | 8.50 (9.17) | 8.17 (8.75) | 7.75 (8.25) |
| Survey weighted % (95% CI) | |||||
| Men, n = 16,884 | 46.9 (46.2–47.5) | 47.2 (45.7–48.7) | 45.0 (43.4–46.6) | 51.5 (50.2–52.8) | 43.6 (42.2–44.9) |
| Women, n = 17,725 | 53.1 (52.5–53.8) | 52.8 (51.3–54.3) | 55.0 (53.3–56.6) | 48.5 (47.2–49.8) | 56.4 (55.1–57.8) |
| Age group, years | |||||
| 40–59, n = 16,549 | 59.9 (59.0–60.9) | 66.8 (65.3–68.2) | 56.5 (54.8–58.2) | 53.3 (51.7–54.9) | 63.9 (62.5–65.3) |
| ≥60, n = 18,060 | 40.0 (39.1–41.0) | 33.2 (31.8–34.7) | 43.5 (41.7–45.2) | 46.7 (45.1–48.3) | 36.1 (34.7–37.5) |
| Race/ethnicity | |||||
| Non-Hispanic White, n = 17,349 | 75.7 (73.8–77.5) | 81.1 (79.5–83.0) | 82.3 (80.5–84.0) | 77.5 (75.5–79.5) | 63.4 (60.7–66.0) |
| Non-Hispanic Black, n = 7472 | 10.2 (9.2–11.4) | 7.8 (6.8–8.8) | 6.0 (5.3–6.9) | 8.8 (7.9–9.8) | 17.4 (15.6–19.3) |
| Mexican-American, n = 6220 | 5.2 (4.5–6.1) | 3.9 (3.4–4.4) | 4.1 (3.3–5.0) | 5.3 (4.4–6.4) | 7.3 (6.2–8.7) |
| All other races/ethnicities, n = 3568 | 8.8 (7.9–9.8) | 7.0 (6.0–8.2) | 7.6 (6.6–8.7) | 8.4 (7.4–9.4) | 11.9 (10.5–13.4) |
| Poverty income ratio, % | |||||
| <130, n = 9031 | 17.0 (15.8–18.2) | 14.5 (13.3–15.7) | 13.1 (11.8–14.6) | 16.2 (14.8–17.8) | 23.1 (21.6–24.7) |
| 130–349, n = 12,506 | 33.5 (32.4–34.6) | 34.1 (32.3–36.0) | 32.5 (30.7–34.5) | 32.9 (31.2–34.6) | 34.3 (32.7–35.9) |
| ≥350, n = 10,008 | 42.7 (41.1–44.4) | 45.4 (43.2–47.7) | 47.6 (45.2–49.9) | 44.0 (41.7–46.3) | 35.0 (33.1–36.9) |
| Missing, n = 3064 | 6.8 (6.3–7.4) | 6.0 (5.1–6.9) | 6.7 (5.9–7.7) | 6.9 (6.1–7.8) | 7.6 (6.9–8.4) |
| Education, years | |||||
| <12, n = 12,215 | 20.9 (19.9–22.0) | 19.1 (17.7–20.6) | 17.5 (16.1–19.0) | 21.0 (19.7–22.3) | 25.4 (23.9–26.8) |
| 12, n = 8384 | 25.8 (24.9–26.7) | 29.4 (27.7–31.1) | 24.7 (23.2–26.3) | 25.1 (23.6–26.5) | 24.5 (23.2–25.9) |
| Some college, n = 7579 | 26.8 (25.9–27.6) | 28.6 (27.0–30.3) | 26.3 (24.6–27.9) | 25.5 (24.2–26.9) | 26.9 (25.5–28.4) |
| ≥College, n = 6330 | 26.5 (25.1–27.9) | 22.9 (20.9–25.0) | 31.5 (29.2–33.8) | 28.4 (26.6–30.2) | 23.2 (21.6–24.9) |
| BMI, kg/m2 | |||||
| <25, n = 9666 | 28.7 (27.9–29.5) | 29.4 (27.8–31.2) | 31.3 (29.7–32.9) | 28.1 (26.8–29.5) | 26.4 (25.2–27.6) |
| 25 to <30, n = 12,527 | 35.5 (34.7–36.3) | 36.5 (34.8–38.2) | 37.3 (35.5–39.1) | 35.5 (34.3–36.7) | 33.2 (31.8–34.7) |
| ≥30, n = 11,586 | 34.3 (33.4–35.3) | 33.0 (31.3–34.7) | 30.2 (28.6–31.9) | 34.9 (33.4–36.5) | 38.5 (37.0–40.0) |
| Missing, n = 560 | 1.4 (1.2–1.6) | 1.0 (0.8–1.4) | 1.2 (0.9–1.6) | 1.4 (1.1–1.9) | 1.8 (1.4–2.4) |
| Alcohol use status | |||||
| Never drinker, n = 5344 | 12.0 (11.2–13.0) | 8.8 (7.9–9.9) | 12.5 (11.3–13.9) | 12.7 (11.2–14.2) | 13.7 (12.4–15.0) |
| Former drinker, n = 7885 | 18.1 (17.3–18.9) | 18.1 (16.8–19.4) | 18.0 (16.6–19.4) | 18.0 (16.8–19.2) | 18.4 (17.3–19.6) |
| Current drinker, n = 19,806 | 65.8 (64.4–67.2) | 69.4 (67.6–71.2) | 65.7 (63.8–67.6) | 65.7 (63.7–67.7) | 62.9 (61.1–64.7) |
| Missing, n = 1574 | 4.0 (3.7–4.4) | 3.6 (3.0–4.4) | 3.7 (3.1–4.5) | 3.6 (3.2–4.2) | 5.0 (4.4–5.6) |
| Any leisure physical activity | |||||
| Yes, n = 17,631 | 56.6 (55.2–58.0) | 58.4 (56.3–60.4) | 62.3 (60.0–64.4) | 57.4 (55.6–59.2) | 49.3 (47.7–51.0) |
| No, n = 16,972 | 43.4 (42.0–44.8) | 41.6 (39.6–43.7) | 37.7 (35.6–39.9) | 42.6 (40.8–44.4) | 50.7 (49.0–52.3) |
| Cigarette smoking status | |||||
| Never smoked, n = 16,656 | 48.2 (47.2–49.1) | 42.5 (40.8–44.3) | 49.2 (47.5–50.9) | 49.8 (48.3–51.3) | 50.3 (48.7–51.9) |
| Former smoker, n = 11,102 | 32.0 (31.2–32.9) | 30.9 (29.4–32.5) | 33.9 (32.4–35.3) | 34.3 (32.9–35.8) | 29.0 (27.8–30.4) |
| Current smoker, n = 6828 | 19.8 (19.0–20.6) | 26.5 (25.2–27.9) | 16.9 (15.7–18.2) | 15.9 (14.7–17.1) | 20.6 (19.4–22.0) |
| Self-report of doctor-diagnosed chronic disease | |||||
| No, n = 16,675 | 53.1 (52.2–54.1) | 55.6 (53.7–57.3) | 55.1 (53.3–56.8) | 51.0 (49.5–52.6) | 51.5 (50.0–53.1) |
| Yes, n = 17,934 | 46.9 (45.9–47.8) | 44.4 (42.7–46.2) | 44.9 (43.2–46.7) | 48.9 (47.4–50.5) | 48.4 (46.9–50.0) |
| Day of recalled intake | |||||
| Monday–Friday, n = 24,009 | 72.7 (71.8–73.6) | 79.6 (78.5–80.7) | 77.6 (76.3–78.9) | 70.6 (69.1–72.0) | 64.7 (63.1–66.2) |
| Saturday–Sunday, n = 10,600 | 27.3 (26.4–28.2) | 20.4 (19.3–21.5) | 22.3 (21.1–23.7) | 29.4 (28.0–30.9) | 35.3 (33.7–36.9) |
The chi-square test of independence was significant (P < 0.0001) for all listed variables. Respondents missing information on education (n = 101), smoking status (n = 23), and physical activity (n = 7) were excluded.
