Your caffeine half-life is not the average one

The standard line is that caffeine has a half-life of about five hours, so a coffee after 2pm is a bad idea. That number is a population average, and the variation around it is unusually large — large enough that the advice is close to meaningless without knowing where you sit.

Where the variation comes from

Caffeine is metabolised almost entirely by the liver enzyme CYP1A2. Common genetic variants in that gene divide people roughly into fast and slow metabolisers, and the difference in clearance rate between the extremes is substantial — several-fold rather than marginal.

Reported half-lives across healthy adults range from around two hours to more than eight. A slow metaboliser drinking coffee at 3pm still has a meaningful fraction of it circulating at midnight. A fast metaboliser can drink an espresso after dinner and sleep.

Neither is unusual. Both people are told the same thing.

The factors that shift it

Hormonal contraception. Combined oral contraceptives inhibit CYP1A2 and roughly double caffeine's half-life. This is a large effect and it is essentially never mentioned when someone is prescribed the pill. If your tolerance changed when you started or stopped contraception, that is why.

Pregnancy. Clearance slows dramatically, with half-life extending substantially by the third trimester. This is part of why intake guidance in pregnancy is lower.

Smoking induces CYP1A2 and speeds clearance considerably. Quitting effectively increases the potency of the same amount of caffeine, which catches people out.

Some medications inhibit the enzyme, including certain antibiotics and fluvoxamine.

Liver function and age shift it more modestly.

What caffeine actually does to sleep

Caffeine blocks adenosine receptors. Adenosine accumulates through the day and is the main driver of sleep pressure — the physical sense of being tired. Blocking it does not remove the accumulated adenosine; it masks it. When the caffeine clears, the pressure is still there, which is where the afternoon crash comes from.

Two effects on sleep matter. It delays sleep onset, which people notice. And it reduces deep sleep even when it does not delay onset, which people do not notice. This is the important one: you can fall asleep fine after an evening coffee and still get worse-quality sleep, then conclude caffeine does not affect you.

Controlled studies have found measurable sleep disruption from caffeine taken six hours before bed in people who reported no subjective effect.

The anxiety side

Caffeine is anxiogenic at sufficient doses. It raises heart rate, and it produces physiological sensations — palpitations, jitteriness — that are easily interpreted as anxiety, which then generates real anxiety in a feedback loop.

Slow metabolisers are more prone to this, as are people with panic disorder. If you are anxious and drinking four coffees a day, reducing is one of the cheapest experiments available. What actually helps anxiety covers the wider list.

There is also a compounding effect worth knowing about: poor sleep raises next-day anxiety and drives higher caffeine intake, which worsens sleep. It is a stable loop and it is difficult to see from inside.

Working out your own number

There is no need to genotype yourself. Run the experiment.

Set a cut-off time and hold it for two weeks. Start with 2pm. Track sleep onset, wake-ups and next-day energy.

Then move it. Try noon for two weeks, or 4pm. Compare against your record, not your impression.

Watch for the confounds. The luteal phase independently worsens sleep, so a fortnight that happens to fall there will look worse regardless — sleep in the luteal phase. Alcohol will swamp the caffeine signal entirely — alcohol, sleep and next-day anxiety. Comparing across whole cycles rather than across two weeks avoids most of this.

Expect withdrawal if you cut back. Headache, fatigue and low mood for a few days to a week is a withdrawal effect, not evidence that you need it. Tapering rather than stopping abruptly avoids most of it.

A note on tolerance

Tolerance develops to some of caffeine's effects — the alertness boost in habitual consumers is substantially the reversal of withdrawal rather than a net gain. Tolerance to the sleep effects is much less complete. So the common belief that "caffeine doesn't affect my sleep because I'm used to it" is usually half right: you have adapted to the subjective stimulation, not to the disruption of deep sleep.

Why this needs a record rather than a rule

The whole point is that the population answer does not apply to you, and the only way to find your answer is to compare your own nights against your own intake — across enough of them to see past the noise from your cycle, alcohol, stress and everything else.

Naked is built to hold that comparison: caffeine, sleep, energy and mood logged together against your cycle, with the relationships surfaced rather than guessed at.

People who run this experiment properly often find their cut-off is much earlier than they assumed, and that moving it is the highest-return sleep change available to them.