Caffeine and Cortisol: The Timing Problem

Almost nobody has a caffeine quantity problem. Most people have a caffeine placement problem — a stimulant dropped on top of the body's own morning stimulant, and a second dose late enough that it is still circulating at bedtime. Both are free to fix and neither requires giving anything up. What follows is the mechanism, what the evidence actually supports, and where the popular advice runs ahead of the data.

The curve you are interfering with

Cortisol is not a stress hormone that happens to have a daily rhythm. It is a rhythm hormone that also responds to stress, and the rhythm is the more important half. Levels fall through the evening to a nadir around midnight, begin climbing in the small hours, and then rise sharply in the thirty to forty-five minutes after you wake — the cortisol awakening response (Stalder et al., Endocr Rev 2025).

That morning surge is not a malfunction. It mobilises glucose, raises blood pressure and alertness, and is the physiological event that turns a sleeping animal into a functioning one. A flattened curve — low peak, high trough — is the pattern associated with poor sleep, chronic stress and a general sense of running on empty while never quite settling. The shape carries more information than any single reading, which is why a lone morning cortisol tells you far less than people assume (morning cortisol versus the saliva curve).

Two things follow. Anything that blunts the peak degrades the signal. And anything that raises cortisol at night degrades it from the other end.

Caffeine does two separate things

These get conflated constantly, and they have different consequences.

It blocks adenosine. Adenosine accumulates in the brain across the waking day and is the molecular substrate of sleep pressure — the longer you have been awake, the more of it there is, and the sleepier you feel. Caffeine occupies adenosine receptors without activating them (Fredholm, Pharmacol Rev 1999). This is the alerting effect, and it is important to be clear about what it does not do: adenosine keeps accumulating behind the blockade. Caffeine hides sleep pressure. It does not discharge it, which is why the arrival of tiredness when it wears off feels abrupt.

It stimulates cortisol. Caffeine acts on the hypothalamic-pituitary-adrenal axis and raises cortisol secretion, dose-dependently and across the waking day (Lovallo et al., Psychosom Med 2005). Partial tolerance develops in habitual users, but it is not complete, and it returns after even a short abstinence — which is why the first coffee back after a week off hits so differently.

The alerting effect is what you are paying for. The cortisol effect is the one that interacts with the clock.

The morning placement, and how strong the evidence really is

The default pattern is alarm, then coffee within minutes. That places a cortisol-raising drug directly on top of the largest natural cortisol event of the day, at the one time when you least need help being alert.

The mechanistic argument for waiting is straightforward. Adding stimulation to a peak that is already maximal buys little; spending the same caffeine ninety minutes later, when the curve is descending, buys more. It also means the natural surge is doing the work it evolved to do rather than being pre-empted, and it separates "I am awake" from "I have had coffee" — which matters, because the second is a dependency and the first is not.

Here is the honest part, which is usually left out. The claim that delaying coffee by ninety minutes measurably preserves the cortisol awakening response over time is mechanistically reasonable and not well tested. The individual components are solid: the awakening response is real, caffeine raises cortisol, and combining two cortisol stimuli produces more cortisol than either alone. The specific long-term claim built on top of them is an inference. What is genuinely well supported is the sleep half of this article, and the reason to try the delay is that it costs nothing, takes a week to evaluate, and the read-out — whether you feel awake before the first cup — is something you can assess yourself.

The end of the day is where the real damage is

Caffeine's half-life is around five to six hours in most adults, which means a 4 PM coffee is still half-present at 10 PM and a meaningful fraction present at midnight. Genetics matter here more than almost anywhere else in nutrition: variation in the liver enzyme that clears caffeine produces fast and slow metabolisers, and a slow metaboliser's afternoon espresso behaves like an early-evening one.

People who insist caffeine does not affect their sleep are usually telling the truth about how it feels and the wrong thing about what happens. Objective measurement shows degraded sleep architecture — less deep sleep, more fragmentation — at doses and timings that subjects did not perceive as disruptive (Drake et al., J Clin Sleep Med 2013). Falling asleep is not the endpoint that matters. Deep sleep is where growth hormone is released and where the bulk of physical recovery happens, and losing it silently is worse than losing sleep loudly, because there is no symptom prompting you to change anything (sleep architecture and hormones).

This is also where the loop closes. Degraded sleep raises next-day fatigue, which raises caffeine intake, which further degrades sleep. The intake is a symptom of the problem it is maintaining.

The practical order

1. Wake, hydrate, get outdoor light on the eyes within the first half hour. Light is the stronger circadian input, and it does the job people are asking coffee to do.
2. Wait 60-90 minutes before the first caffeine. Cheap to test, easy to reverse.
3. Take the bulk of the day's intake in the mid-morning window.
4. Cap total intake around 400 mg for most adults — lower if you are anxious, pregnant, or on medication that competes for the same clearance pathway.
5. Stop by 2 PM, or by noon if you are a slow metaboliser or a light sleeper. If you only change one thing, change this one; it is the best-evidenced item on the list.
6. Treat a rising requirement as data. Needing more caffeine to reach the same baseline is usually a sleep debt or an HPA axis problem asking to be looked at, not a tolerance problem asking for a bigger dose (the HPA axis in detail).

What to expect when you change it

Moving the first cup later is usually uneventful — a slightly duller first hour for two or three days, then nothing. Moving the cut-off earlier is the one that produces a real adjustment. Expect three to seven days of afternoon flatness and, in heavier users, a headache for the first two, as adenosine receptor density readjusts.

What changes first is sleep latency and night-time waking, usually within the first week. Deep sleep and the sense of waking rested take longer, and they are the point. What does not change is anything driven by an untreated cause — sleep apnoea, alcohol in the evening, a bedroom that is too warm, or chronic stress will all sit exactly where they were (sleep, cortisol and recovery). Caffeine timing is a real lever, and a small one relative to those.

Bottom line

Caffeine blocks adenosine and raises cortisol, and the second of those interacts with a daily rhythm that matters. Taking it during the natural morning surge spends it at the least useful moment and blurs the line between being awake and being caffeinated; taking it after mid-afternoon degrades deep sleep whether or not you notice. Delay the first cup by an hour or so, keep total intake moderate, and stop early enough that it has cleared before bed. The morning-delay case is mechanistic and worth a week's trial; the afternoon cut-off is the one the objective data supports. And if you need caffeine to function rather than to perform, the problem is upstream of the coffee.

Educational content, not medical advice. Laboratory interpretation and any treatment decision are made by a licensed physician after individual evaluation. Individual results vary.