The distributions of all examined covariates differed among quartiles of the reported time of the first eating episode (P < 0.0001; Table 1). In the first quartile of time of the first eating episode, respondents were more likely to be aged 40–59 years, be non-Hispanic Whites, have a BMI <25 kg/m2, be current alcohol users, report some leisure-time physical activity, be current smokers, and have no chronic disease at baseline. Respondents with the lowest and the highest levels of education were less likely to be in the first quartile of time of the first eating episode. Respondents who recalled a weekend day of intake were more likely to be in the fourth quartile of time of the first eating episode.
All-cause mortality outcome
The covariate-adjusted hazard of mortality increased with a later time of the first eating episode of the day (Table 2). Overall, respondents reporting in the first 2 quartiles of time of the first eating episode had a 12% lower hazard of mortality from all causes relative to those in the fourth quartile (P for trend across quartiles = 0.0005). The observed associations were unchanged by further adjustments for time of last eating episode and dietary variables (total energy intake and HEI-2015). In all iterations, the association remained significant when the time of first eating was operationalized as a continuous exposure variable (≤0.006). The association between the time of the first eating episode and mortality did not differ by sex (P for interaction = 0.9), age (P for interaction = 0.1), or BMI (P for interaction = 0.9); hence, results in Table 2 are not stratified by sex, age, or BMI.
TABLE 2.
Covariate-adjusted risk of mortality from all causes (HR and 95% CI) in relation to approximate quartiles of recalled time of first eating episode in a national cohort1
| Survey-weighted, sex-specific approximate quartiles of reported time of first eating episode in the 24-hour recall | Trend2 across quartiles | ||||
|---|---|---|---|---|---|
| Quartile 1 | Quartile 2 | Quartile 3 | Quartile 4 | ||
| Model 13 | 0.88 (0.81–0.95) | 0.88 (0.81–0.95) | 0.93 (0.86–1.01) | 1.0 | P = 0.0005 |
| Model 24 | 0.89 (0.82–0.97) | 0.88 (0.82–0.96) | 0.94 (0.87–1.01) | 1.0 | P = 0.001 |
| Model 35 | 0.88 (0.81–0.95) | 0.88 (0.81–0.95) | 0.93 (0.86–1.01) | 1.0 | P = 0.0006 |
| Model 46 | 0.88 (0.81–0.96) | 0.88 (0.81–0.95) | 0.94 (0.87–1.02) | 1.0 | P = 0.0008 |
Estimates are HRs and 95% CIs from proportional hazards regression models with person-months of follow-up from the NHANES exam date to date of death or end of the follow-up period (31 December 2015) as the dependent variable. Respondents missing information on education, smoking status, and physical activity (n = 122) were excluded. n = 34,487; events = 10,238
Trend was operationalized as sex-specific median reported time in each quartile of reported time of first eating episode.
3Independent variables in Model 1 included: time of first eating episode (quartiles), age (continuous), age2, sex (men, women), race (non-Hispanic White, non-Hispanic Black, Mexican-American, Others), poverty income ratio (<130%, 130%–349%, ≥349%, missing), education (<12 years, 12 years, some college, ≥college), BMI (<25, 25–29.9, or ≥30 kg/m2 or missing), leisure time physical activity (yes, no), any self-reported doctor-diagnosed chronic disease (yes, no), smoking status (never, former, current smoker), and alcohol use status (never, former, current drinker, unknown).
Model 2 included all covariates in Model 1 and 24-hour energy intake (kcal).
Model 3 included all covariates in Model 2 and reported time of the last eating episode of the 24-hour recall.
Model 4 included all variables in Model 3 and Healthy Eating Index–2015 score.
Sensitivity analysis
The lower hazard of mortality associated with an earlier time of the first eating episode remained significant after exclusion of deaths due to accidental or unknown causes or deaths that occurred in the first 2 years of follow-up; in models stratified by self-report of a doctor-diagnosed morbidity at baseline; and after exclusion of reporters of low or high energy intake on the recall day (Table 3). The association between mortality and the time of the first eating episode remained significant when exposure was operationalized as weekday-/weekend- and gender-specific quartiles (P for trend = 0.001; Table 4).
TABLE 3.
