Dietary circadian rhythms and cardiovascular disease risk in the prospective NutriNet-Santé cohort

Cardiovascular diseases are the leading global cause of mortality. Dietary behaviors, including the timing of food intake, can synchronize peripheral circadian clocks that regulate cardiometabolic functions such as blood pressure. Modern lifestyles and the rise of fasting practices have increased mistimed eating behaviors like late-night eating and breakfast skipping, which have been linked to adverse cardiometabolic outcomes. Prior studies and meta-analyses suggest breakfast consumption benefits cardiometabolic health, while late-night eating is associated with risk factors. However, there is no consensus on meal definitions, and prior research often relies on self-identified meal labels and broad time windows, risking misclassification. Few studies have examined meal timing as a continuous exposure, and none have evaluated its association with incident CVD using hard endpoints. This study aims to investigate whether the timing of the first and last daily meals, the number of eating occasions, and nighttime fasting duration are associated with the incidence of overall CVD, cerebrovascular disease, and coronary heart disease in the NutriNet-Santé cohort.

Observational and interventional evidence indicates breakfast skipping is associated with overweight/obesity, higher CVD risk, and type 2 diabetes, though some trials show no effect on body weight. Late-night eating has been linked to arterial stiffness, dyslipidemia, obesity, metabolic syndrome in women, and increased coronary heart disease risk in one cohort. A key limitation is inconsistent meal definitions leading to classification bias. Only two cross-sectional studies have assessed continuous meal timing with cardiometabolic risk factors, not with incident CVD. Time-restricted eating (TRE), extending nighttime fasting beyond 12 hours, improves several cardiometabolic markers in animal models and some human studies, though results are mixed and hard endpoints are lacking. Evidence gaps include the direct link between nightly fasting duration and CVD risk, and the role of eating frequency in the context of meal timing and fasting.

Design and population: Prospective analysis within the French web-based NutriNet-Santé cohort (launched 2009), including adults aged ≥18 years who provided informed consent. Ethics approvals obtained (IRB Inserm and CNIL). As of Oct 5, 2021, 103,389 participants without prevalent CVD at baseline were included. Exposure assessment: Diet was assessed from baseline using biannual series of three 24-hour dietary records (including at least one non-working day). Participants reported all foods/beverages and their consumption times; portion sizes estimated via validated photographs or containers. Nutrient intakes derived from the NutriNet-Santé food composition database; under-reporters identified using Goldberg cut-offs. Circadian dietary behaviors: For the first two years of follow-up, averaged per-participant: time of first meal, time of last meal, and number of eating occasions (any intake ≥1 kcal; water excluded). Participants with first meal after 3PM or last meal before 3PM were excluded to remove extremely disrupted patterns (e.g., likely night-shift workers). Nighttime fasting duration was calculated as 24 h minus the elapsed time between first and last eating occasions. Outcomes: Incident CVDs were self-reported biannually and via a continuous online platform, verified by medical records reviewed by physicians, and complemented through linkage with national health insurance databases (SNIIRAM) and the national cause-specific mortality registry (CépiDC). Events were coded using ICD-10. Outcomes included: overall CVD; cerebrovascular diseases (stroke, transient ischemic attack); coronary heart diseases (myocardial infarction, angina pectoris, acute coronary syndrome, angioplasty). Statistical analysis: Cause-specific Cox proportional hazards models estimated hazard ratios (HR) and 95% CIs for associations of meal timing and eating frequency with overall CVD, cerebrovascular disease, and coronary heart disease. Exposures were modeled as continuous (per hour or per occasion) and categorical (approximate tertiles). Competing risks were handled by censoring at alternative first CVD events. Proportional hazards (Schoenfeld residuals) and linearity (splines) assumptions were checked. Primary models mutually adjusted the circadian nutritional behaviors (first meal, last meal, number of eating occasions) and included covariates: age (time scale), sex, education, income, baseline BMI (continuous), family history of CVD, alcohol consumption (categorized), binge drinking episodes, energy intake excluding alcohol (continuous), smoking status, pack-years (continuous), physical activity (IPAQ categories), number of dietary records (continuous), and healthy/Western dietary patterns (PCA-derived continuous scores). Multiple imputation by chained equations handled missing data (20 datasets, Rubin’s rules). Nighttime fasting duration was analyzed continuously and categorically (≤12 h, 12–13 h, >13 h) with similar covariate adjustment including time of first meal but excluding time of last meal to avoid collinearity. Effect modification was explored for sex, baseline BMI, menopausal status (women), and chronotype (subsample with sleep data). Numerous sensitivity analyses tested robustness: additional dietary adjustments (e.g., saturated fat, sodium, sugars, red/processed meat, sugary drinks, fruits/vegetables, nuts, whole grains, yogurt, ultra-processed foods), region and profession, weight change during follow-up, eating jet lag, exclusion of early cases, adjustment for baseline comorbidities, sleep-related variables (duration, quality, chronotype), interval between last meal and bedtime, number of nighttime awakenings, sleep apnea, inclusion of extreme meal timings, marital status/children, medication counts, seasonal adjustments, unusual dietary days, restrictive diets, and exclusions of prevalent obesity, type 2 diabetes, or sleep apnea. A falsification endpoint (basal cell carcinoma) was also assessed.

