Circadian Rhythms and Food: How Biological Time Shapes Metabolism

Body temperature is not exactly the same at eight in the morning and eight in the evening. Hormones are released in daily patterns. Sleepiness changes predictably across the day, as do alertness, digestion and the way the body handles glucose after a meal.

Some of these changes continue even when researchers remove familiar signals such as clocks and regular daily schedules. They are generated in part by the circadian system: an internal timing system that produces biological rhythms lasting approximately 24 hours.

Sleep is closely connected with this system, but sleep and circadian rhythms are not the same thing. Going to bed at a particular hour is a behaviour. A circadian rhythm is an internally generated biological pattern whose timing can be measured even under carefully controlled laboratory conditions.

Food has become an important part of this research because the physiological response to a meal is not constant across the day. Circadian phase describes a person's position within their own approximately 24-hour biological cycle rather than the hour shown on a clock. The same food can produce different glucose responses depending on that phase, and changing the timing of meals can alter rhythmic activity in some metabolic tissues.

The study of these relationships is usually described as chrononutrition. It includes questions about when people eat, how regularly they eat, how eating relates to sleep and whether changing meal timing affects metabolic health.

Moving dinner from one hour to another can also change the overnight fasting period, sleep proximity, energy intake or the number of opportunities to eat. Researchers therefore have to separate effects that occur together in ordinary life but may arise through different mechanisms.

Where circadian rhythms come from

A small region of the hypothalamus called the suprachiasmatic nucleus (SCN) plays a central role in coordinating circadian timing.

Light is the strongest environmental signal reaching this central pacemaker. The retina contains intrinsically photosensitive retinal ganglion cells, often abbreviated as ipRGCs. These cells contain melanopsin, a photopigment that responds to light, and they also receive information originating from the rods and cones involved in vision. Signals from this retinal system reach the SCN and help synchronize internal timing with the external light-dark cycle. [1,2]

The SCN is only part of the system.

Cells in the liver, pancreas, skeletal muscle, adipose tissue and many other organs also contain molecular machinery capable of producing rhythmic activity. Genes including CLOCK, BMAL1, PER and CRY participate in a wider network of feedback loops. CLOCK and BMAL1 proteins promote the production of PER and CRY proteins. As PER and CRY accumulate, they suppress CLOCK and BMAL1 activity; when their levels fall, that activity can rise again. These interactions contribute to approximately 24-hour patterns of gene expression and cellular activity. [1]

Scientists now study circadian organization as a coordinated network rather than treating daily physiology as the output of one clock in the brain. Signals from the SCN interact with hormones, behaviour and processes within individual tissues. Different organs do not necessarily reach their biological peaks at the same moment or respond identically to the same signal. [1]

Meals can influence some of these peripheral rhythms.

A controlled human experiment tested the effect of meal timing on different circadian rhythms. Researchers studied ten healthy young men during a 13-day laboratory protocol. For six days, three daily meals were delayed by five hours. The later schedule delayed a rhythm in blood glucose and the rhythm of PER2 expression measured in adipose tissue. The melatonin and cortisol rhythms measured by the researchers did not shift significantly. [3]

The measured rhythms did not all respond alike to the later meals. Changing meal time affected the timing of some measured peripheral markers without producing a corresponding significant shift in every marker examined. The experiment involved only ten men and lasted less than two weeks, so it cannot tell us whether such changes improve or worsen health over years.

Why the response to a meal can change across the day

Two meals can contain exactly the same amount and type of carbohydrate. If blood glucose rises differently after the two meals, the difference does not necessarily come from the food itself. The physiological state in which the food is processed can also differ.

After carbohydrate-containing food is digested, glucose enters the bloodstream. The pancreas responds by releasing insulin. Insulin then helps regulate how glucose is taken up and used or stored by tissues. Several parts of this process can vary with circadian timing.

Controlled laboratory studies have tried to separate circadian effects from ordinary differences between morning and evening behaviour.

