Does it matter when you eat?

In short

In the TREAT trial, twelve weeks of 16:8 eating matched three structured meals to within 0.26 kg (95% CI −1.30 to 0.78, P = .63).
Give both arms the same calorie prescription for a year and the window adds −1.8 kg (95% CI −4.0 to 0.4). Under isocaloric feeding it adds 0.3 kg (95% CI −1.2 to 1.9), in the wrong direction.
TREAT's awkward secondary finding was lean mass: a between-group appendicular lean mass index difference of −0.16 kg/m² (95% CI −0.27 to −0.05).
Against no intervention, an uncounted eight-hour window still gave −4.61 kg at a year (95% CI −7.37 to −1.85), cutting 425 kcal a day.

Two different questions, usually answered as one

Compress your eating into eight hours and you will probably eat less. That is the difficulty with this literature: a shorter window is not only a change in timing but, in practice, a change in intake, so any trial pitting it against unrestricted eating measures both at once. To isolate the clock you must remove the calories as a variable — prescribe the same deficit to both arms, or feed people a fixed amount and vary only the hours.

A window against a meal schedule

The TREAT trial randomised 116 adults with overweight or obesity to twelve weeks of either 16:8 time-restricted eating — eat ad libitum from 12:00 pm until 8:00 pm — or three structured meals a day.¹ Weight fell by 0.94 kg in the time-restricted arm (95% CI −1.68 to −0.20) and 0.68 kg in the meal-timing arm (95% CI −1.41 to 0.05): a difference of −0.26 kg (95% CI −1.30 to 0.78, P = .63). Fasting glucose, insulin, HOMA-IR, HbA_1c, triglycerides and cholesterol fractions moved in neither group, and estimated energy intake did not differ.

The result that drew attention was secondary. Among the 50 participants scanned by DXA in person, the between-group difference in appendicular lean mass index was −0.16 kg/m² (95% CI −0.27 to −0.05, P = .005). Decomposing the time-restricted arm's own 1.70 kg of loss, the authors reported that 1.10 kg (approximately 65% of weight lost) was lean mass; only 0.51 kg of weight loss was fat mass, against a normal expectation they put at 20% to 30%. That 65% is a within-group breakdown in a small subcohort, not a randomised contrast; the between-group figure is sturdier, and points the same way.

Holding the deficit constant

A twelve-month trial assigned 139 adults with obesity to eating between 8:00 a.m. and 4:00 p.m. with calorie restriction, or to identical calorie restriction with no window.² Both arms were prescribed 1,500–1,800 kcal a day for men and 1,200–1,500 for women, and adherence was near-identical at 84.0% and 83.8% of days. Weight fell 8.0 kg (95% CI −9.6 to −6.4) with the window and 6.3 kg (95% CI −7.8 to −4.7) without: a net difference of −1.8 kg (95% CI −4.0 to 0.4, P = 0.11), with body composition and metabolic risk factors improving equally in both. The authors' conclusion: These results indicate that caloric intake restriction explained most of the beneficial effects seen with the time-restricted–eating regimen.

A second trial removed even the possibility of an intake difference by providing all the food. Forty-one adults with obesity and prediabetes or diet-controlled diabetes spent twelve weeks on a ten-hour window with 80% of calories before 1 p.m., or on a usual pattern of up to sixteen hours with most calories after 5 p.m.³ Weight fell 2.3 kg on the early window and 2.6 kg on the late one — a difference of 0.3 kg (95% CI −1.2 to 1.9) favouring the longer window — and glycaemic measures did not differ. Small and narrow, but the only design in which timing was genuinely all that changed.

Table 1. Randomised comparisons of a restricted eating window against an active comparator; a negative difference favours the window. TRE = time-restricted eating; CR = calorie restriction; ER = energy restriction; UEP = usual eating pattern.
Trial	Design	Difference in weight change (95% CI)
Lowe et al., 2020 JAMA Intern Med	16:8 TRE vs 3 structured meals n = 116, 12 weeks	−0.26 kg (−1.30 to 0.78)
Liu et al., 2022 N Engl J Med	8-h window + CR vs CR alone n = 139, 12 months, matched calorie targets	−1.8 kg (−4.0 to 0.4)
Lin et al., 2023 Ann Intern Med	8-h window, no counting, vs 25% CR n = 90, 12 months	0.81 kg (−3.07 to 4.69)
Maruthur et al., 2024 Ann Intern Med	Early 10-h window vs late UEP n = 41, 12 weeks, isocaloric feeding	0.3 kg (−1.2 to 1.9)
Jamshed et al., 2022 JAMA Intern Med	Early 8-h window + ER vs ≥12-h window + ER n = 90, 14 weeks	−2.3 kg (−3.7 to −0.9)

Figure 1. Five randomised contrasts against the dashed line of no difference. The comparators differ by row, so read this as how consistently the intervals straddle zero, not as a pooled estimate. Only Jamshed and colleagues, who moved the window to the morning, excluded zero.