Sensitivity analysis: covariate-adjusted HR and 95% CI of mortality from all causes in relation to approximate quartiles of recalled time of first eating episode in a national cohort1
| Survey-weighted, sex-specific approximate quartiles of reported time of first eating episode in the 24-hour recall | Trend2 across quartiles | ||||
|---|---|---|---|---|---|
| Quartile 1 | Quartile 2 | Quartile 3 | Quartile 4 | ||
| Excluded 257 deaths from accidental and unknown causes: n = 34,487; events = 9981 | 0.86 (0.79–0.93) | 0.87 (0.81–0.95) | 0.93 (0.86–1.01) | 1.0 | 0.00007 |
| Excluded first 2 years of follow-up: n = 31,447; events = 8961 | 0.90 (0.83–0.99) | 0.92 (0.83–1.0) | 0.95 (0.87–1.04) | 1.0 | 0.02 |
| No self-report of doctor-diagnosed chronic disease at baseline: n = 16,613; events = 3738 | 0.86 (0.75–0.99) | 0.88 (0.79–0.99) | 0.87 (0.77–0.99) | 1.0 | 0.04 |
| Self-report of doctor-diagnosed chronic disease at baseline: n = 17,874; events = 6500 | 0.90 (0.81–0.99) | 0.88 (0.79–0.97) | 0.98 (0.89–1.07) | 1.0 | 0.01 |
| Excluded reporters of low or high energy intake3: n = 32,531; events = 9662 | 0.90 (0.82–0.98) | 0.88 (0.81–0.95) | 0.95 (0.87–1.02) | 1.0 | 0.004 |
Estimates are HRs and 95% CIs from proportional hazards regression models with number of person-months of follow-up from the NHANES exam date to the date of death or end of the follow-up period (31 December 2015), as the dependent variable. Independent variables included: time of first eating episode (quartiles), age (continuous), age2, sex (men, women), race (non-Hispanic White, non-Hispanic Black, Mexican-American, Others), poverty income ratio (<130%, 130%–349%, ≥349%, missing), education (<12 years, 12 years, some college, ≥college), BMI (<25, 25–29.9, or ≥30 kg/m2 or missing), any leisure time physical activity (yes, no), any self-report of doctor-diagnosed chronic disease (yes, no), smoking status (never, former, current smoker), alcohol use status (never, former, current drinker, unknown), 24-hour energy intake (continuous), time of last eating episode (continuous), and Healthy Eating Index–2015 (continuous).
Trends were operationalized with sex indicating the specific median-reported time in each quartile of reported time of first eating episode.
Defined as <800 or >4200 kcal/24-hour recall in men and <600 or >3500 kcal/24-hour recall in women.
TABLE 4.
Covariate-adjusted HR and 95% CI of mortality from all-causes in relation to survey-weighted, sex and weekday/weekend specific quartiles of recalled time of first eating episode in a national cohort1
| Survey-weighted, sex and weekday/weekend-specific, approximate quartiles of reported time of first eating episode in the 24-hour recall | Trend2 across quartiles | ||||
|---|---|---|---|---|---|
| Quartile 1 | Quartile 2 | Quartile 3 | Quartile 4 | ||
| Median and interquartile range (minutes) of time of first eating episode on weekdays (Monday-Friday) | |||||
| Weekdays, men3 | 05:30 (120) | 07:00 (30) | 08:00 (30) | 0905 (90) | |
| Weekdays, women3 | 06:00 (90) | 07:00 (30) | 08:00 (30) | 0945 (120) | |
| Median and interquartile range (minutes) of time of first eating episode on weekend days (Saturday-Sunday) | |||||
| Weekends, men4 | 05:45 (90) | 07:00 (30) | 08:00 (30) | 1000 (120) | |
| Weekends, women4 | 06:00 (120) | 07:25 (30) | 08:00 (30) | 0945 (120) | |
| Model 15 | 0.89 (0.83–0.97) | 0.87 (0.81–0.94) | 0.96 (0.88–1.04) | 1.0 | P = 0.001 |
| Model 26 | 0.91 (0.84–0.98) | 0.87 (0.81–0.94) | 0.96 (0.89–1.04) | 1.0 | P = 0.003 |
| Model 37 | 0.89 (0.82–0.97) | 0.86 (0.80–0.93) | 0.96 (0.88–1.04) | 1.0 | P = 0.001 |
| Model 48 | 0.90 (0.83–0.97) | 0.87 (0.81–0.94) | 0.96 (0.89–1.04) | 1.0 | P = 0.001 |
Estimates are HRs and 95% CIs from proportional hazards regression models with number of person-months of follow-up from the NHANES exam date to the date of death or end of the follow-up period (31 December 2015), as the dependent variable. n = 34,487; events = 10,238.
Trends were operationalized as the sex-specific, weekday-/weekend-specific median reported time in each quartile of reported time of first eating episode.
Median time (IQR), in minutes, of first eating episode on weekdays (Monday–Friday).
Median time (IQR), in minutes, of first eating episode on weekend days (Saturday–Sunday)
Independent variables in Model 1 included the time of first eating episode (quartiles), age (continuous), age2, sex (men, women), race (non-Hispanic White, non-Hispanic Black, Mexican-American, Others), poverty income ratio (<130%, 130%–349%, ≥349%, missing), education (<12 years, 12 years, some college, ≥college), BMI (<25, 25–29.9, or ≥30 kg/m2 or missing), leisure time physical activity (yes, no), any self-reported doctor-diagnosed chronic disease (yes, no), smoking status (never, former, current smoker), and alcohol use status (never, former, current drinker, unknown).
Model 2 included all covariates in Model 1 and 24-hour energy intake (kcal; continuous).
Model 3 included all covariates in Model 2 and the reported time of the last eating episode of the 24-hour recall (continuous).
Model 4 included all variables in Model 3 and the Healthy Eating Index–2015 (continuous).
The results for the tests for proportional hazards showed that the HRs of mortality by quartiles of the first ingestive episode varied significantly (P < 0.001) across the 3 follow-up intervals (<60, 61–128, and ≥129 months). For example, for Model 1, the Q1 to Q4 HRs were 0.71 (95% CI: 0.61–0.83), 0.75 (95% CI: 0.66–0.87), 0.87 (95% CI: 0.77–0.98), and 1.0, respectively, for the earliest interval; 0.99 (95% CI: 0.85–1.17), 0.98 (95% CI: 0.85–1.10), 0.98 (95% CI: 0.86–1.13), and 1.0, respectively, for the second interval; and 0.97 (95% CI: 0.84–1.12), 0.93 (95% CI: 0.79–1.09), 0.97 (95% CI: 0.85–1.18), and 1.0, respectively, for the latest interval.
Dietary and temporal eating attributes
Intakes of energy and energy-adjusted dietary fiber, potassium, magnesium, servings of whole grains, and fruits, and the HEI-2015 score decreased, but the energy density (kcal/g) of beverages (but not foods) increased with a later time of the first eating episode (P < 0.0001; Table 5). The percentage of energy from fat, teaspoons of added sugar, sodium, and vegetable servings were not related to the time of the first eating episode.