In one experiment, participants received identical test meals at different points in their internal cycle: their biological morning and biological evening. Post-meal glucose was higher during the biological evening than during the biological morning. Researchers also experimentally inverted the participants' behavioural cycle, creating circadian misalignment. Post-meal glucose increased under that condition as well. [4]

Those two observations can look similar on a glucose graph while arising partly through different physiology.

Subsequent experimental work examined insulin secretion and insulin sensitivity. During the biological evening, poorer glucose tolerance was associated with reduced responsiveness of pancreatic β cells, the cells responsible for secreting insulin. During experimentally induced circadian misalignment, lower estimated insulin sensitivity contributed to the altered response. [4,5]

Statements such as "glucose control is worse at night" leave the mechanism unspecified. A higher glucose response does not identify a single mechanism. Insulin secretion, tissue sensitivity to insulin and circadian phase can contribute differently depending on the experimental condition.

The experiments were also short laboratory studies. They tell us about physiological responses under controlled conditions. They do not tell us that eating a particular meal in the evening will cause diabetes, nor do they establish an hour after which food suddenly becomes metabolically harmful.

Glucose outcomes in earlier and later eating trials

A 2026 systematic review and meta-analysis examined 44 human intervention studies that directly compared earlier and later food intake. [6]

Most were short. Thirty-seven lasted no more than one week, while only seven continued for between two and twelve weeks.

Across the shorter experiments, earlier food intake was associated with a pooled reduction of approximately 1.51 mmol/L in glucose measured two hours after eating and 1.03 mmol/L in peak post-meal glucose compared with later intake.

The studies did not produce uniform results. Statistical heterogeneity was high for the two-hour glucose analysis (I² = 79%) and substantial for peak glucose (I² = 61%). I² estimates how much of the variation in study results reflects differences between studies beyond what would be expected from sampling variation alone. Differences in participants, meals, timing protocols and study designs contributed to a body of evidence that cannot be reduced to one predictable response.

The longer studies were less conclusive. With only seven interventions lasting two to twelve weeks, they did not establish that the short-term glucose differences persist over longer periods.

There is also a practical difference between a pooled experimental comparison and moving one person's dinner from, for example, 20:00 to 19:00. The meta-analysis estimates an average difference across specific interventions. It does not provide a formula that predicts how much an individual's glucose response will change for every hour a meal is moved.

Time-restricted eating adds another variable

Much of the current discussion around meal timing involves time-restricted eating (TRE).

TRE limits daily food intake to a defined window. Someone using an eight-hour window might eat between 08:00 and 16:00, between 13:00 and 21:00, or within another eight-hour period.

Shortening the eating window can change meal timing and several other behaviours simultaneously.

A person may eat fewer meals or snacks simply because there is less time available for eating. Total energy intake may fall even when calorie reduction was not prescribed. The overnight fasting period becomes longer. The final meal may move farther away from sleep. An early window may also change social eating patterns differently from a late one.

For that reason, a TRE study is not automatically an experiment on the circadian clock alone.

Early, late and self-selected eating windows

A large randomized trial published in 2025 studied 197 adults with overweight or obesity. Participants received usual care alone or usual care combined with one of three eight-hour eating schedules: early TRE, late TRE or a self-selected TRE window. Usual care included education about a Mediterranean dietary pattern. [7]

The primary outcome was change in visceral adipose tissue, the fat stored around internal organs in the abdominal cavity.

After twelve weeks, none of the three TRE schedules produced a statistically significant additional reduction in visceral adipose tissue compared with usual care. The early, late and self-selected schedules also did not differ significantly from one another on this primary outcome.

Participants assigned to TRE lost more body weight than those receiving usual care alone, and the early-TRE group showed favourable changes in some secondary measures, including fasting glucose and subcutaneous fat (fat beneath the skin).

Primary and secondary outcomes answer different questions within a trial. The primary outcome is specified as the main test around which the study is designed. Secondary outcomes provide additional information but should not replace the primary result when the intervention is evaluated.

