What happens during a fast, hour by hour

The switch you have read about does not exist

Almost every fasting timeline online tells the same story: you burn through your stored carbohydrate, and once it runs out, somewhere around hour twelve, you flip into fat burning. It is a tidy story. It is also not what the measurements show.

When researchers used carbon-13 magnetic resonance to watch glucose production directly, they found that gluconeogenesis, the process of manufacturing glucose from non-carbohydrate sources, was already supplying about half of total glucose production, 55% give or take, in the hours between dinner and dawn. Thirteen people. At that point liver glycogen was still close to maximal. The liver had not run out of anything. It was simply doing both jobs at once, as it always does.

By 22 hours gluconeogenesis accounts for around two thirds of glucose production, and by 42 hours more than 90%. Two independent research groups using different methods arrived at closely matching numbers, which is about as good as this kind of evidence gets.

So the honest version is less dramatic and more useful: there is no switch to wait for. There is a dial that starts turning the moment you stop eating.

Hours 0 to 12: digestion, then the quiet handover

For the first few hours after a meal your liver stops releasing glucose almost entirely, within about 30 minutes, and starts storing it instead. Liver glycogen peaks roughly five and a quarter hours after eating, which is later than most timelines suggest.

What follows is not a cliff. Insulin begins to fall, glucagon rises, and the liver gradually shifts from storing to producing. The ratio between those two hormones, rather than insulin alone, tracks what the liver does.

Fat release from your fat cells is extraordinarily sensitive to insulin. Half-maximal suppression happens at insulin levels inside the normal fasting range, which is why lipolysis picks up early and gradually rather than waiting for some threshold.

Hours 12 to 24: where most of the change actually happens

If there is a period that deserves the attention the internet gives to hour 16, this is it. In a study of six healthy young men fasted from 12 to 72 hours, fat release roughly doubled, and about 60% of that entire increase happened between hours 12 and 24. The single largest jump came between hours 18 and 24.

Roughly 70% of the total fall in insulin also happens in the first 24 hours. Ketones become measurable well before this, but the meaningful rise belongs to this window.

It is worth being clear about the size of the evidence, though. That widely repeated finding about hours 18 to 24 comes from six people. Six young men, specifically. It is a careful study and it is almost certainly directionally right, but it is not a law of nature.

Ketones: the numbers nobody actually has

Every hour-by-hour ketone chart you have seen online is interpolated. No single human study reports mean ketone levels at all the hours those charts show, so the smooth curve is drawn between a handful of real measurements taken at scattered timepoints. The precision is added afterwards.

The best single anchor point that exists comes from a 1982 study which measured beta-hydroxybutyrate at 30 hours of fasting and found a striking pattern: 0.9 mM in men, 1.7 mM in women, and 3.7 mM in children. Women reached roughly twice the ketone level of men at the same hour.

Even the question of when ketones plateau is unsettled. Tracer work puts it at around three days; other work has ketones still climbing at five to eight days and level by ten. Those cannot all be right, and the field has not resolved it.

Hours 24 to 48: growth hormone, and a number worth retiring

You will have seen the claim that fasting raises growth hormone by 1,300% in women and 2,000% in men. Those figures come from a 2011 hospital press release describing a conference presentation. They were never published in a peer-reviewed paper.

The peer-reviewed measurements are more modest and more interesting. Careful sampling every five minutes across 24 hours found growth hormone production rose about fivefold after two days of fasting, driven by more frequent and larger pulses. A separate five-day study found roughly a threefold rise.

Here is the part that almost never gets repeated: IGF-1, the hormone through which growth hormone drives tissue building, does not rise. It was unchanged at 56 hours in one study and fell by day five in another. Growth hormone during a fast is better understood as a starvation adaptation that helps limit protein loss than as an anabolic bonus.

The enormous percentages you see quoted are an artefact of arithmetic. Fed-state growth hormone troughs are close to zero, and dividing by a number close to zero produces a very large percentage from a modest absolute change.

Does fasting speed up or slow down your metabolism?

This is the clearest example of a question where the honest answer is that we do not know, and where confident articles on both sides are simply picking a study.

One group measured resting energy expenditure rising about 14% by day three of a fast in eleven people. Another group measured resting metabolic rate falling about 8% at exactly the same timepoint in six people. Same duration, same kind of healthy lean participants, comparable technique, opposite conclusions. A third analysis found metabolic rate per unit of body mass essentially constant across three weeks.

The longer-term picture is better established, though it rests on one unusual group. Fourteen contestants from a televised weight-loss competition still had resting metabolic rates several hundred calories a day below baseline six years on, even though most had regained the weight. That is the most extreme cohort in the literature rather than a forecast for an ordinary loss, and nothing about it is specific to fasting.

Hours 48 to 72: protein, and the limits of sparing

Fasting is often described as protein-sparing, which is true but partial. Protein breakdown does not slow at 72 hours; leucine turnover rose 31% and leucine oxidation rose 46% at three days. Net amino acid release from muscle increases. Nitrogen continues to leave the body indefinitely.

What genuinely does the sparing is fat. When researchers pharmacologically blocked fat release by about half in fasted people, protein breakdown rose roughly 70%. That is a causal demonstration, and it is the strongest single piece of mechanistic evidence in this whole area.

The practical reading: fasting slows protein loss compared with what would otherwise happen, and having more body fat to draw on makes that sparing more effective. It does not stop protein loss.

The timeline is mostly a male timeline

Women reach roughly twice the ketone concentration of men at 30 hours of fasting, in a study of ten men and ten women. Fat release and ketone production rise earlier in women, tracking an earlier glucagon rise. Women also route more of their mobilised fat into ketone production than men do, although the same study found no sex difference in how much went into liver triglyceride export.

And yet the studies that define the 24 to 72 hour part of the timeline, the growth hormone work, the metabolic rate work, the protein metabolism work, are almost entirely studies of men. Nine men. Six men. Eight men.

So a fasting timeline is not a description of what happens to a person. It is a description of what happened to a few dozen mostly male volunteers, several decades ago, presented as universal.

Four things to take from this

The changes are continuous, not stepped. Nothing waits for a round number on a clock.

The 12 to 24 hour window is where most of the measurable metabolic shift happens, which is a reasonable argument for the popular 16 to 20 hour eating windows, if not for the specific numbers.

Individual variation is large and systematic, and sex is one of the biggest sources of it. Your own experience of a 16-hour fast may legitimately differ from the person next to you.

And the most repeated numbers in this field, the 1,300% growth hormone figure and the hour at which autophagy supposedly begins, are the two least supported things in it.