Longevity Lab Report

Analysis

The Dinner Trial: The Same Meal at 10 PM Raised Overnight Glucose 18 Percent Over Eating It at 6

Windows are the popular unit. The interval between dinner and sleep is the measured one, and the effects are modest, partly calorie-driven, and real enough to grade.

Saved articles

Ask ChatGPT about this article ↗

Copy the prompt, then paste it into ChatGPT to explore the findings and sources.

Open ChatGPT ↗

The findings

The dinner question is narrower than the fasting-window question, and it has its own evidence: in the one direct trial, the identical meal eaten at 10 PM produced an overnight glucose curve about 18 percent higher than eating it at 6 PM, and the graded variable is the gap between your last meal and sleep.

Most writing about time-restricted eating argues about the window: eight hours or ten, noon to eight or nine to five. The dinner question is smaller and more practical than that argument. It asks when the last meal should end relative to sleep, and it turns out the evidence speaks to that narrower question more directly than it speaks to windows in general. Longevity Lab Report grades time-restricted eating C as a weight-loss adherence tool, C- for metabolic improvement independent of weight loss, and D+ for the longevity and autophagy claims made for it. Inside those grades sits a dinner-specific finding worth separating out: the gap between your last meal and your sleep is the variable the trials actually measure moving.

The one direct trial

The cleanest dinner experiment is Gu and colleagues' 2020 study in the Journal of Clinical Endocrinology and Metabolism, which fed the same people the same meal at 6 PM and at 10 PM and measured what followed. The late dinner produced an overnight glucose curve 17.8 percent higher and a post-meal glucose peak 18 percent higher than the early dinner, with free fatty acids 23 percent higher as well. Nothing else changed: same meal, same people, same total food. Only the clock moved, and the metabolic response moved with it, in the direction every circadian researcher would predict and at a size larger than most expected.

The newest item points the same way at lower certainty. A 2026 secondary analysis by Reutrakul and colleagues in Nutrients followed 44 adults with obesity for 12 weeks and found that each additional hour between the last meal and sleep was associated with lower overnight glucose. It is an exploratory analysis of a trial designed for other questions, in non-diabetic participants, with no measurement of melatonin onset, and this page prints it at exactly that weight: consistent with the direct trial, not a replacement for one.

What the window trials add, and what they take away

The strongest positive metabolic result in the whole time-restricted eating literature is an early-window trial: Sutton and colleagues' 2018 study, in which men with prediabetes ate their last food before 3 PM on a calorie-matched schedule and improved insulin sensitivity and blood pressure without losing weight. It is the trial our graded record's dose note rests on: window placement is the dose variable with any signal, and the only positive metabolic trial used an early window.

The counterweight is equally specific. When calories are held equal, the window itself adds little: the TREAT trial's noon-to-8 schedule was null, and Liu and colleagues' 12-month trial in the New England Journal of Medicine in 2022 found no added benefit from an 8-hour window on top of calorie restriction. A network meta-analysis in BMJ Medicine in January 2026, pooling 41 randomized trials and 2,287 participants, ranked early windows first, with about 1.15 kg more weight loss than late windows, and mid windows finishing between 5 and 7 PM second. Read together, the literature says something narrower than either camp's headline: eating earlier is mildly better than eating later, much of the benefit travels with eating less, and the dinner-to-sleep interval is where the timing effect, such as it is, actually lives.

One small trial sharpens the asymmetry. In a metabolic-chamber study at Hohenheim with 17 adults, a day that skipped breakfast produced a 46 percent higher glucose response to lunch and a 54 percent worse insulin-response index, while a day that skipped dinner produced neither, and burned slightly more fat. Seventeen people in a chamber is a mechanism demonstration, not a prescription, but it undercuts the assumption that the two ends of the window are interchangeable: they are not, and dinner is the end the evidence keeps pointing at.

Late is a relative term

Every rule on page one of the search results says some version of "finish eating by 7 PM," and the rule is cruder than the biology. The circadian system does not read the wall clock; it reads melatonin onset, which varies by hours between people. A 7 PM dinner is early for a person whose melatonin rises at 11 and late for a person whose rises at 8. The trials measure clock times because trials must, but the variable they are proxying is the interval between the last calories and the body's night. That is why this page grades the interval rather than the hour: finish eating roughly three hours before your own sleep is the practical form of the finding, and it moves with the reader instead of against them.

Who should not improvise this

Meal timing interacts with medication in ways windows do not. Anyone taking glucose-lowering medication, insulin above all, should not shift dinner earlier or later without their clinician, because the trials exclude them and the hypoglycemia risk is theirs, not the literature's. Pregnancy, a history of eating disorders, and shift work that scrambles sleep itself are all cases where the interval advice needs a professional in the room. For everyone else, moving dinner is among the lowest-risk experiments in this library, which is part of why the modest evidence still earns a practical answer.

Our grades and rationale

The grades are the record's: C for weight loss as a calorie-adherence tool, C- for metabolic improvement independent of weight loss, D+ for longevity and autophagy benefits. The dinner-interval evidence does not lift them, and this page does not pretend otherwise: one small direct trial, one exploratory analysis, an early-window mechanism trial, and pooled rankings that put early ahead of late by about a kilogram. What the interval evidence does is locate the benefit the C grades contain. If a reader takes one action from the time-restricted eating literature, the trials point at the last meal's distance from sleep, not at any particular window's branding.

Frequently asked questions

How long before bed should I stop eating? The evidence supports an interval of roughly three hours as the practical target: the direct trial compared a four-hour difference in dinner time, and the 2026 analysis found overnight glucose improving with each added hour of interval. Three hours is the form of the finding a household can actually run.

Does the eating window matter at all, then? As an adherence tool, yes: a window is how many people eat less without counting. As a metabolic intervention independent of calories, the large calorie-matched trials are null, and the grade reflects that.

Is breakfast-skipping the same as dinner-shifting? No. The one head-to-head chamber trial found breakfast-skipping worsened the day's glucose handling while dinner-skipping did not, and the pooled rankings favor early windows over late ones.

What about the traditional rule of eating before 7 PM? It is a rough proxy for the interval, right for people who sleep around 10 and wrong in both directions for everyone else. Grade the gap, not the hour.

Limitations: what we do not know

We do not know whether the overnight-glucose differences compound into outcomes anyone feels: no dinner-timing trial runs long enough or measures hard endpoints. We do not know the interval's effect in people with diabetes, who are excluded from the trials and most likely to be reading. And the newest item is a secondary analysis, the weakest design in this piece, printed as such. What would change the grades: a randomized trial of dinner timing itself, powered on glycemic outcomes over months, in either direction.

The record

The grades on this page, the dose note the early-window trial rests on, and the record's own citation layer are carried in full on the time-restricted eating record, which is updated as new trials land.

Protocols mentioned

Grades are research summaries, not medical advice. Decisions about your own health belong with your clinician.

Keep exploring

View all

The weekly newsletter

Keep up with what’s changing.

New research, useful tools, and the people shaping health and longevity. Selected and explained in one email.

See what you’ll receive ↗

Free. Confirm your email to subscribe. Unsubscribe anytime.