Night eating in timing, frequency, and food quality and risks of all-cause, cancer, and diabetes mortality: findings from national health and nutrition examination survey

Meal timing is an important determinant of health. Night eating—consuming food between late evening and early night—has been linked to adverse outcomes such as cancer and diabetes, but its long-term mortality implications are unclear. Prior research suggests that the timing, frequency, and quality of food intake may modify health effects. However, no population-based study has comprehensively examined night eating across specific timing windows, frequency, and food quality in relation to all-cause, cancer, and diabetes mortality. This study investigates these associations in a nationally representative US cohort to inform public health strategies aimed at reducing excess mortality risk associated with night eating.

Existing evidence indicates later meal timing is associated with higher adiposity and risk of type 2 diabetes. Higher daily eating frequency has been associated with lower mortality in some populations, though results may differ at night. A higher percentage of energy intake at night associates with increased risks of T2DM and cancer, whereas small, low-energy nighttime snacks may be neutral or beneficial for cardiometabolic health. Dietary energy density is positively related to risks of T2DM and several cancers. These mixed findings suggest that timing, frequency, and food quality could differentially influence health impacts of night eating, motivating a nuanced analysis.

Design and data source: Observational analysis using NHANES 2002–2018 with linkage to the National Death Index through December 31, 2019, following STROBE-NUT guidelines. Population: Adults ≥20 years with at least one valid 24-h dietary recall at baseline. Exclusions: unreliable dietary recalls, age <20, implausible energy intake (>5000 or <500 kcal/day), missing mortality data, incomplete dietary intake data, and pregnancy at baseline. Final sample: 41,744 participants. Dietary assessment: First 24-h dietary recall conducted in person at mobile exam centers using standardized methods; nutrient values from FNDDS; foods mapped to FPED 2017–2018 groups; HEI-2015 calculated for diet quality; dietary intake adjusted for total energy by residual method; supplement use recorded separately. Exposures: Night eating defined as intake between 22:00 and 4:00. Timing categorized as: no eating, 22:00–23:00, 23:00–00:00, 00:00–1:00, 1:00–2:00, 2:00–3:00, 3:00–4:00 (latest time used if >1 event). Frequency categorized as no eating, one time, or two times and over. Food quality: Using energy intake, low-energy-density foods (fruits, vegetables, whole grains, dairy, protein foods), and high-energy-density foods (refined grains, added sugars, oils, solid fats, alcohol). Each parameter dichotomized (energy <200 vs ≥200 kcal; food intakes below median vs at/above median). Latent class analysis (model selection via AIC/BIC) yielded four classes: VL-energy (very low energy and very low high-density foods), L-energy (low energy, very low low-density foods, moderate high-density foods), M-energy (moderate energy with moderate low- and high-density foods), H-energy (high energy with high low- and high-density foods). Outcomes: All-cause mortality; cancer mortality (ICD-10 C00–C9); diabetes mortality (ICD-10 E10–E14) from NDI Public-use Linked Mortality Files. Covariates: Age, sex, race/ethnicity, education, income, smoking, drinking, BMI, physical activity (MET-h/week), sleep hours, dietary energy intake, HEI-2015, supplement use, glycohemoglobin, triglycerides, fasting glucose, total cholesterol (TCHO), OGTT, hypertension, hyperlipidemia, CVD, diabetes, cancer. Missing data handled via multiple imputation by chained equations (most covariates <10% missing; sleep hours ~20%). Statistical analysis: NHANES survey weights, strata, and clusters applied. Weighted Cox proportional hazards models estimated aHRs with 95% CIs. Models: (1) age, sex; (2) + education, race/ethnicity, income, BMI; (3) + energy intake, alcohol, smoking, physical activity, histories (diabetes, hypertension, CVD, cancer, hyperlipidemia), HEI-2015, supplement use. Subgroup analyses by age (<65/≥65), sex, BMI categories, smoking, drinking, HEI-2015 (<70/≥70), sleep hours (<6/≥6). Linear regression assessed associations of night eating metrics with HbA1c, triglycerides, fasting glucose, TCHO, OGTT (with model 3 additionally adjusting for fasting time). Sensitivity analyses: (a) excluded participants with >50% daily energy at night; (b) included 21:00–22:00 time category; (c) additional adjustment for sleep time.

