Meal timing and circadian rhythm: what the trials measured on a clock

Four days of eating before 2 pm cut mean 24-hour glucose by 4 mg/dL in eleven people, with cortisol and clock-gene rhythms shifting. A 59-person liver trial improved HbA1c only in the early window. And a 17-person microbiome crossover found almost nothing.

mixedStudies disagree, or the mechanism is contested.Revised 17 Oct 2026

A plain white ceramic bowl on a light wooden table beside a wall clock with a completely blank face and a folded linen cloth

Every trial below held the food constant and moved only the clock. That is the only reason any of them can say anything about timing.

Meal timing is the rare corner of this literature where the intervention is free and the design is clean: hold the food exactly constant, move the clock, and see what changes. Four trials below did that. They do not all point the same direction, which is the most interesting part.

Eleven people, four days, continuous glucose monitoring

Jamshed and colleagues, in Nutrients 2019, ran what they describe as the first study to determine how time-restricted feeding affects gene expression, circulating hormones and diurnal patterns in cardiometabolic risk factors in humans. Eleven overweight adults, a four-day randomized crossover, eating either between 08:00 and 14:00 or between 08:00 and 20:00, with continuous glucose monitoring and blood draws for hormones and whole-blood gene expression.

The metabolic numbers: mean 24-hour glucose fell by 4 ± 1 milligrams per decilitre with early time-restricted feeding (p = 0.0003), and glycemic excursions by 12 ± 3 milligrams per decilitre (p = 0.001). The molecular numbers: in the morning before breakfast, ketones, cholesterol, and the expression of SIRT1 — a stress-response and ageing gene — and LC3A, an autophagy gene, all increased (all p < 0.04); in the evening, brain-derived neurotrophic factor tended to increase (p = 0.10) and MTOR expression increased (p = 0.007). Early feeding also altered the diurnal patterns of cortisol and the expression of several circadian clock genes (p < 0.05).

Eleven people and four days. The glucose effect is statistically clean and biologically small, and both of those descriptions are true.

Fifty-nine people, twelve weeks, a liver

The CHRONO-NAFLD trial, published in Alimentary Pharmacology & Therapeutics in 2025, is the larger of the set. Fifty-nine people with metabolic dysfunction-associated steatotic liver disease completed it — 27 males, mean age 52.9 years, BMI 32.1 kg/m2. All three arms were on a hypocaloric Mediterranean-type diet; the only difference was the window. Control was unrestricted, early time-restricted feeding ran 08:00–18:00, late time-restricted feeding ran 12:00–22:00, both a ten-hour window, over 12 weeks.

All three groups showed significant twelve-week reductions in body weight, anthropometry and blood pressure. Glycated haemoglobin and insulin resistance — measured three ways, by the Matsuda index, by HOMA-IR and by the fasting-glucose-to-insulin ratio — improved at 12 weeks in the early window group only.

Read that sentence carefully twice. The diet did the weight loss. The clock was associated with the glycemic change.

Seventeen people and an almost empty result

Papadopoulou and colleagues, in Annals of Nutrition and Metabolism 2020, asked a microbiome question. Seventeen healthy adults randomly consumed an isocaloric diet for seven days, twice, alternating lunch with dinner so that 60 per cent of daily energy came at midday in one block and in the evening in the other, with a two-week washout. Fecal samples from the last three days of each intervention went through 16S rRNA amplicon sequencing, qPCR for representative bacterial groups, and short-chain fatty acid output.

The result is a null result and the authors say so plainly: meal timing had no significant effect on fecal SCFA output, energy loss in feces, microbial community profiling, or relative abundance. No effects on blood biomarkers of cardiometabolic health. The single exception is a specific one — the absolute concentration of Escherichia coli was significantly higher after the large lunch intervention (p = 0.02). Their conclusion uses the word “minimal” for the acute effects of main meal timing on the microbiome and on cardiometabolic biomarkers.

If you have read somewhere that meal timing “resets your gut clock”, this is the study to check it against.

The eight-week pair, and its silence

Allison and colleagues’ 2021 Current Biology crossover is discussed in this blog’s entry on sleep and glucose, but the relevant detail is different here. Twelve healthy adults, two eight-week counterbalanced conditions, identical meals, identical sleep-wake cycles, identical exercise. Intake limited to 0800–1900 h, or to 1200–2300 h. Effect sizes ran d = 0.397 to 1.019, which is a real difference, and the delayed schedule showed higher weight, HOMA-IR, triglycerides and lower resting energy expenditure, respiratory quotient, fasting glucose, insulin, total and HDL cholesterol and adiponectin.

What did not change, in the same abstract: circadian phase and amplitude of melatonin, cortisol, ghrelin, leptin and glucose. Twelve weeks of eating ten hours later did not measurably move a single clock hormone.

That contrast is the whole state of the field in one sentence. Move the meals and the metabolic markers move. Whether that happens because the clock moved is exactly the part nobody has established.

What this evidence does not show

It does not show that an early eating window causes weight loss: in the largest trial here, the weight loss happened in all three arms on the same diet. It does not show a link between meal timing and the gut microbiome, and the one study that measured it found nothing beyond a single bacterial count. It does not show that the metabolic effect is mediated by circadian misalignment, because the eight-week study found no change in clock hormones and the four-day study found gene-expression changes that have not been shown to produce the glucose effect. Eleven people over four days and twelve people over eight weeks are also, on their own, the smallest kind of evidence.

This article summarises published trials. It is not medical advice and it is not a recommendation for any individual.