Here, the trial did not demonstrate an additional reduction in visceral fat, its primary outcome, while differences appeared in weight and some secondary metabolic measurements.

Comparing eating schedules with approximately similar food intake

Another randomized trial approached the problem differently. It studied 31 women with overweight or obesity in a crossover design, meaning that each participant experienced both eating schedules. [8]

For two weeks, the women followed an early window from 08:00 to 16:00. For another two weeks, they followed a late window from 13:00 to 21:00.

They were asked to maintain their habitual food quantity and quality. The study was intended to keep energy intake similar between schedules so the researchers could compare their timing.

Perfectly equal intake was not achieved. A modest energy deficit occurred during the early condition, and participants lost weight during both conditions.

The study found no significant difference in its measure of insulin sensitivity between the early and late schedules. Several other cardiometabolic measurements also showed no clinically meaningful differences between conditions. [8]

The researchers examined timing-related measures as well. The midpoint of sleep occurred about 15 minutes later during the late-eating condition.

They also estimated the phase of a rhythm measured in blood monocytes, a type of circulating immune cell. The estimated phase was 24 minutes later during late TRE, but the 95% confidence interval ranged from −5 to 54 minutes, with P = 0.10. Because that interval includes no difference, the molecular result remains statistically uncertain.

The behavioural sleep result and the molecular estimate therefore should not be treated as the same observation.

What larger analyses of TRE have found

Individual trials can produce different results because they use different populations, eating windows, control groups and outcomes. Meta-analyses attempt to combine these studies, although the interpretation still depends on which trials are being compared.

A 2026 network meta-analysis included 41 randomized controlled trials involving 2,287 participants. [9]

Across comparisons with usual diets, TRE was associated with improvements in several outcomes, including body weight and some metabolic measures.

The analysis also compared earlier and later TRE schedules. The network analysis combined the limited trials directly comparing early and late TRE with indirect comparisons through other interventions or control conditions. Earlier TRE was associated with approximately 1.15 kg lower body weight and 3.32 μIU/mL lower fasting insulin than later TRE. The authors rated the certainty of evidence for these two comparisons as high.

"High certainty" has a specific meaning in evidence assessment. It indicates confidence in the estimated comparison for those outcomes within the evidence reviewed. It does not establish why the difference occurred.

The included trials differed in participants, diets and intervention protocols. Earlier windows can change food intake, fasting duration, sleep proximity and other behaviours alongside circadian timing. The pooled difference therefore cannot be attributed automatically to better alignment with the circadian system.

A separate 2026 systematic review and meta-analysis included 30 trials and 1,341 adults with overweight or obesity. [10]

TRE was associated with lower pooled body weight and fat mass when compared with both isocaloric (designed for similar energy intake between groups) and non-isocaloric control strategies. The analysis also reported reductions in fat-free mass, the portion of body weight made up of muscle, organs, bone and water rather than fat. A change in this measure does not by itself establish a loss of skeletal muscle.

Isocaloric comparisons attempt to reduce the influence of calorie intake, but they do not make every trial a pure experiment on circadian timing. Adherence, actual intake, fasting duration and other behavioural differences remain relevant to interpretation.

Longer studies are beginning to answer different questions

Many meal-timing studies last days or weeks. That is enough to measure post-meal glucose or short-term changes in weight, but chronic metabolic diseases develop over much longer periods.

A 2026 randomized trial followed 247 adults at elevated risk of type 2 diabetes and compared a self-selected nine-hour eating window ending by 19:00 with individualized dietetic guidance. [11]

The study used HbA1c as an outcome. HbA1c reflects average blood-glucose exposure over the preceding several months and therefore provides a different view from measuring glucose for a few hours after one meal.

At four months, TRE met the trial's prespecified criterion for noninferiority to individualized dietetic guidance. In this type of analysis, the question is whether the intervention performs no worse than the comparator by more than a predefined margin. TRE was not superior to dietetic guidance.

