https://www.nature.com/articles/s41430-026-01796-1
chatGPT(6AstraMaxPaid):
This study links meal timing with mortality, but it does not establish that changing meal times extends life or that time-restricted eating is harmful. Its most useful finding is that eating-window duration needs to be interpreted alongside when that window occurs.
The paper is Meal timing and eating window with all-cause and cardiovascular mortality and life expectancy: population-based cohort study, by Hu and colleagues, published in the European Journal of Clinical Nutrition in 2026.
Summary. The researchers analysed 31,044 US adults aged 40 or older from NHANES surveys conducted between 1999 and 2018. During a median follow-up of 8.6 years, there were 7,129 deaths, including 2,259 cardiovascular deaths.
Meal timing was primarily determined from one 24-hour dietary recall at baseline. A “meal” meant any food or drink containing calories, including snacks. The eating window was the interval between the first and last caloric intake.
The analyses adjusted for demographic factors, income, smoking, alcohol, physical activity, calorie intake, diet quality, BMI and existing chronic disease.
The principal results were:
| Eating pattern | All-cause mortality HR, with 95% CI | Cardiovascular mortality HR, with 95% CI |
|---|---|---|
| First intake at 08:00-10:00 | 1.10 (1.02-1.18) | 1.16 (1.01-1.32) |
| First intake at 10:00-12:00 | 1.19 (1.05-1.34) | 1.26 (1.05-1.51) |
| First intake at noon or later | 1.29 (1.10-1.51) | 1.46 (1.07-2.00) |
| Last intake before 19:00 | 1.13 (1.03-1.25) | 1.15 (0.96-1.39) |
| Last intake at midnight or later | 1.27 (1.06-1.54) | 1.38 (1.03-1.84) |
First-intake comparisons use 07:00-08:00 as the reference; last-intake comparisons use 19:00-20:00. A hazard ratio, or HR, of 1.29 represents a 29% higher relative mortality hazard, not a 29-percentage-point increase in the probability of dying.
Later first intake showed a progressively higher mortality association. Last intake showed a U-shaped pattern, with the lowest estimated hazard around 20:00. However, neither the 20:00-22:00 nor the 22:00-midnight category had significantly higher all-cause mortality than the reference category. The cardiovascular result for eating before 19:00 was also not statistically significant.
Eating-window results were less consistent. After adjustment for first-meal timing, a window of eight hours or less was not significantly associated with mortality overall:
- All-cause mortality: HR 1.06, with a confidence interval of 0.92-1.22.
- Cardiovascular mortality: HR 1.07, with a confidence interval of 0.82-1.38.
Subgroup analyses nevertheless found higher mortality among people who started eating before 09:00 and finished within eight hours. Among those starting at 09:00 or later, windows longer than 14 hours were associated with higher cardiovascular mortality.
The authors also modelled remaining life expectancy at age 50. Relative to the reference groups, it was estimated to be 2.46 years shorter for people whose first intake was at noon or later, and 2.35 years shorter for those whose last intake was at midnight or later. These are modelled differences between groups, not measured benefits achievable by changing meal times.
Novelty. The contribution is mainly a more detailed epidemiological analysis of an existing question.
Earlier work had already associated first-intake timing with mortality. Kant and Graubard’s 2022 study examined 34,609 adults using NHANES data and found a modest survival advantage associated with earlier first intake. A 2024 NHANES analysis also associated late-night eating with mortality. The present study therefore extends previous findings using substantially overlapping source populations; it is not wholly independent replication. (PubMed)
Its more distinctive additions are:
- Finer timing categories, separating a moderately late breakfast from starting at noon or later.
- Joint consideration of first intake, last intake and window duration, including possible differences between early and late starters.
- Life-expectancy modelling, translating mortality associations into estimated years of remaining life.
The timing-by-duration analysis is arguably the most useful addition. The life-expectancy estimates are more striking, but they do not strengthen the underlying causal evidence.