TABLE 5.
Covariate-adjusted means and 95% CIs of dietary and temporal attributes self-reported in the in-person 24-hour recall, by approximate quartiles of time of first eating episode1
| Survey-weighted, sex-specific approximate quartiles of time of first eating episode in the 24-hour dietary recall | Trend2 across quartiles | ||||
|---|---|---|---|---|---|
| Dietary Variable | Quartile 1 | Quartile 2 | Quartile 3 | Quartile 4 | |
| Dietary energy, nutrient, food group, and overall diet quality variables | |||||
| 24-hour energy intake, kcal | 2150 (2120–2180) | 2030 (2010–2060) | 2040 (2010–2060) | 1920 (1890–1950) | <0.0001 |
| Energy from fat, % | 34 (33–34) | 34 (33–34) | 34 (33–34) | 34 (34–34) | 0.8 |
| Fiber, g | 16 (16.0–16.7) | 17 (16.6–17.3) | 17 (16.6–17.4) | 16 (15.6–16.2) | 0.01 |
| Vitamin C, mg | 88 (84–91) | 92 (89–96) | 91 (88–94) | 86 (83–89) | 0.2 |
| Potassium, mg | 2860 (2820–2900) | 2810 (2770–2840) | 2770 (2730–2800) | 2630 (2600–2670) | <0.0001 |
| Magnesium, mg | 302 (297–306) | 299 (295–303) | 295 (291–300) | 282 (278–287) | <0.0001 |
| Sodium, mg | 3290 (3240–3330) | 3340 (3290–3390) | 3320 (3280–3360) | 3300 (3260–3340) | 0.9 |
| Added sugar, tsp | 16.6 (16.1–17.2) | 16.3 (15.8–16.8) | 16.4 (16.0–16.8) | 17.0 (16.5–17.4) | 0.2 |
| Whole grains, oz equivalents | 0.86 (0.81–0.91) | 0.91 (0.85–0.96) | 0.91 (0.86–0.96) | 0.76 (0.72–0.80) | 0.0002 |
| Fruit serving, cup equivalents | 1.07 (1.02–1.13) | 1.08 (1.04–1.13) | 1.07 (1.02–1.11) | 0.99 (0.94–1.03) | 0.002 |
| Vegetable serving, cup equivalents | 1.60 (1.56–1.64) | 1.67 (1.63–1.72) | 1.67 (1.62–1.71) | 1.60 (1.56–1.64) | 0.8 |
| Healthy Eating Index–2015 | 52.1 (51.6–52.6) | 52.7 (52.2–53.3) | 52.7 (52.2–53.2) | 51 (50.5–51.5) | <0.0001 |
| Energy density of foods and beverages, kcal/g | 0.84 (0.83–0.85) | 0.88 (0.87–0.89) | 0.91 (0.90–0.92) | 0.96 (0.95–0.97) | <0.0001 |
| Energy density of foods,3 kcal/g | 1.90 (1.87–1.92) | 1.85 (1.83–1.86) | 1.84 (1.82– 1.86) | 1.88 (1.86–1.91) | 0.7 |
| Energy density of beverages,4, 5 kcal/g | 0.24 (0.24–0.25) | 0.24 (0.24–0.25) | 0.26 (0.26–0.27) | 0.29 (0.28–0.29) | <0.0001 |
| Temporal variables | |||||
| Length of the 24-hour ingestion period, hours | 15.0 (14.9–15.1) | 12.9 (12.9–13.0) | 12.1 (12.0–12.1) | 10.0 (9.9–10.1) | <0.0001 |
| 24-hour clock time of last eating episode | 19:55 (19:51–19:59) | 19:59 (19:56–20:02) | 20:09 (20:06–2012) | 20:18 (20:14–20:22) | <0.0001 |
| Number of eating episodes,6 n | 5.6 (5.5–5.6) | 5.2 (5.1–5.3) | 4.9 (4.9–5.0) | 4.4 (4.4–4.5) | <0.0001 |
| Intervals between eating episodes, hours | 2.9 (2.88–2.95) | 2.7 (2.6–2.7) | 2.6 (2.59–2.64) | 2.4 (2.37–2.42) | <0.0001 |
Estimates are predicted margins (adjusted means) and 95% CIs from linear regression models with each dietary or temporal variable as a continuous dependent variable. The independent variables included time of first eating episode (quartiles), age (continuous), sex (men, women), race (non-Hispanic White, non-Hispanic Black, Mexican-American, Others), poverty income ratio (<130%, 130%–349%, ≥349%, missing), education (<12 years, 12 years, some college, ≥college), BMI (<25, 25–29.9, or ≥30 kg/m2 or missing), leisure time physical activity (yes, no), any self-reported doctor-diagnosed chronic disease (yes, no), month of recall (November–April, May–October), day of recall (Monday–Friday, Saturday–Sunday), and 24-hour energy intake (kcal; continuous; for all variables except % energy from fat, energy density of foods and beverages, and temporal variables). n = 34,504 participants with complete covariate information.
Trends were operationalized as the sex-specific median reported time in each quartile of the reported time of first eating episode.
n = 34,465.
n = 33,970.
Plain water was not considered a beverage.
An eating event where plain water was the only item was not considered an eating episode.
An increasing time of the first eating episode was related with a later time of the last ingestive episode, a shorter duration of the 24-hour ingestion period, fewer eating episodes, and shorter intervals between eating episodes (P < 0.0001; Table 5).
Cardiometabolic biomarker outcomes
BMI, fasting triglycerides, fasting glucose, fasting insulin, fasting C-peptide, and C-reactive protein values increased with increasing quartiles of time of the first eating episode (P ≤ 0.001; Table 6). Although these trends were highly significant, the magnitudes of observed differences in biomarker concentrations with an increasing time of the first eating episode were small. Serum total cholesterol was not associated with the time of first eating.
TABLE 6.
