Advice about diet has traditionally concerned what and how much to eat. Time restricted eating (TRE) concerns when. The person eats freely, or nearly so, within a fixed daily window and takes only water or non-caloric drinks for the remaining hours. The approach grew out of experiments in mice, in which confining access to a high-fat diet to the active phase prevented obesity, hyperinsulinaemia and hepatic steatosis although the animals ate as much as those fed at will (Hatori et al., 2012; Chaix et al., 2014). The striking feature of those experiments, benefit without a reduction in energy intake, is the claim that human research has since tried to reproduce.

Interest spread quickly. A study using a smartphone application to log eating occasions suggested that many adults spread their intake across most of their waking hours and that shortening this span was feasible (Gill and Panda, 2015), and an influential review placed TRE within a wider account of the metabolic effects of fasting (de Cabo and Mattson, 2019). Trials followed in people with obesity, prediabetes and type 2 diabetes. Their results have been read both as confirmation and as refutation, often on the basis of the same papers.

This review asks a narrow question: what do recent trials show about TRE and glycaemic control, and how much of the apparent effect survives once the designs are taken into account. Weight, blood pressure and lipids are mentioned only where they bear on the interpretation of glucose outcomes.

This is a narrative review of published human trials together with the mechanistic work that motivates them. It is selective by intent. Trials were chosen because they are widely cited, because their design isolates a particular question, or because they enrolled people with impaired glucose regulation. No formal search strategy, screening record or pooled estimate is offered. Detailed accounts of the wider field are available elsewhere (Manoogian et al., 2022; Ezpeleta et al., 2024). The organising idea here is that each trial design removes some sources of confounding and leaves others, so that results should be compared within design classes before they are compared across them.

The physiological case for TRE rests on the daily rhythm of glucose metabolism. In healthy adults, glucose tolerance is better in the morning than in the evening, a pattern attributed to diurnal variation in insulin sensitivity and in beta cell responsiveness (Van Cauter et al., 1997). Laboratory protocols that separate the internal clock from behaviour confirm that the circadian system itself contributes to this variation and that eating and sleeping out of phase with the clock raises postprandial glucose (Scheer et al., 2009; Morris et al., 2015). A meal eaten late in the biological day is therefore handled less efficiently than the same meal eaten early.

Two predictions follow, and they are often blurred. One is that moving intake earlier should improve glycaemia regardless of the length of the window. The other is that lengthening the daily fast should help regardless of its position, through lower insulin exposure or a shift toward fat oxidation. Most free-living trials have used a window from about noon to early evening, which lengthens the fast but places intake in the later part of the day. Such a schedule is convenient and socially acceptable, but it tests the second prediction and works against the first. Figure 1 arranges the possible designs by these two properties and by whether energy intake is matched.

Table 1 lists the principal trials discussed in this section, grouped by the comparison each makes. The sections that follow take the groups in order of increasing control over energy intake.

Several features of this literature make individual results unreliable guides. Samples are small and durations short in exactly the designs that could isolate a timing effect, because controlled feeding is expensive. Adherence to the window is usually self-reported, and energy intake in free-living arms is estimated by food records whose errors are larger than the effects sought. The definition of the intervention varies in window length, clock position and whether energy is prescribed, so that trials bearing the same label test different hypotheses.

Outcome measures add further scatter. Fasting glucose, glycated haemoglobin, clamp-derived insulin sensitivity, oral glucose tolerance and sensor-derived mean glucose respond over different time scales. Glycated haemoglobin reflects average glucose over the preceding two to three months (Nathan et al., 2008) and cannot register a change in a study lasting a few weeks. Fasting samples taken after a longer overnight fast in the intervention arm than in the control arm are not strictly comparable. In trials enrolling people with diabetes, reductions in medication made for safety can mask an effect or, if unrecorded, create an apparent one. Finally, many trials report numerous metabolic outcomes without a prespecified glycaemic primary endpoint, which invites selective emphasis.

Taken by design class, the trials support three statements. Time restricted eating often leads people to eat less without counting calories, and when it does, weight and glycaemic markers improve to about the extent expected from the weight change. When energy intake is matched by prescription, the addition of a time window has not improved glycaemic control in the larger and longer trials. Evidence for a weight-independent benefit of an early window exists but rests on small, short feeding studies and is contradicted by a larger one. Current nutrition guidance for diabetes, which treats a range of eating patterns as acceptable and emphasises the one a person can sustain, is consistent with this reading (Evert et al., 2019).

The gaps are specific. An adequately powered controlled feeding trial comparing early and late windows of equal length, with a prespecified glycaemic endpoint, would settle whether clock position matters. Trials in type 2 diabetes need systematic recording of medication changes and hypoglycaemia. Objective measurement of adherence, for example by continuous glucose monitoring or timestamped photographs, should replace diaries. Follow-up beyond a year is almost absent, as are data on older adults, shift workers and people using insulin, for whom the balance of benefit and risk may differ.

The signal in recent trials of time restricted eating is that it offers one more way to lower energy intake, with glycaemic benefits in proportion. The noise comes from short, small and heterogeneous studies that have been read as evidence for something more. A specific effect of meal timing on glucose regulation is biologically plausible and has been seen under tightly controlled conditions, but it has not been reproduced at scale. Clinicians can offer the approach to people who find it easier than calorie counting, with attention to medication in those with diabetes, while describing its advantages over other methods of energy restriction as unproven.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.