Critique. The study has substantial strengths: a large sample, thousands of deaths, prospective mortality ascertainment, appropriate attention to NHANES sampling design, and extensive adjustment. The authors also report sensitivity analyses excluding early deaths and baseline cardiovascular disease or cancer, using two dietary recalls where available, and adjusting for additional social and sleep-related factors.
Several limitations materially affect interpretation.
-
One day’s intake is a weak measure of long-term behaviour.
A participant could be classified as a late eater because of an unusual workday, illness, travel or a missed breakfast. Eating habits could then change during years of follow-up.
Using two recalls in a sensitivity analysis helps, but does not establish a sustained eating pattern. The suggestion that older adults have more stable schedules does not validate one day’s recall as a measure of their habitual timing.
Counting every caloric drink as a meal also means these categories do not necessarily correspond to breakfast or dinner as people ordinarily understand them.
-
Confounding and reverse causation remain substantial concerns.
Late starters differed in income, activity, diet quality and other characteristics. For example, 31.5% of those starting at noon or later were in the lowest income category, compared with 13.8% of those starting at 07:00-08:00.
People with short eating windows also tended to have lower calorie intake, poorer diet quality and more existing disease. Illness, frailty or poor appetite could therefore produce a short eating window while independently increasing mortality.
Statistical adjustment reduces these problems but cannot remove them completely, particularly when health and lifestyle variables are measured imperfectly. Excluding deaths within two years does not eliminate illnesses that affect eating for longer periods.
-
Window duration and meal timing are mathematically linked.
At a fixed starting time, extending the eating window means finishing later. Consequently, the analysis cannot cleanly treat duration and timing as three separate biological influences.
The subgroup findings illustrate this:
- Starting before 09:00 and eating for no more than eight hours means finishing before 17:00.
- Starting at 09:00 or later and eating for more than 14 hours means finishing after 23:00.
These results may therefore reflect factors associated with very early finishing or late-night intake, rather than effects of fasting duration itself.
-
The subgroup estimates are much less secure than the total sample suggests.
The early-starting, eight-hour-or-shorter group contained only 44 cardiovascular deaths. Its cardiovascular HR of 1.84 had a wide confidence interval of 1.06-3.16.
The late-starting group with windows longer than 14 hours contained only 32 cardiovascular deaths.
Numerous subgroup, outcome and interaction analyses increase the opportunity for chance findings. The additional split at an 11-hour window was explicitly exploratory and followed inspection of the nonlinear patterns. These findings need confirmation in independent data.
-
The study does not directly test intentional time-restricted eating.
A short eating window arising from poor appetite or irregular routines is not equivalent to a planned intervention with adequate nutrition.
Randomised evidence also addresses a different question. For example, a five-week trial found that early time-restricted eating improved insulin sensitivity and several metabolic measures. Such short trials cannot establish longevity benefits, but this observational mortality study cannot establish that those interventions are harmful either. (Nature Communications)
-
The life-expectancy language overstates what the design can establish.
The life-table estimates carry forward the uncertainty and possible bias in the observational associations. Their confidence intervals do not account for every possible source of confounding or exposure misclassification.
The discussion’s suggestion that optimal meal timing offers an effective way to enhance longevity therefore goes beyond the evidence. The study estimates differences between people with different reported eating patterns; it does not estimate the proven effect of moving someone’s breakfast or dinner.
-
The proposed mechanisms remain untested.
Circadian misalignment, altered lipid handling and muscle loss are plausible explanations discussed by the authors, but none was demonstrated as the pathway connecting meal timing to death in this cohort. Clock time also does not precisely identify an individual’s biological circadian phase.
A stronger follow-up would repeatedly measure eating and sleep timing, distinguish deliberate restriction from illness-related reduced intake, assess frailty and nutritional adequacy, and replicate the timing-by-duration findings in a separate cohort. Those steps would help determine whether meal timing contributes to mortality or mainly identifies people with different underlying health and circumstances.