Covariate-adjusted mean and 95% CI of body mass index and serum concentrations of cardiometabolic biomarkers by approximate quartiles of reported time of first eating episode in a 24-hour dietary recall1
| Survey-weighted, sex-specific, approximate quartiles of time of first reported eating episode in the 24-hour recall | Trend2 across quartiles | ||||
|---|---|---|---|---|---|
| Biomarker | Quartile 1 | Quartile 2 | Quartile 3 | Quartile 4 | |
| BMI, kg/m2, n = 33,937 | 28.6 (28.3–28.9) | 28.3 (28.1–28.5) | 28.8 (28.6–29.0) | 29.1 (28.9–29.3) | 0.0002 |
| Serum cholesterol, mg/dL, n = 32,785 | 207 (205–208) | 207 (205–208) | 207 (206–209) | 207 (206–209) | 0.5 |
| HDL cholesterol, mg/dL, n = 32,704 | 54 (53–54) | 53 (53–54) | 53 (52–53) | 53 (52–54) | 0.04 |
| Fasting triglycerides,3, 4 mg/dL, n = 12,336 | 121 (118–124) | 125 (122–128) | 128 (125–132) | 131 (127–134) | <0.0001 |
| Glycated hemoglobin,3 %, n = 33,206 | 5.64 (5.6–5.6) | 5.64 (5.6–5.7) | 5.64 (5.6– 5.7) | 5.66 (5.6–5.7) | 0.2 |
| Fasting glucose,3, 4 mg/dL, n = 15,209 | 103 (102–105) | 104 (103–105) | 105 (104–106) | 106 (105–107) | 0.0001 |
| Fasting insulin,3, 4 μU/mL n = 15,055 | 8.8 (8.5–9.1) | 9.4 (9.1–9.7) | 9.9 (9.6–10.2) | 10.0 (9.7–10.3) | <0.0001 |
| Fasting C-peptide,3– 5 nmol/L, n = 7937 | 0.73 (0.7–0.75) | 0.75 (0.72–0.78) | 0.77 (0.74–0.80) | 0.77 (0.75–0.79) | 0.001 |
| C-reactive protein,3, 6 mg/dL, n = 26,763 | 0.21 (0.20–0.22) | 0.23 (0.22–0.24) | 0.23 (0.22–0.24) | 0.24 (0.23–0.25) | <0.0001 |
Estimates are predicted margins (adjusted means) from linear regression models, with each biomarker as a continuous dependent variable. Independent variables were time of first eating episode (quartiles), age (continuous), sex (men, women), race (non-Hispanic White, non-Hispanic Black, Mexican-American, Others), poverty income ratio (<130%, 130%–349%, ≥349%, missing), education (<12 years, 12 years, some college, ≥college), leisure time physical activity (yes, no), self-reported doctor-diagnosed chronic disease (yes, no), smoking status (never, former, current smoker), alcohol use status (never, former, current drinker, unknown), BMI (<25, 25–29.9, or ≥30 kg/m2 or missing; except models with BMI as dependent), and hours of fasting before phlebotomy (tertiles).
Trends were operationalized as the sex-specific median reported time in each quartile of reported time of first eating episode.
Back-transformed geometric means and 95% CIs.
Fasting subsample weights for serum triglycerides, glucose, insulin, and C-peptide.
Not available for surveys conducted after 2003–2004.
Not available for survey cycles conducted in 2011–2012 and 2013–2014.
Discussion
In this observational study, there was a small relative survival advantage for ≥40-year-old adults who reported their first eating episode of the day before ∼08:00. Study results also suggest that reporting an earlier time of the first ingestive episode was associated with higher-quality diets, and lower metabolic risks, indicated by an array of cardiometabolic biomarkers.
To our knowledge, this is among the first published reports to examine the relationships between the clock time of the first eating episode and risks of mortality and metabolic biomarkers; therefore, there are no directly comparable studies for corroborative assessments. However, there is a sizable body of literature on the relationships between chronotype—mostly based on individual preferences for sleep and wake times or self-identification as either a morning or evening type—and a variety of health outcomes (13–20) and all-cause mortality (21, 22). Having an evening chronotype was associated with a 30% increased age-adjusted risk of all-cause mortality in a small Finnish cohort of men (n ≈ 2144) (21). However, the multiple covariate-adjusted mortality risk was 10% higher at 6.5 years of follow-up among evening types relative to morning types in the large British Biobank cohort (22), which is comparable to the extent of risk reduction observed in the association with an early time of first eating in the current study.
Our study suggests that reporters of a later time of the first eating episode also reported adverse physical activity and dietary behaviors. In unadjusted analysis in Table 1, any leisure time physical activity was less frequently reported in association with a later time of first eating. The energy-adjusted intakes of key micronutrients (potassium and magnesium) and food groups (whole grains and fruits) and the overall diet quality (assessed as HEI-2015) were lower in these respondents, and the energy density of all reported foods and beverages, driven largely by a higher energy density of beverages, was higher. Although the relationships between our study exposure—time of the first eating episode—and dietary attributes have been examined in relatively few studies (37), our results of qualitative differences in diets by the time of the first eating episode are in accord with previous reports that compared morning and evening chronotypes (38–42).
The study findings of a higher BMI and higher serum concentrations of cardiometabolic biomarkers with a later time of the first eating episode suggest mechanisms through which the time of the first eating episode may relate to metabolic health. Prior studies have reported a higher prevalence of obesity and higher serum glucose and C-reactive protein in association with an evening chronotype or later time of eating (14–17). We found only 1 published study that examined the self-reported clock time of breakfast as an exposure; in this study, clock-time of breakfast mediated the relation of a morning-evening chronotype with the risk of obesity in type 2 diabetics (43). Dashti et al. (44) found that early eaters, based on the midpoint between the clock times of breakfast and lunch, had significantly lower fasting insulin and triglyceride concentrations (44). It is noteworthy that McHill et al. (45) found no cross-sectional association between the clock time of eating and body weight in a study where a later circadian time, established using melatonin onset, was related to higher body fat. However, it is also possible that dietary and physical activity behaviors associated with a later time of the first eating episode may independently contribute to adverse metabolic profiles.
We note that the mean duration of the 24-hour eating window was 5 hours longer in the first quartile of time of the first eating episode relative to the last quartile of time of the first eating episode. However, the average clock time of the last eating episode differed by <30 minutes between extreme quartiles of the time of the first eating episode. We have previously noted this relative lack of variation in the reported time of the evening eating event despite a gradual increase in the time of reported breakfast over the past 3 decades in the US population (27, 28).
