Fasting and time-restricted eating: what the trials show, protein timing, and who should not fast

Three different things

The word "fasting" covers three practices with different evidence bases, and most arguments about it are people comparing different ones.

• Calorie restriction (CR): eating 10–25% less than you would otherwise, every day, indefinitely. The intervention with the animal lifespan data.
• Time-restricted eating (TRE): eating the same food inside a window of 6–10 hours, usually 16:8. A timing intervention, not necessarily a calorie one.
• Intermittent and prolonged fasting: alternate-day fasting, 5:2, or fasts of two to five days, a few times a year. Larger stimulus, larger drop-out.

Calorie restriction: the evidence

In rodents, restricting calories by 20–40% extends median and maximal lifespan reliably — the oldest result in the field, from 1935. In rhesus monkeys the two long studies disagreed (Wisconsin found longer life, the NIA did not), and the difference seems to have been the control diet and when restriction started. The human trial is CALERIE 2 (Kraus 2019, Lancet Diabetes & Endocrinology): 218 non-obese adults randomised to 25% calorie restriction or usual diet for two years. They managed about 12% restriction in practice, lost 7.5 kg, and improved blood pressure, LDL, insulin sensitivity and hsCRP, with a slower rise in a composite biological-age score. A 2022 follow-up in Science (Spadaro et al.) found thymic fat replaced by tissue and higher naive T-cell output in the restricted group — a human echo of the animal data.

CALERIE also showed the cost: the restricted group lost lean mass and bone density along with fat, and the trial excluded anyone over fifty and anyone lean. Lifelong 12% restriction in a 60-year-old with declining muscle is a different proposition from two years in a 35-year-old.

Time-restricted eating: what the trials actually found

The pattern across all of them: when calories are matched, the window adds little to weight. When the window is early — eating finished by mid-afternoon — there is a modest, repeatable metabolic benefit that fits the circadian biology of insulin sensitivity, which is highest in the morning. And a late window, the 16:8 most people actually do (noon to 8 p.m.), is simply a way of skipping breakfast, which some people find makes eating less easier and others find makes eating more at night easier.

Protein and muscle: the real cost

The problem with every form of fasting for a person over forty is not hunger, it is the amino acids. Muscle-protein synthesis is triggered by a dose of roughly 0.4 g/kg of protein with 2.5–3 g of leucine, and it saturates — the muscle cannot bank a double portion. Fitting 1.6–2.2 g/kg into an eight-hour window means three large protein meals in eight hours, which most people do not manage; a six-hour window makes it harder still, and TREAT showed the lean-mass cost of not managing it. Alternate-day fasting makes the arithmetic impossible on the fasting days.

The compromise most protocols land on: a 10–12-hour window, early rather than late, with three protein-anchored meals, a hard stop three hours before bed, and resistance training twice a week. That captures most of the circadian and overnight-fasting benefit — the overnight fast is the part with the strongest evidence — without the muscle cost.

Who should not fast

• Anyone pregnant or breastfeeding, and anyone under 18.
• Anyone with a history of restrictive or binge eating. Fasting is a structured restriction and reliably re-opens the pattern. This is the exclusion we apply most often.
• Anyone on insulin or a sulphonylurea, or any glucose-lowering medicine, without their prescriber adjusting the dose.
• Underweight adults, and older adults with low muscle mass (appendicular lean mass below the sarcopenia cut-offs on DEXA). The lean-mass cost is the one they cannot afford.
• Anyone training hard twice a day or in a heavy block: the fuelling does not fit the window.
• People with reflux, gallstones or gout, where fasting and the first meal after it provoke the problem; and anyone with a heart rhythm problem doing prolonged fasts, because of electrolyte shifts.
• Women with cycle disruption: an aggressive window on top of training is a common cause of a lost period, and the bone cost follows.

What to measure

• DEXA before and after any fasting phase — lean mass and bone, not just fat. The scale cannot tell you which you lost.
• Protein, tracked for a week, to confirm 1.6 g/kg actually fits the window.
• Fasting insulin, HbA1c, triglycerides: the markers that an early window should improve; if they have not moved in three months, the window is not the lever.
• A two-week CGM to see the real effect of your window on the overnight and dawn glucose.
• Sleep from a wearable, because a late last meal or a hungry night shows there first.
• Resting heart rate, cycle regularity, grip: the early warnings that the restriction is too much.

Where peptides and therapies fit

Fasting and the compound layer intersect in two places. On a fat-loss plan, the fat-loss goal page describes how the eating window, the protein floor and the lifting are set first, and the tirzepatide and retatrutide pages explain why appetite suppression makes the protein arithmetic harder, not easier — the GLP-1 transition guide covers the phasing. On a recomposition or lean-mass plan, the protocol will usually say not to fast at all, because the steady amino-acid supply that growth-hormone-axis compounds depend on does not fit a short window. The metabolic-health article covers the insulin biology underneath all of it.

Common mistakes

• A late 16:8 that is really "no breakfast, big dinner at 9 p.m." — the window with the least evidence and the worst effect on sleep.
• Fasting instead of lifting. The lean mass goes with the fat.
• Two protein meals in six hours and calling it 1.6 g/kg.
• Prolonged fasts for "autophagy" that nobody can measure in a person, with real electrolyte risk.
• Fasting through a hard training block and blaming the programme for the plateau.
• Ignoring the history. If restriction has ever been a problem, the answer is no, whatever the trial says.