Cortisol does not sit flat overnight and switch on at the alarm. Output falls to its lowest point in the hours around midnight, holds there briefly, then climbs through the second half of the night so that concentrations are already high by the time you wake.² By 3am the axis is usually in the early part of that climb, which is the first reason the hour is so consistent across people who have never spoken to each other about it.

Underneath the daily shape sits a faster one. Cortisol is released in discrete pulses roughly every 60 to 120 minutes, generated below the hypothalamus, and the amplitude of those pulses varies across the 24 hours.² Through the night the pulses are small and infrequent, and that relative quiet appears to be part of what allows sleep to stay consolidated. When pulses arrive at night that should not be there, sleep becomes easier to break and awakenings run longer.³

This is the part worth holding onto. The problem is rarely that cortisol is high. It is that the shape is wrong for the hour. A number sitting inside its reference range at 8am tells you almost nothing about what the same axis was doing at 3am, which is why a normal morning cortisol and a broken night sit together so often on the same person.

When researchers sampled plasma ACTH and cortisol serially across 24 hours in young people with chronic insomnia and in matched controls, total output was higher in the insomnia group and the number of secretory pulses per 24 hours was higher too. The circadian shape itself was preserved. The elevation was concentrated in the evening and the first half of the night, and the people with the most disturbed sleep secreted the most cortisol.⁴

The authors read that as a state of central nervous system arousal rather than a consequence of lost sleep, and the distinction has practical weight. If the driver were sleep loss, cortisol would be flat or low. It was neither. The axis was running warm before the person got into bed.

A second study looked at where the correlation actually sits. In men with severe chronic primary insomnia, both evening and overnight cortisol were raised, and evening cortisol tracked the number of nocturnal awakenings. That same relationship held in the healthy controls.⁵ The evening reading predicted the night, in people with a diagnosis and in people without one.

This is where the popular version of the story goes wrong. Wake at 3am, spit in a tube, find cortisol elevated, conclude that cortisol woke you. The physiology does not support that conclusion on its own. Spontaneous awakening is itself followed by a brief rise in ACTH and cortisol, measured in healthy men whose sleep had been left completely undisturbed.⁶ The act of waking moves the number you are trying to interpret.

That is the practical case for collecting a curve at fixed times across an ordinary day and night, instead of sampling in the moment something goes wrong. What you want to see is the shape the axis holds when nothing is being asked of it.

There is also a loop running underneath all of this. After a night of partial or total sleep restriction, cortisol the following evening ran 37 and 45 per cent higher than baseline, and the usual evening shutdown was delayed by at least an hour.⁷ Set that beside the finding that evening cortisol tracks the number of awakenings,⁵ and one broken night has a plausible mechanism for biasing the next one. Run that for a few weeks and an occasional 3am wake becomes the thing that happens every night.

Start with glucose. In healthy volunteers, a controlled fall in blood glucose during sleep woke almost everyone it was applied to, about as readily in early sleep as in late sleep. What changed was the hormonal answer. The rises in adrenaline, noradrenaline, ACTH, cortisol and growth hormone were distinctly weaker in the second half of the night than in the first.⁸ The counter-regulatory response is quieter late, which is one reason the late-night version of this tends to present as a proper awakening rather than a stir.

Alcohol is the one people discount. Across every dose studied, it shortens the time it takes to fall asleep, consolidates the first half of the night, and increases disruption in the second half.⁹ Someone who falls asleep quickly after two glasses and wakes at 3am is describing that curve rather than an unusual one.

The vasomotor version has a physiology of its own. In women in the menopausal transition and after it, roughly half of nocturnal hot flushes came with an arousal or awakening, and those were accompanied by a heart-rate rise of about 20 per cent and systolic and diastolic pressure rises of around 6 and 5 mmHg, sustained for several minutes.¹⁰ The heat is the trigger. The cardiovascular surge is why the person is properly awake afterwards rather than drifting back down.

Breathing is the one to rule out first when the description fits. Obstructive events cluster in REM sleep, and REM is weighted towards the second half of the night. In just over a thousand consecutive adults referred for a sleep study, REM-related sleep-disordered breathing accounted for 13.5 to 36.7 per cent of cases depending on the definition used, and was more common in women and in younger people.¹¹ Waking with a gasp, a dry mouth, or unrefreshed after eight hours is worth a sleep study before it is worth a cortisol test.

A single morning blood cortisol gives you one point on a curve that has at least four clinically interesting parts: the waking response, the daytime slope, the evening level and the overnight floor. The collection that answers a 3am question is a multi-point sample across an ordinary day, timed relative to waking rather than to the clock, with the first samples inside the 30 to 45 minutes after waking so the awakening response is captured properly.¹² Both saliva and dried urine do this well; the method comparison sits in the cortisol testing guide, and the practical version is in how to test cortisol at home.

Two things are worth running alongside it, because the table above is a differential and not a ranking. Fasting glucose with fasting insulin, plus overnight glucose where a continuous monitor is available, covers the metabolic driver. A sleep study covers the breathing one, and should come first whenever the description includes snoring, gasping, or waking unrefreshed after a full night.

What you are reading in the result is not one flagged value. It is whether the evening sits high, whether the waking response is blunted or exaggerated, and whether the daytime slope has gone flat. Those three read together point at which lever is worth pulling, and they give you a baseline to re-measure against in six months. That last part is where the value compounds, and it is the same logic that applies to deep sleep and recovery: one reading is a snapshot, and the trajectory is the finding.

If the bloods have already come back clean and the nights are still broken, that combination is common enough to have its own pattern, and it is covered in why you are exhausted with normal blood tests.