We make no assertions about biologic or circadian times of eating or chronotypes in this study. The NHANES surveys included in our study did not collect information about usual sleep and wake times, which can be used to determine the midpoint of sleep and classification of early or late chronotypes, nor were subjects asked to self-classify themselves as morning or evening types. Moreover, the surveys did not include biomarkers of dim light melatonin onset to enable an assessment of chronotype, nor is this type of measurement a logistical possibility in a population survey such as the NHANES. Instead, we used the available clock time of the first eating episode as a proxy for an early or late preference. By design, the exposure in our study is not based on named eating events, such as breakfast, snack, and so forth. Names of eating events are social and cultural constructs (17), and exposure based on meal name(s) may exclude a substantial percentage of the population who skip meals such as breakfast or lunch (27, 28). The first eating episode of the day is preceded by an overnight fast, and the time of morning eating events has been shown to differ significantly between early and late chronotypes (14, 43, 46). Xiao et al. (14) examined chronotypes based on midpoints of sleep, dietary intake, and body weight over 1 year and found that the mean clock time of breakfast differed between early and late chronotypes by an hour. However, the magnitudes of differences in mean clock times of lunch and dinner between early and late chronotypes, although significant, were much smaller (0.2 and 0.3 hours, respectively). Similarly, Dashti et al. (44) found that the mean breakfast time differed by >1 hour between early and late eaters, but the mean lunch time differed by only 7 minutes and the dinner time differed by 37 minutes.
The strengths of our study include a large, nationally representative cohort of US adults aged ≥40 years with long-term follow-up data. The availability of detailed dietary intake, measured BMI, metabolic biomarkers, and information on a variety of potential confounders, collected using standardized protocols by the trained NHANES staff, is another strength. However, study results should be interpreted with due consideration for the following limitations. First, the study is observational in nature and provides no information on causality. Second, the dietary exposure of reported clock time of eating was determined from a single recall of the previous 24 hours, which is prone to both systematic and random errors. Although dietary recall methods in the NHANES have been validated (47–49), the focus of such validations was on the intakes of energy and nutrients, not the time of eating. Moreover, although the estimation of “usual dietary intakes” using repeat dietary recalls has received considerable attention (50), such methods have not been studied for the principal exposure of our study: the reported clock time of ingestion of foods and beverages. In a small sample of 18- to 21-year-old college students (n = 14), McHill et al. (51) reported considerable intraindividual variability in the clock time of eating events recorded in diet records. Interestingly, the within-individual variability in the clock time of the first eating event was smaller than the variability in the clock time of other eating events. Given that our study population was ≥40 years old, we can speculate that the variability of the time of eating may be smaller. Nevertheless, random (intraindividual variability) and systematic measurement errors in determination of the time of eating are undoubtedly present, leading to misclassification into categories of the time of first eating and possibly attenuating or increasing associations with an outcome.
Another limitation is that the testing for proportional hazards indicated that the effect of the time of the first eating episode was essentially confined to the earliest follow-up interval. This could result from reverse causality from individuals with reported chronic conditions or latent illness at baseline, who more likely experienced mortality in the first compared with the later follow-up intervals and may have had later times of first eating episodes. This was examined in sensitivity analyses by stratifying the analyses by individuals who did and did not report chronic conditions, where analyses in both strata showed similar trends in mortality. We also excluded the first 2 years of follow-up to examine the possible effects of latent illness, which showed a statistically significant but numerically diminished trend in the hazards. The lack of proportional hazards could also be due in part to changes in the time of the first eating episode over the follow-up period, but only a single baseline dietary assessment was collected.
We also note that the days of recall in our study included both weekdays and weekend days. Notably, however, the observed associations were not changed by the use of weekday-/weekend-specific quartiles of the time of the first eating episode (Table 4). Nevertheless, it is likely that the time of the first eating episode, the temporal exposure in this study, resulted in some misclassification. Finally, although we adjusted for several potential established confounders of the association between the time of eating and mortality, residual confounding due to unknown or unmeasured variables cannot be excluded.
We limited the study analytic cohort to participants aged 40 years and older. We hypothesized that by age 40, routines of work and leisure may be stable and, in turn, eating patterns, including the time of eating, may be well established. This hypothesis is supported by findings of decreased variability in chronotypes among older US men and women in a large US survey (52) and the suggestion of the persistence of chronotype with age in European men (21). However, as we acknowledge above, possible effects of random measurement errors and within-person variability in various aspects of temporality of eating behaviors are likely present and require further study. Another related area of study would be the type of memory—that is, episode-specific or generic—involved in recall of the time of eating.
In conclusion, the self-report of an earlier clock time of the first eating episode in a 24-hour recall by ≥40-year old Americans was associated with an approximately 12% lower risk of all-cause mortality. An early time of the first eating episode was also a correlate of qualitative dietary attributes and a better cardiometabolic profile. The observed associations, although statistically significant, were weak and should be considered exploratory. Given the observational nature of our study, the findings need to be replicated in other large, well-established cohorts.
Supplementary Material
Acknowledgments
We thank Lisa Kahle, Information Management Systems, Silver Spring, MD, for expert SAS and SUDAAN programming support.
The authors’ responsibilities were as follows—AKK: conceptualized the study question, designed the research, analyzed the data, wrote the manuscript, and had primary responsibility for the final content of the manuscript; BIG: provided guidance on the study design and analytic strategy, drafted the manuscript, and reviewed the manuscript for important intellectual content; and both authors: read and approved the final manuscript.
Notes
Supported in part by the intramural research program of the Department of Health and Human Services, National Cancer Institute, NIH.
Author disclosures: The authors report no conflicts of interest.
Supplemental Figure 1 is available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.
Abbreviations used: FPED, Food Patterns Equivalents Database; HEI-2015, Healthy Eating Index–2015; MEC, mobile examination center; MPED, MyPyramid Equivalents Database; NCHS, National Center for Health Statistics.
Contributor Information
Ashima K Kant, Department of Family, Nutrition, and Exercise Sciences, Queens College of the City University of New York, Flushing, NY, USA.
Barry I Graubard, Division of Cancer Epidemiology and Genetics, Biostatistics Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA.
Data Availability
This manuscript is based on public domain data from the National Health and Nutrition Examination Surveys (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx).