- Sample and outcomes: 41,744 participants; median follow-up 8.7 years; 6066 deaths (1381 cancer; 206 diabetes). - Timing of night eating vs no night eating: All-cause mortality showed a positive trend (P-trend=0.002 in fully adjusted model). Notable aHRs (Model 3): 23:00–00:00 aHR 1.29 (95% CI 1.08–1.54); 00:00–1:00 aHR 1.38 (95% CI 1.02–1.88); 1:00–2:00 aHR 1.49 (95% CI 0.97–2.30, trend highest but not statistically significant); 3:00–4:00 aHR 1.22 (95% CI 0.80–1.86); 2:00–3:00 aHR 1.12 (95% CI 0.73–1.71); 22:00–23:00 aHR 1.04 (95% CI 0.94–1.16). Diabetes mortality increased for earlier night windows: 22:00–23:00 aHR 1.68 (95% CI 1.07–2.62); 23:00–00:00 aHR 2.31 (95% CI 1.21–4.40). Cancer mortality overall not associated by trend, but eating at 1:00–2:00 showed higher risk: aHR 2.09 (95% CI 1.03–4.25). - Frequency of night eating: Compared to none, one time associated with higher all-cause mortality aHR 1.10 (95% CI 1.01–1.20) and diabetes mortality aHR 1.72 (95% CI 1.20–2.48). Two times or over associated with higher all-cause mortality aHR 1.38 (95% CI 1.09–1.75); cancer mortality not significantly associated; diabetes mortality for two or more not reported as significant. - Food quality (latent classes) vs no night eating: High dietary-energy-density (H-energy) class associated with higher risks: all-cause aHR 1.21 (95% CI 1.06–1.38); cancer aHR 1.30 (95% CI 0.99–1.71, borderline); diabetes aHR 1.97 (95% CI 1.13–3.45). Moderate (M-energy) class: all-cause aHR 1.28 (95% CI 1.03–1.58); trends toward cancer aHR 1.22 (95% CI 0.87–1.73) and diabetes aHR 2.01 (95% CI 0.69–5.83). Very low (VL-energy) and low (L-energy) classes showed no associations. - Biomarkers: Night eating correlated positively with glycohemoglobin for timing at 23:00–00:00, 00:00–1:00, 1:00–2:00; with frequency of one or ≥2 times; and with L-, M-, and H-energy classes. Fasting glucose and OGTT were positively associated with timing at 23:00–00:00 and 00:00–1:00. Triglycerides showed inverse associations with timing at 23:00–00:00, frequency of ≥2 times, and H-energy class. Total cholesterol was not associated. - Subgroups and sensitivity: Associations were consistent across subgroups of age, sex, BMI, smoking, drinking, HEI-2015, and sleep duration (no significant interactions). Sensitivity analyses (excluding >50% nocturnal energy, adding 21:00–22:00 window, adjusting for sleep) yielded similar results.

The study demonstrates that night eating is associated with increased risks of all-cause and diabetes mortality, with risk varying by specific night-time windows, frequency, and food quality. Eating between 23:00 and 1:00 was particularly associated with higher all-cause mortality, and eating between 22:00 and 00:00 with higher diabetes mortality, while cancer mortality showed limited timing-specific association (notably at 1:00–2:00). More frequent night eating elevated all-cause and diabetes mortality risks. High dietary energy density at night conferred greater risks, whereas low energy intake at night was not associated with excess mortality. These findings suggest that circadian misalignment, higher nocturnal energy intake, and altered glucose metabolism (as reflected by higher HbA1c, fasting glucose, OGTT) may underlie the observed associations. The results provide actionable insights: if night eating occurs, earlier timing (before 23:00) and low energy-density choices may mitigate risk.

Night eating is associated with increased all-cause, cancer, and diabetes mortality, with risk patterns depending on timing, frequency, and food quality. Eating before 23:00 and choosing low dietary energy-density foods may reduce excess mortality risk among night eaters. These results inform chrono-nutrition strategies and public health guidance on nocturnal eating behaviors.

- Night eating and diet were assessed using a single baseline 24-h dietary recall, which may not capture long-term habits. - Only the first 24-h recall was used to obtain specific timing, potentially limiting precision. - Residual and unmeasured confounding cannot be excluded despite extensive adjustments. - Lack of detailed sleep timing and occupational (e.g., shift work) data, which may influence associations; only sleep duration was adjusted. - Limited diabetes mortality events in specific timing categories reduced statistical power and may attenuate some estimates.