Where the literature genuinely disagrees

Against that run of nulls sits a 2026 systematic review of 30 randomised trials and 1,341 participants, which split its analysis by whether the control was isocaloric.⁷ Against non-isocaloric controls the time-restricted groups lost 2.82 kg more (95% CI −3.49 to −2.15) — the confounded comparison. But against isocaloric controls the pooled difference was still −1.46 kg (95% CI −2.65 to −0.26), with fat mass down 1.50 kg (95% CI −2.77 to −0.24). The authors conclude that time-restricted eating yields favorable anthropometric and clinical outcomes, even when intake is isocaloric between the intervention and control groups.

Why does that not settle it? Because the pooled estimate rests mainly on short trials with modest samples, in which "isocaloric" is often verified by food logs rather than provided meals. The trials with the longest follow-up and the tightest control of intake are precisely the ones that find nothing — which suggests any timing effect is small enough to be swamped by measurement. One finding pointed the same way throughout: under isocaloric conditions fat-free mass also fell more with time restriction, by 0.41 kg (95% CI −0.79 to −0.03) — the direction TREAT flagged.

Morning windows and evening ones

If timing does anything, circadian biology predicts it should favour eating earlier — and that is the one place the evidence shows a signal. A 14-week trial randomised 90 adults with obesity to a window of 7:00 to 15:00 or a self-selected window of at least twelve hours, both arms given identical energy-restriction counselling.⁵ The early window was more effective for weight (−2.3 kg; 95% CI −3.7 to −0.9; P = .002) and lowered diastolic blood pressure by a further 4 mm Hg (95% CI −8 to 0; P = .04), but did not significantly affect body fat (−1.4 kg; 95% CI −2.9 to 0.2; P = .09) or the ratio of fat loss to weight loss (−4.2%; 95% CI −14.9 to 6.5).

The mechanistic case is older and much smaller. A supervised crossover trial fed eight overweight men with prediabetes all of their food for five weeks on a six-hour early schedule with dinner before 15:00, then five weeks on a twelve-hour schedule, with weight held stable.⁶ The early schedule cut insulin resistance, measured by the three-hour incremental AUC ratio, by 36 ± 10 U/mg (p = 0.005) and fasting insulin by 3.4 ± 1.6 mU/l (p = 0.05) without changing fasting glucose; morning systolic and diastolic pressure fell 11 ± 4 and 10 ± 4 mm Hg (both p = 0.03), and evening desire to eat dropped 22 ± 7 mm (p = 0.007). Eight men, surrogate endpoints, every meal supplied: a hypothesis, not a plan.

Not better than a deficit is not the same as useless

The most practically useful trial had three arms and ran for a year.⁴ Ninety adults with obesity were told to eat ad libitum between noon and 8:00 p.m. without counting calories, to cut energy intake by 25% daily, or to eat over ten hours or more. At twelve months the window group had lost 4.61 kg against control (95% CI −7.37 to −1.85) and the counting group 5.42 kg (95% CI −9.13 to −1.71); between the active arms the difference was 0.81 kg (95% CI −3.07 to 4.69, P = 0.68). Energy intake fell 425 kcal a day with the window and 405 kcal with counting.

A rule about hours produced almost exactly the deficit that arithmetic produced, without the arithmetic. That is an adherence result, not a metabolic one — and adherence is where most diets fail. "Adds nothing beyond a calorie deficit" is compatible with "is a good way of reaching one".

The bottom line

The clock is a delivery mechanism, not an ingredient. Once the deficit is matched, a window buys somewhere between nothing and a couple of kilograms, and the intervals in the best-controlled trials include zero. If a shorter window is what makes eating less feel manageable, adopt it — earlier in the day rather than later, with protein kept up and resistance training kept in, because the lean-mass signal is small but consistent. If it makes you miserable, nothing here says you are giving up an advantage by dropping it.

What these trials keep returning to is total intake — the one variable almost nobody estimates accurately without writing it down. Kettle logs food macros, including energy, alongside strength training, runs and rides, with the data living on your iPhone.