References
- 1. Asher G, Sassone-Corsi P. Time for food: the intimate interplay between nutrition, metabolism, and the circadian clock. Cell. 2015;161(1):84–92. [DOI] [PubMed] [Google Scholar]
- 2. Mattson MP, Allison DB, Fontana L, Harvie M, Longo VD, Malaisse WJ, Mosley M, Notterpek L, Ravussin E, Scheer FA et al. Meal frequency and timing in health and disease. Proc Natl Acad Sci. 2014;111(47):16647–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Johnston JD. Physiological links between circadian rhythms, metabolism and nutrition. Exp Physiol. 2014;99(9):1133–7. [DOI] [PubMed] [Google Scholar]
- 4. Challet E. The circadian regulation of food intake. Nat Rev Endocrinol. 2019;15(7):393–405. [DOI] [PubMed] [Google Scholar]
- 5. Wehrens SMT, Christou S, Isherwood C, Middleton B, Gibbs MA, Archer SN, Skene DJ, Johnston JD. Meal timing regulates the human circadian system. Curr Biol. 2017;27(12):1768–75.e3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Poggiogalle E, Jamshed H, Peterson CM. Circadian regulation of glucose, lipid, and energy metabolism in humans. Metabolism. 2018;84:11–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Richter J, Herzog N, Janka S, Baumann T, Kistenmacher A, Oltmanns KM. Twice as high diet-induced thermogenesis after breakfast vs dinner on high-calorie as well as low-calorie meals. J Clin Endocrinol Metab. 2020;105(3):e211. [DOI] [PubMed] [Google Scholar]
- 8. Morgan LM, Shi JW, Hampton SM, Frost G. Effect of meal timing and glycaemic index on glucose control and insulin secretion in healthy volunteers. Br J Nutr. 2012;108(7):1286–91. [DOI] [PubMed] [Google Scholar]
- 9. Jakubowicz D, Barnea M, Wainstein J, Froy O. High caloric intake at breakfast vs. dinner differentially influences weight loss of overweight and obese women. Obesity. 2013;21(12):2504–12. [DOI] [PubMed] [Google Scholar]
- 10. Leung GKW, Huggins CE, Ware RS, Bonham MP. Time of day difference in postprandial glucose and insulin responses: systematic review and meta-analysis of acute postprandial studies. Chronobiol Int. 2020;37(3):311–26. [DOI] [PubMed] [Google Scholar]
- 11. Aoyama S, Shibata S. Time-of-day-dependent physiological responses to meal and exercise. Front Nutr. 2020;7:18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. St-Onge MP, Ard J, Baskin ML, Chiuve SE, Johnson HM, Kris-Etherton P, Varady K; American Heart Association Obesity Committee of the Council on Lifestyle and Cardiometabolic Health; Council on Cardiovascular Disease in the Young; Council on Clinical Cardiology; and Stroke Council . Meal timing and frequency: implications for cardiovascular disease prevention: a scientific statement from the American Heart Association. Circulation. 2017;135(9):e96–e121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Partonen T. Chronotype and health outcomes. Curr Sleep Med Rep. 2015;1(4):205–11. [Google Scholar]
- 14. Xiao Q, Garaulet M, Scheer F. Meal timing and obesity: interactions with macronutrient intake and chronotype. Int J Obes. 2019;43(9):1701–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Maukonen M, Kanerva N, Partonen T, Männistö S. Chronotype and energy intake timing in relation to changes in anthropometrics: a 7-year follow-up study in adults. Chronobiol Int. 2019;36(1):27–41. [DOI] [PubMed] [Google Scholar]
- 16. Almoosawi S, Prynne CJ, Hardy R, Stephen AM. Time-of-day and nutrient composition of eating occasions: prospective association with the metabolic syndrome in the 1946 British birth cohort. Int J Obes. 2013;37(5):725–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Almoosawi S, Vingeliene S, Gachon F, Voortman T, Palla L, Johnston JD, Van Dam RM, Darimont C, Karagounis LG. Chronotype: Implications for epidemiologic studies on chrono-nutrition and cardiometabolic health. Adv Nutr. 2019;10(1):30–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Leech RM, Timperio A, Worsley A, McNaughton SA. Eating patterns of Australian adults: associations with blood pressure and hypertension prevalence. Eur J Nutr. 2019;58(5):1899–909. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Kogevinas M, Espinosa A, Castelló A, Gómez-Acebo I, Guevara M, Martin V, Amiano P, Alguacil J, Peiro R, Moreno V et al. Effect of mistimed eating patterns on breast and prostate cancer risk (MCC-Spain study). Int J Cancer. 2018;143(10):2380–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Hurley S, Goldberg D, Von Behren J, Clague DeHart J, Wang S, Reynolds P. Chronotype and postmenopausal breast cancer risk among women in the California Teachers Study. Chronobiol Int. 2019;36(11):1504–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Broms U, Pitkäniemi J, Bäckmand H, Heikkilä K, Koskenvuo M, Peltonen M, Sarna S, Vartiainen E, Kaprio J, Partonen T. Long-term consistency of diurnal-type preferences among men. Chronobiol Int. 2014;31(2):182–8. [DOI] [PubMed] [Google Scholar]
- 22. Knutson KL, von Schantz M. Associations between chronotype, morbidity and mortality in the UK Biobank cohort. Chronobiol Int. 2018;35(8):1045–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. CDC . National Health and Nutrition Examination Survey. [Internet]. Available from: https://www.cdc.gov/nchs/nhanes/about_nhanes.htm. [Google Scholar]
- 24. CDC . National Center for Health Statistics (NCHS). National Health and Nutrition Examination Survey data. Hyattsville (MD): US Department of Health and Human Services, Centers for Disease Control and Prevention, NHANES III and NHANES 1999–2014. [Internet]. Available from: https://wwwn.cdc.gov/nchs/nhanes/default.aspx. [Google Scholar]
- 25. CDC . National Health and Nutrition Examination Survey. NHANES response rates and population totals. NCHS. [Internet]. Available from: https://wwwn.cdc.gov/nchs/nhanes/ResponseRates.aspx. [Google Scholar]
- 26. CDC . National Health and Nutrition Examination Survey. NHANES survey methods and analytic guidelines. NCHS. [Internet]. Available from: https://wwwn.cdc.gov/nchs/nhanes/analyticguidelines.aspx. [Google Scholar]
- 27. Kant AK. Eating patterns of US adults: meals, snacks, and time of eating. Physiol Behav. 2018;193(Pt B):270–8. [DOI] [PubMed] [Google Scholar]