At twelve months, the trial no longer established noninferiority. Changes in HbA1c were small and were not considered clinically meaningful in either group. [11]

This trial should not be interpreted as a direct test of early versus late eating. Participants selected their own nine-hour window, subject to the requirement that it ended by 19:00, and the comparison group received active individualized dietary support.

The trial adds information that short glucose experiments cannot provide: what happens when an eating-window intervention is maintained and compared with another structured dietary approach over months rather than days.

Eight o'clock does not represent the same biological time for everyone

Meals can be recorded by clock time without revealing where they fall in a person's biological cycle.

Two people can eat dinner at 20:00. One normally goes to sleep at 21:30. The other goes to sleep at 01:00. Their meals occur at the same clock time but at very different positions relative to their usual sleep period.

Researchers use several measurements to study these differences. Chronotype describes individual tendencies toward earlier or later patterns of sleep and activity. Circadian phase can be assessed more directly in laboratory research using physiological markers.

Chronotype and circadian phase are related concepts, but they are not interchangeable.

Shift work is another setting in which clock time and circadian phase can diverge. Eating during a night shift may occur alongside nighttime light exposure, sleeping during the day and repeated changes in the relationship between behaviour and internal circadian timing. Experiments show that circadian misalignment can alter glucose regulation. They do not show that one late meal explains the metabolic effects observed in people working irregular schedules. [4,5,13]

Research on chronotype has raised the possibility that individual biological timing could eventually become relevant to nutrition interventions. Current evidence has not produced a validated method for calculating an individual's optimal breakfast, lunch or dinner time from chronotype.

Sleep cannot be separated completely from meal-timing research

Changing an eating window may also change the relationship between food and sleep. Researchers have therefore examined whether TRE itself affects sleep.

A 2026 systematic review and meta-analysis included 13 studies with 638 participants. [12]

When researchers looked at changes occurring within TRE groups, they found small improvements in sleep duration and questionnaire-rated sleep quality.

Controlled comparisons were less conclusive. When analyses were restricted to controlled trials, there was no significant between-group difference in sleep duration or questionnaire-rated sleep quality. The review also found no significant effects on sleep onset or wake time.

Within-group and between-group analyses answer different questions in intervention studies. If participants sleep better after beginning an intervention, that change does not by itself show that the intervention caused the improvement. A control group helps estimate what would have happened without the intervention under otherwise comparable study conditions.

Genetics explains part of the clock, not a personal dinner time

Circadian rhythms are also a genetic phenomenon.

The molecular clock depends on genes and proteins that interact to generate rhythmic cellular activity. Genetic studies have identified variation in circadian-related genes and associations with chronotype, sleep characteristics and susceptibility to some diseases. [1]

That type of research can help scientists understand why circadian traits differ between individuals and how molecular timing interacts with physiology.

It is much further from telling someone when to eat.

A genetic association does not by itself demonstrate how a variant changes the response to breakfast or dinner. A molecular mechanism does not automatically establish a clinically meaningful dietary effect. And none of the meal-timing trials discussed here validates a genetic test that can calculate an individual's optimal eating window.

For nutrigenomics, that boundary is particularly relevant. Personalized nutrition requires more than identifying biological variation. The variation has to predict a meaningful response to an intervention with enough reliability to improve a real decision.

Current meal-timing research has not reached that point.

What remains uncertain

Whether the short-term differences in post-meal glucose persist over years, translate into meaningful changes in disease risk, or depend strongly on an individual's biological timing remains unresolved.

Answering these questions requires records of when people actually eat rather than only the schedule they were assigned. Energy intake, sleep, light exposure, physical activity and work schedules can alter the interpretation of meal timing. Researchers also need consistent definitions of early eating, late eating and time-restricted eating. Standardized terminology and better tools for recording eating times in everyday settings were among the priorities identified by a National Heart, Lung, and Blood Institute workshop held in 2023 and reported in 2025. [13]

Scientific note

This article is an educational account of current research. It does not provide individualized dietary or medical advice.