Caffeine and sleep: the cutoff time is only half a rule

Almost every piece of sleep advice contains the same instruction: no caffeine after two. Or three. Or noon, depending who is writing. The hour changes and the confidence does not, which is usually a sign that the number is a convention rather than a finding.

The underlying pharmacology is genuinely well understood. The problem is that it does not produce a single cutoff hour, and converting it into one throws away the part that would actually help you.

Caffeine halves, it does not wear off

Caffeine is eliminated in a first-order process, which means the amount removed is proportional to the amount present. In plain terms: it halves at a roughly constant interval rather than draining at a constant rate.

For a typical healthy adult that interval — the half-life — is around five hours. A 95 mg filter coffee at two in the afternoon leaves about 47 mg at seven, 24 mg at midnight, and 12 mg at five in the morning. Nothing dramatic happens at any point. There is no hour at which the caffeine is gone, only hours at which there is progressively less of it.

This is why “no caffeine after 2 pm” is an odd instruction. It implies a boundary the pharmacology does not have. What it is really encoding is an assumption about your dose, your bedtime and your clearance rate — and it does not tell you any of them, so you cannot tell whether it applies to you.

The six-hour finding, and its limits

The most-cited experiment on timing gave participants 400 mg of caffeine at zero, three, and six hours before their usual bedtime, and compared each against placebo. All three timings produced significant sleep disturbance — including the dose taken a full six hours before bed. ¹ The authors concluded that this supports advice to avoid substantial caffeine for at least six hours before sleeping.

That result gets quoted constantly, and it is worth being precise about what it does and does not establish.

400 mg is a large dose. It is roughly four cups of filter coffee, or a strong pre-workout, taken at once. It is also the figure health authorities cite as a reasonable daily ceiling for healthy adults ⁴ — so the study administered an entire day’s allowance in a single sitting. Whether one afternoon coffee behaves the same way is a reasonable question the study does not answer.

The sample was small. Twelve healthy normal sleepers. ¹ That is enough to demonstrate an effect exists and not enough to tell you its size in the general population, let alone in you specifically.

Six hours is a floor, not a safe point. The finding was that six hours still disrupted sleep, not that six hours is sufficient. If anything it argues the conventional cutoffs are too generous rather than too strict.

A separate study answers the dose objection reasonably well. Giving 200 mg — about two coffees, a far more ordinary evening dose — to twelve younger and twelve middle-aged habitual coffee drinkers, in a double-blind crossover design, lengthened the time taken to fall asleep, reduced sleep efficiency, and cut total sleep duration in both age groups. ³ So the effect is not an artefact of an implausibly large dose. It is smaller samples again, but the direction is consistent.

None of this makes either study wrong. Both are well-designed work that replaced a widely repeated rule of thumb with actual evidence, which is more than most sleep-hygiene advice can claim. They just do not support a universal clock time, and the people quoting them usually turn them into one.

Why your half-life is probably not five hours

Five hours is a population midpoint. The usual range reported in healthy non-smoking adults runs from about three to seven hours, and the modifiers move it further than most people expect:

• Smoking induces the liver enzyme that clears caffeine and roughly halves the half-life. A smoker’s evening coffee behaves very differently from a non-smoker’s.
• Oral contraceptives inhibit that same pathway and commonly double it.
• Pregnancy extends it substantially, particularly later on — which is part of why intake guidance during pregnancy drops to around 200 mg a day. ⁵
• Impaired liver function extends it considerably.
• Genetics, chiefly variation in the CYP1A2 enzyme, accounts for much of the remaining spread between otherwise similar people.

Put two healthy adults side by side at the extremes of the normal range and the same afternoon coffee leaves one of them with a trace at midnight and the other with most of a cup’s worth. Both are normal. Neither is doing anything wrong. A single cutoff hour cannot be correct for both, and the honest version of the advice acknowledges that rather than picking one and presenting it as physiology.

The threshold that does not exist

The obvious next question is how much residual caffeine is acceptable at bedtime. It would be convenient if there were a number below which sleep is unaffected.

There is not one, and it is worth being direct about that rather than inventing a figure. Sensitivity varies severalfold between individuals. Some people fall asleep after an espresso at nine; others lie awake having had tea at four. Habitual intake, age, anxiety levels and the same genetic variation that drives clearance all contribute, and no consensus cutoff emerges from any of it.

What is useful is comparison against yourself. If you know that on nights you slept badly you had roughly 80 mg on board at bedtime, and on nights you slept well you had 20, that is a personally meaningful signal even though it generalises to nobody else. Our caffeine half-life calculator exists to make that number visible — and it deliberately shows a range rather than a single figure, because the range is the honest answer.

Using this without overthinking it

A few things follow from the pharmacology that are more useful than a cutoff hour:

Dose matters as much as timing. Moving a 200 mg pre-workout from 4 pm to 2 pm helps less than halving it. Two half-lives from a large dose still leaves more than one half-life from a small one.

Count everything. Pre-workout, cola, tea, and dark chocolate all contribute, and pre-workout in particular is frequently double a coffee. People who believe they stopped at two often did not.

Watch the strong-coffee problem. A cup of filter coffee is conventionally listed near 95 mg, but real cups range from roughly 70 to 200 depending on beans, grind, brew time and cup size. If your coffee is strong, every estimate you have read — including ours — is low.

Feeling fine is not evidence. Caffeine can measurably affect sleep depth in people who report falling asleep normally. ² Subjective sleepiness is not a reliable readout of what happened once you were asleep.

If you are pregnant, on medication that affects liver metabolism, or managing a condition that does, the general figures here are a poor guide and your clinician’s advice supersedes them.

Bottom line

Caffeine does not expire at a certain hour. It halves at an interval that varies about threefold between healthy people and considerably more once smoking, contraceptives, pregnancy or liver function are involved. A fixed cutoff time is a reasonable default for someone with an average clearance rate and an average dose, and a poor one for everybody else.

The more useful habits are reducing the dose rather than only shifting it, counting the sources you forgot about, and paying attention to what your own bad nights have in common. That is less satisfying than a rule, but it has the advantage of being true.