- 28. Kant AK, Graubard BI. 40-year trends in meal and snack eating behaviors of American adults. J Acad Nutr Diet. 2015;115(1):50–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Kant AK, Graubard BI. Secular trends in patterns of self-reported food consumption of adult Americans: NHANES 1971–1975 to NHANES 1999–2002. Am J Clin Nutr. 2006;84(5):1215–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. National Center for Health Statistics . Office of Analysis and Epidemiology, public-use linked mortality file, 2015. Hyattsville (MD): CDC. [Internet]. Available from: https://www.cdc.gov/nchs/data-linkage/mortality-public.htm. [Google Scholar]
- 31. National Center for Health Statistics . Office of Analysis and Epidemiology. The linkage of National Center for Health Statistics survey data to the National Death Index—2015 linked mortality file (LMF): methodology overview and analytic considerations, March 2019. Hyattsville (MD): CDC; 2019. [Internet]. Available from: https://www.cdc.gov/nchs/data-linkage/mortality-methods.htm. [Google Scholar]
- 32. USDA . Food patterns equivalents database. Baltimore (MD): Agricultural Research Service. Food Surveys Research Group. [Internet]. Available from: https://www.ars.usda.gov/northeast-area/beltsville-md-bhnrc/beltsville-human-nutrition-research-center/food-surveys-research-group/docs/fped-methodology/. [Google Scholar]
- 33. Krebs-Smith SM, Pannucci TE, Subar AF, Kirkpatrick SI, Lerman JL, Tooze JA, Wilson MM, Reedy J. Update of the Healthy Eating Index-2015. J Acad Nutr Diet. 2018;118(9):1591–602. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Kant AK, Graubard BI. Energy density of diets reported by American adults: association with food group intake, nutrient intake, and body weight. Int J Obes. 2005;29(8):950–6. [DOI] [PubMed] [Google Scholar]
- 35. National Health and Nutrition Examination Surveys. NHANES laboratory data. CDC. [Internet]. Available from: https://wwwn.cdc.gov/nchs/nhanes/search/datapage.aspx?Component=Laboratory. [Google Scholar]
- 36. Graubard BI, Korn E. Predictive margins with survey data. Biometrics. 1999;55(2):652–9. [DOI] [PubMed] [Google Scholar]
- 37. Gontijo CA, Cabral BBM, Balieiro LCT, Teixeira GP, Fahmy WM, Maia YCP, Crispim CA. Time-related eating patterns and chronotype are associated with diet quality in pregnant women. Chronobiol Int. 2019;36(1):75–84. [DOI] [PubMed] [Google Scholar]
- 38. Kanerva N, Kronholm E, Partonen T, Ovaskainen ML, Kaartinen NE, Konttinen H, Broms U, Männistö S. Tendency toward eveningness is associated with unhealthy dietary habits. Chronobiol Int. 2012;29(7):920–7. [DOI] [PubMed] [Google Scholar]
- 39. Maukonen M, Kanerva N, Partonen T, Kronholm E, Konttinen H, Wennman H, Männistö S. The associations between chronotype, a healthy diet and obesity. Chronobiol Int. 2016;33(8):972–81. [DOI] [PubMed] [Google Scholar]
- 40. Maukonen M, Kanerva N, Partonen T, Kronholm E, Tapanainen H, Kontto J, Männistö S. Chronotype differences in timing of energy and macronutrient intakes: a population-based study in adults. Obesity. 2017;25(3):608–15. [DOI] [PubMed] [Google Scholar]
- 41. Mazri FH, Manaf ZA, Shahar S, Mat Ludin AF. The association between chronotype and dietary pattern among adults: a scoping review. Int J Environ Res Public Health. 2019;17(1):68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Dashti HS, Chen A, Daghlas I, Saxena R. Morning diurnal preference and food intake: a Mendelian randomization study. Am J Clin Nutr. 2020;112(5):1348–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Nimitphong H, Siwasaranond N, Saetung S, Thakkinstian A, Ongphiphadhanakul B, Reutrakul S. The relationship among breakfast time, morningness-eveningness preference and body mass index in type 2 diabetes. Diabet Med. 2018;35(7):964–71. [DOI] [PubMed] [Google Scholar]
- 44. Dashti HS, Gómez-Abellán P, Qian J, Esteban A, Morales E, Scheer F, Garaulet M. Late eating is associated with cardiometabolic risk traits, obesogenic behaviors, and impaired weight loss. Am J Clin Nutr. 2021;113(1):154–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. McHill AW, Phillips AJ, Czeisler CA, Keating L, Yee K, Barger LK, Garaulet M, Scheer FA, Klerman EB. Later circadian timing of food intake is associated with increased body fat. Am J Clin Nutr. 2017;106(5):1213–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Phoi YY, Rogers M, Bonham MP, Dorrian J, Coates AM. A scoping review of chronotype and temporal patterns of eating of adults: tools used, findings, and future directions. Nutr Res Rev. 2021;1–24.. doi:10.1017/S0954422421000123. [DOI] [PubMed] [Google Scholar]
- 47. Conway JM, Ingwersen LA, Moshfegh AJ. Accuracy of dietary recall using the USDA five-step multiple-pass method in men: an observational validation study. J Am Diet Assoc. 2004;104(4):595–603. [DOI] [PubMed] [Google Scholar]
- 48. Conway JM, Ingwersen LA, Vinyard BT, Moshfegh AJ. Effectiveness of the US Department of Agriculture 5-step multiple-pass method in assessing food intake in obese and nonobese women. Am J Clin Nutr. 2003;77(5):1171–8. [DOI] [PubMed] [Google Scholar]
- 49. Moshfegh AJ, Rhodes DG, Baer DJ, Murayi T, Clemens JC, Rumpler WV, Paul DR, Sebastian RS, Kuczynski KJ, Ingwersen LA et al. The US Department of Agriculture automated multiple-pass method reduces bias in the collection of energy intakes. Am J Clin Nutr. 2008;88(2):324–32. [DOI] [PubMed] [Google Scholar]
- 50. National Cancer Institute . Division of Cancer Control and Population Sciences. Epidemiology and genomics research program. Usual dietary intakes. [Internet]. Available from: https://epi.grants.cancer.gov/diet/usualintakes/. [Google Scholar]
- 51. McHill AW, Hilditch CJ, Fischer D, Czeisler CA, Garaulet M, Scheer F, Klerman EB. Stability of the timing of food intake at daily and monthly timescales in young adults. Sci Rep. 2020;10(1):20849. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Fischer D, Lombardi DA, Marucci-Wellman H, Roenneberg T. Chronotypes in the US–Influence of age and sex. PLoS One. 2017;12(6):e0178782. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
This manuscript is based on public domain data from the National Health and Nutrition Examination Surveys (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx).