Blood Sugar Over 24 Hours: A Day You Can Rearrange

Our previous model followed one meal for four hours. That is the right frame for understanding a spike, and the wrong frame for almost every decision you actually make, because you do not eat one meal. You arrange a day.

And arrangement turns out to matter on its own. The same food, the same person, the same total carbohydrate, moved around the clock, produces measurably different days — because insulin sensitivity is not a constant. It has a daily rhythm, the liver has its own schedule, each meal changes what the next one does, and last night's sleep is still setting the terms.

The model below runs midnight to midnight. Start with the standard day, then change one thing: the dashed line stays behind as the day you would have had.

A day is one curve, not four

The first thing worth noticing is what the model does not have: any code that connects one meal to the next. It runs a single continuous simulation from midnight to midnight, and the connections appear on their own.

Lunch lands on a body that is still carrying insulin from breakfast. Insulin action outlasts the glucose that triggered it, and the liver, having been told to stop producing glucose two hours ago, has not yet started again. So the same meal produces a smaller rise than it would have on an empty morning. This is the second-meal effect, first described by Staub and Traugott in the 1920s, and it is why a day cannot be understood as a stack of independent meals.

It also cuts the other way. Meals placed close together stack: the second arrives while the first is still being absorbed, and the curve never gets back down between them. Drag two meals within about two hours of each other in the model and you can watch two moderate excursions merge into one long one.

The same dinner is a different dinner at 21:30

The ribbon under the chart is the part most people have never seen. Insulin sensitivity follows a daily rhythm: highest in the morning, declining through the afternoon, lowest in the late evening and small hours, then recovering before you wake. The beta cell keeps roughly the same schedule, releasing insulin more sluggishly at night for the same glucose rise.

Two systems falling together is why the effect is larger than either alone. An identical meal eaten late meets both less capacity to clear glucose and a slower response to the arrival of it. Controlled studies giving the same meal early and late typically report excursions 10 to 25 percent higher in the evening, and the model sits inside that range.

Press last meal ninety minutes earlier and watch what moves. The peak comes down, the minutes above 140 fall, and the total food has not changed by a gram. This is the cheapest intervention on the page: not eating less, just earlier.

Why morning can be the highest reading of the day

Set the model to prediabetes and look at the small hours. Glucose starts climbing around 03:00, with no meal anywhere near it, and by waking it is well above where it was at midnight. Nothing was eaten. This is the dawn phenomenon.

Before you wake, cortisol and growth hormone rise on a fixed schedule and instruct the liver to release stored glucose, so that you come round with fuel already in circulation rather than having to eat before you can function. It is a feature. In someone whose insulin still restrains the liver, the released glucose is cleared about as fast as it appears and the rise is a handful of mg/dL — the healthy setting in the model barely moves.

Where that restraint has weakened, the same signal meets a liver that will not stop, and the climb is large. This is the mechanism behind one of the most demoralising experiences in early type 2 diabetes: eating nothing after dinner, sleeping eight hours, and waking to the highest number of the day. It is not a failure of discipline. It is an endocrine schedule.

One short night changes the next day

Pull the sleep slider down to five hours and change nothing else. Every curve in the day gets taller. The food is identical, the timing is identical, and the day is worse — because sleep restriction reduces insulin sensitivity, by roughly 15 to 25 percent in experimental studies of a few short nights in healthy young adults.

That is a large enough effect to move someone with normal glucose tolerance towards the impaired range for a day or two, and it recovers with normal sleep. It is worth sitting with the implication: sleep is a glucose intervention, and it acts on the same lever — how well insulin works — that everything else on this page is fighting over.

The model understates it, if anything. It applies the penalty evenly across the day, when the real effect is largest in the morning, and it does not attempt the other half of the problem: short sleep also changes appetite, and the food you choose after a bad night is usually not the food in the standard preset.

Where the walks go

Press ten minutes after every meal. Thirty minutes of walking, split three ways and placed where the glucose actually is, does more to the day than the same thirty minutes in one block at a convenient hour — because contraction-mediated glucose disposal only helps while there is glucose to dispose of.

The mechanism is the second door covered in the post-meal model: muscle contraction brings GLUT4 transporters to the cell surface without needing insulin at all. Across a day it compounds, because each walk also leaves the transporters partly in place and the glycogen partly empty, so the next meal lands on a body that is still primed.

Try the walks on the healthy setting and then on the insulin-resistant one. On the healthy day almost nothing happens, because there was never a problem to solve. The effect scales with how little margin you have.

What eating early does, and what it does not

This is where the popular version and the model part company, and the model is worth trusting here because you can watch the mechanism.

Moving a late meal earlier reliably helps: better sensitivity, better secretion, lower peak. But compressing a day into a short window is a different operation, and it has a cost the enthusiasm rarely mentions — the meals stack. Four sittings squeezed into eight hours means each one arrives while the previous is still being absorbed, and some of what you gained by eating early is handed back. Build it in the model and watch both effects fight.

Which is roughly what the trial evidence shows. Time-restricted eating produces weight loss mainly when it reduces how much people eat; when calories are matched, its independent effects are smaller and less consistent than the headlines suggest. The part with the better support is timing — earlier is better than later — rather than the width of the window.

The day adds up to a number

Every arrangement you try changes the mean glucose in the readouts, and mean glucose is the quantity a blood test eventually measures. Glucose sticks to haemoglobin in proportion to how much of it there is and how long it is there; three months later, that accumulated stickiness is your HbA1c.

So the model can put a day on the same scale as the test — with an assumption stated plainly on the panel: if every day looked like this one. No real three months is one day repeated ninety times. And the conversion is a population average: HbA1c also depends on how long your red cells live, which is why two people with the same mean glucose can differ by several tenths of a percent.

And it comes with a lesson you will notice immediately when you press the buttons: the average barely moves. Three walks and an earlier dinner take the insulin-resistant day from an implied 6.0% to 5.9%, while the time spent above 140 falls by more than half and the highest point drops by more than 10 mg/dL. Arrangement changes the shape of a day far more than its mean.

That is not a disappointment, it is the division of labour. What moves the average is how much carbohydrate arrives and how well insulin works — the food and the phenotype. What arrangement buys you is a flatter day at the same average: fewer hours above the spike threshold, lower peaks, and a smaller insulin bill for getting there. Both matter, and the model lets you see which lever is doing what.

Frequently asked questions

The honest limits of this model

The engine is the same one as the post-meal model — a two-compartment minimal model with biphasic secretion, insulin-suppressible hepatic output and two disposal routes — extended with a circadian sensitivity rhythm, a dawn rise in hepatic output, a sleep-debt penalty, renal glucose loss at high concentrations, and as many meals and walks as you care to place. The day is run twice and only the second pass is shown, so midnight reflects the night before rather than an arbitrary starting point.

What that means for the numbers: the healthy day lands near a mean of 98 mg/dL and spends effectively all of it in range, which is about right. The impaired phenotypes are tuned to reproduce their diagnostic curve shapes, and that makes their whole-day averages sit at the pessimistic end of what continuous monitors actually record — a prediabetes preset eating 260 g of mostly refined carbohydrate with no activity is a deliberately unflattering day, not a typical one.

Absent entirely: food order within a meal, alcohol, caffeine, acute stress, illness, menstrual-cycle effects, medication of any kind, and the gut microbiome — which in the largest studies of identical meals explains a good share of why two people respond differently. Body weight is fixed. Type 1 diabetes is not modelled at all, and would behave differently in ways that matter.

Use it to see how the pieces interact and which levers are worth pulling. What your own day looks like takes a continuous monitor, and what to do about it takes a clinician who knows the rest of your picture.