• Cohort: 103,389 participants (79% women), mean baseline age 42.6 (SD 14.5) years; median follow-up 7.2 years (Q1–Q3: 3.1–10.1), 699,547 person-years; 2,036 incident CVD cases (988 cerebrovascular; 1,071 coronary heart disease).
• Overall CVD: Each 1-hour delay in time of first meal associated with higher risk (HR 1.06, 95% CI 1.01–1.12; p=0.02). Time of last meal (continuous) not significant; last meal after 9PM vs before 8PM: HR 1.13 (0.99–1.29), p-trend=0.06. Number of eating occasions: null (HR 0.99, 0.96–1.02; p=0.5).
• Cerebrovascular disease: Time of first meal not associated overall. Each 1-hour delay in last meal associated with higher risk (HR 1.08, 1.01–1.15; p=0.02). Last meal after 9PM vs before 8PM: HR 1.28 (1.05–1.55), p-trend<0.01. Number of eating occasions: null (HR 0.97, 0.93–1.01; p=0.1).
• Coronary heart disease: No overall associations with meal timing or eating frequency in primary analyses.
• Nighttime fasting duration: Each additional hour associated with lower cerebrovascular disease risk (HR 0.93, 0.87–0.99; p=0.02). No associations with overall CVD (HR 0.98, 0.94–1.02; p=0.4) or coronary heart disease (HR 1.03, 0.96–1.09; p=0.4). Categorically for cerebrovascular disease vs ≤12 h: 12–13 h HR 0.78 (0.66–0.91); >13 h HR 0.80 (0.63–1.01); p=0.02.
• Sex differences: Significant interaction between sex and time of last meal for overall CVD (p=0.01) and coronary heart disease (p=0.004). Associations generally stronger in women for overall CVD and cerebrovascular disease with later first/last meals. Notably, in women: first meal after 9AM vs before 8AM for overall CVD HR 1.24 (1.02–1.51); last meal after 9PM vs before 8PM for overall CVD HR 1.26 (1.05–1.51) and for cerebrovascular disease HR 1.31 (1.03–1.67). In men: time of first meal and coronary heart disease showed a positive association (per 1 h delay HR 1.11, 1.01–1.22; p=0.03), whereas later last meals were inversely associated with CHD (per 1 h delay HR 0.91, 0.84–0.99; p=0.02). Nighttime fasting and CHD showed interaction by sex (p=0.02): women HR 0.93 (0.85–1.03), men HR 1.10 (1.01–1.19) per 1 h increase.
• Sensitivity analyses: Results robust to extensive adjustments and exclusions; some attenuation in sleep-questionnaire subsample (likely reduced power). Falsification endpoint (basal cell carcinoma) showed no associations. Longer interval between last meal and bedtime was inversely associated with overall CVD in ancillary analyses.

Findings indicate that later timing of the first and last daily meals is associated with higher risk of overall CVD, with particularly strong signals in women, and that longer nighttime fasting is associated with lower cerebrovascular disease risk. These results support the hypothesis that aligning food intake with endogenous circadian rhythms benefits cardiovascular health. Potential mechanisms include time-of-day differences in insulin sensitivity and glucose tolerance, and circadian regulation of blood pressure and metabolism, whereby late eating may induce circadian misalignment and adverse metabolic effects. The associations were not explained by weight change, suggesting additional pathways. While sleep patterns, chronotype, and potential night-shift work could confound associations, extensive adjustments and sensitivity analyses, including chronotype and sleep measures in a subsample, yielded generally consistent results. The findings, together with emerging interventional evidence, suggest that adopting earlier eating windows—potentially early time-restricted eating—may be more favorable than breakfast skipping for cardiovascular prevention. However, sex-specific differences and mixed findings for coronary outcomes in men highlight the need for further mechanistic and longitudinal research.

In this large prospective cohort, later times of first and last meals were associated with higher overall CVD risk, and longer nighttime fasting was linked to lower cerebrovascular disease risk. Results were generally stronger in women. These findings support considering meal timing—beyond diet quality—in cardiovascular prevention strategies and suggest potential benefits of earlier eating windows and longer overnight fasting coupled with an early last meal. Future research should replicate these findings in diverse populations, incorporate objective measures of sleep and activity timing, better capture shift work and light-at-night exposure, explore mechanistic pathways, and test meal timing interventions, including early time-restricted eating, on CVD hard endpoints.

Key limitations include the observational design with potential residual confounding; volunteer cohort with overrepresentation of women and healthier profiles, limiting generalizability; limited and self-reported sleep/chronotype data available only in a subsample; lack of detailed shift work history and data on light-at-night, recreational drug use, and timing of physical activity, medications, or alcohol; potential measurement error in self-reported behaviors; averaging meal timing over the first two years which may miss within-person variability (though eating jet lag was assessed); and possible reverse causation despite sensitivity exclusions. Some analyses in subsamples suffered reduced power. Despite comprehensive adjustment and a falsification endpoint, unmeasured confounding cannot be fully excluded.