Does eating early help?

A cohort of 31,044 US adults aged 40 and older from NHANES (1999–2018) was followed for a median 8.6 years, tracking 7,129 deaths, with meal timing estimated from 24-hour dietary recalls. Compared with a last meal at 19:00–20:00, all-cause mortality was 13% higher for a last meal before 19:00 (hazard ratio 1.13, 95% CI 1.03–1.25) and 27% higher for eating after midnight (1.27, 1.06–1.54). A later first meal carried a similar pattern in one direction: 10% higher mortality for 8:00–10:00, 19% for 10:00–12:00, and 29% for after noon, all versus a 7:00–8:00 first meal. At age 50, estimated life expectancy was 2.46 years shorter for a first meal after noon and 2.35 years shorter for a last meal after midnight, versus the reference groups [3]. The study found no clean floor where "earlier is always better" — the earliest last-meal group did worse than the mid-evening one, not better.

In 103,389 adults in the French NutriNet-Santé cohort, followed a median of roughly 7 years with meal timing from repeated 24-hour dietary records, a last meal after 9pm carried 28% higher risk of cerebrovascular disease (cerebrovascular hazard ratio 1.28, 95% CI 1.05–1.55) than a last meal before 8pm, with each added hour of delay adding about 8% (hazard ratio 1.08, 1.01–1.15, p=0.02). The link to overall cardiovascular disease was weaker and not quite statistically significant (hazard ratio 1.13, 0.99–1.29). The authors’ own framing: coupling a longer overnight fast with an earlier last meal looked more useful for prevention than skipping breakfast [1].

Eight men with prediabetes were randomized, in a crossover design, to 5 weeks each of early time-restricted feeding (a 6-hour eating window with dinner finished before 3pm) or a control 12-hour eating window, at a fixed calorie level to prevent weight loss. Early time-restricted feeding lowered systolic blood pressure by 11 ± 4 mmHg and diastolic by 10 ± 4 mmHg versus the control schedule (both p=0.03) — a drop the authors compare to the effect size of an ACE-inhibitor blood pressure medication — alongside improved insulin sensitivity and lower oxidative stress markers, all without weight loss [4]. This is the only randomized evidence on this page, which makes it the strongest for causation. It is also the narrowest protocol: dinner before 3pm is a far tighter window than "eat dinner a few hours before bed," and the trial ran 8 men for 5 weeks.

Thirteen healthy young women completed a randomized crossover trial with two conditions: dinner 1 hour before habitual bedtime, or 5 hours before, each held for 4 days with matched 709–740 kcal meals. Overnight sleep EEG on the late-dinner condition showed significantly shorter total sleep time (p=0.013), lower sleep efficiency (p=0.040), more time awake after falling asleep (p=0.017), a higher arousal index (p=0.041), and more sleep-stage shifts (p=0.003) than the early-dinner condition. Postprandial glucose showed only a transient interaction effect, and next-morning glucose tolerance did not differ [5].

In 71,838 Japanese adults (28,625 men, 43,213 women) aged 40 to 79, free of cardiovascular disease and cancer at baseline and followed 19 years, people with an irregular supper time had 44% higher hemorrhagic stroke mortality than those eating supper consistently before 8pm (hazard ratio 1.44, 95% CI 1.05–1.97). There was no significant difference between early (before 8pm) and late (after 8pm) supper timing on its own for coronary heart disease, other stroke types, or total cardiovascular mortality [2].

All of the cohort numbers above come from self-reported 24-hour dietary recalls, not observed eating, and none can fully rule out reverse causation: shift workers, people with insomnia, night-shift jobs, depression, or undiagnosed illness plausibly eat later or more irregularly for reasons that independently affect mortality. That is the likely explanation for the NHANES U-shape, where the earliest last-meal group — plausibly including people who eat little dinner because they feel unwell, skip meals under stress, or have unusual schedules — did worse than the mid-evening group, not better. The NutriNet-Santé and Japan cohorts used different definitions (a single 9pm or 8pm cutoff versus continuous timing) and different populations, which is part of why their conclusions about "how early is early enough" do not line up neatly.

Insulin sensitivity, glucose tolerance and blood pressure all follow a daily circadian rhythm driven by the same clock genes that govern sleep, and eating late pushes food intake into the part of the day when the body is metabolically least prepared to handle it — the mechanistic case reviewed in Poggiogalle et al. [6]. This is consistent with, but does not by itself prove, any of the specific numbers above; it is the working explanation for why timing might matter at all, independent of what or how much is eaten.

Put together: the field agrees eating very late (after midnight, or after 9pm for stroke specifically) tracks with worse outcomes, and the one randomized trial shows a tight early-eating window can causally lower blood pressure. It does not agree that earlier is linearly better all the way down, and one large cohort suggests a consistent time matters as much as an early one. Nothing here has been turned into a years-based number the way sleep, exercise or smoking have; this stays an article, not a scored factor on your Health Score.