The short version
- Plasma caffeine half-lives ranged from 2.7 to 9.9 hours across ten healthy young men given an identical 5mg/kg dose, a nearly fourfold difference between individuals.
- Oral caffeine is essentially completely bioavailable at 108.3%, and peaks in plasma about 30 minutes after drinking it.
- A 400mg dose of caffeine taken 0, 3 or 6 hours before habitual bedtime each significantly disturbed sleep compared with placebo, measured in participants' own homes.
- A double-espresso-equivalent dose taken 3 hours before bedtime delayed the circadian melatonin rhythm by about 40 minutes, nearly half the delay caused by three hours of 3,000 lux evening light.
- Caffeine is cleared mainly by CYP1A2, and a systematic analysis of 141 pharmacokinetic publications identifies smoking, oral contraceptives and CYP1A2-metabolised drugs as sources of large inter-individual variation.
Most people know caffeine has a half-life of about five hours the way they know a marathon is 26 miles: roughly right, and not much use on its own.
The number that matters is not the average. It is the spread.
The measurement
Blanchard and Sawers gave ten healthy men aged 18 to 30 a 5mg/kg dose of caffeine as either an oral solution or an intravenous infusion, a week apart, in randomised crossover, then sampled plasma over 24 hours.
Three findings. Absorption was very fast, peaking at 29.8 ± 8.1 minutes. Oral bioavailability was essentially complete, 108.3%, meaning swallowing it is as good as injecting it. And the plasma half-lives varied from 2.7 to 9.9 hours — a nearly fourfold range across ten similar young men.
Run the arithmetic on the slow end. Take 200mg at 2pm with a half-life of eight hours and you still have roughly 150mg circulating at 8pm and about 100mg at bedtime. On the fast end, the same coffee is down to trace amounts by dinner. Same drink, same person's height and weight, completely different evening.
What that does to sleep, measured rather than assumed
Sleep hygiene advice about evening caffeine was, for a long time, advice without a trial behind it. Drake and colleagues fixed that. They gave 400mg of caffeine at 0, 3 and 6 hours before habitual bedtime, versus placebo, and measured sleep in participants' own homes both by self-report and with a validated portable monitor.
All three timings significantly disturbed sleep relative to placebo (p<0.05 for all). The authors' conclusion is that the magnitude of reduction in total sleep time means caffeine taken six hours before bedtime has important disruptive effects on sleep, and that this provides empirical support for refraining from substantial caffeine for a minimum of six hours before bed.
Six hours before an 11pm bedtime is 5pm. For a lot of people that is the second coffee, not the first.
It does not just shorten sleep, it moves the clock
This is the finding that surprises people. In a double-blind, placebo-controlled, roughly 49-day within-subject study, consuming a caffeine dose equivalent to a double espresso three hours before habitual bedtime induced a ~40-minute phase delay of the circadian melatonin rhythm. For scale, the authors note that was nearly half the delay produced by three hours of bright light at around 3,000 lux beginning at bedtime.
Caffeine is not only blocking the sleepiness signal. It is nudging the timing of your entire night later, which then makes tomorrow morning harder, which is usually answered with more caffeine.
Why your number is not your friend's
Caffeine is cleared almost entirely by the liver, mostly through CYP1A2. A systematic analysis compiling caffeine pharmacokinetic data from 141 publications identifies large inter-individual variability as the open problem in the field, and specifically examines how caffeine pharmacokinetics are altered by smoking and by oral contraceptives, and how drugs metabolised via CYP1A2 interact with it.
The practical translation, without overclaiming numbers the abstract does not give: smoking speeds clearance up, so quitting can make your usual coffee suddenly feel like too much. Hormonal contraception slows it down. Several common medications compete for the same enzyme. Pregnancy slows it dramatically. None of this is visible on a label, which is why the only reliable instrument is your own sleep.
The rule we would actually give
- Set a cut-off time rather than a cup limit. Eight hours before bed is a conservative start; six is the evidence-supported minimum.
- Change the timing first and hold the dose steady for two weeks, so you can tell which variable did the work.
- If mornings are brutal and evenings feel wired, suspect the phase delay rather than the dose.
- Judge it on sleep duration over a fortnight, not on how you feel the next morning — the morning after is exactly when caffeine's own withdrawal is loudest.
One honest caveat. Fixing your caffeine timing will improve your sleep, and better sleep will flatten out the worst of an afternoon energy dip. It will not do anything at all if the tiredness has a medical cause. If you have already moved your cut-off earlier for a month and you are still flattened by 3pm, stop adjusting your coffee and go and get your thyroid and a basic panel checked.
Good questions
How late is too late for coffee?
Six hours before bed is the evidence-supported minimum, and eight is a safer starting point. A 400mg dose taken six hours before bedtime still significantly disturbed sleep versus placebo in a home-based trial. If your bedtime is 11pm, that puts the cut-off between 3pm and 5pm, which is earlier than most habits allow for.
Why does caffeine not seem to affect my sleep?
It may genuinely clear fast in you, since half-lives vary nearly fourfold between individuals, or you may be measuring the wrong thing. People are poor judges of their own sleep quality, and the trial that measured it objectively found disruption even where participants underestimated it. Track total sleep time over two weeks rather than trusting the impression.
Does caffeine mess with your body clock?
Yes, separately from keeping you awake. A dose equal to a double espresso three hours before bed shifted the melatonin rhythm about 40 minutes later, roughly half the effect of three hours of bright evening light. That delay makes the next morning harder, which is often answered with more caffeine, which is how the pattern sustains itself.
Does the pill change how I handle caffeine?
It slows clearance. Oral contraceptives are one of the specific factors identified as altering caffeine pharmacokinetics through CYP1A2, alongside smoking, pregnancy and several prescription drugs. If your tolerance has changed after starting or stopping hormonal contraception, that is a real pharmacological effect rather than imagination.
If I quit smoking, why does my usual coffee suddenly feel like too much?
Because smoking induces the enzyme that clears caffeine, so stopping slows your clearance and the same cup produces higher blood levels for longer. It is one of the better-documented influences on caffeine pharmacokinetics. Cutting your caffeine dose during a quit attempt is a reasonable thing to plan for rather than be surprised by.
Is switching to decaf in the afternoon worth it?
If you like the ritual, yes, and it costs you nothing. Decaf is not caffeine-free but the dose is small enough to sit well below the levels used in the sleep trials. The point of an afternoon cut-off is the pharmacology, not the mug, so anything that keeps the habit and drops the dose does the job.
Sources
- Blanchard J, Sawers SJ. The absolute bioavailability of caffeine in man. Eur J Clin Pharmacol, 1983. View study
- Drake C, Roehrs T, Shambroom J, Roth T. Caffeine effects on sleep taken 0, 3, or 6 hours before going to bed. J Clin Sleep Med, 2013. View study
- Burke TM, Markwald RR, McHill AW, Chinoy ED, Snider JA, et al. Effects of caffeine on the human circadian clock in vivo and in vitro. Sci Transl Med, 2015. View study
- Grzegorzewski J, Bartsch F, Köller A, König M. Pharmacokinetics of caffeine: a systematic analysis of reported data for application in metabolic phenotyping and liver function testing. Front Pharmacol, 2021. View study
These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.
