Caffeine Cutoff Time: How Late Is Too Late for Sleep

Most caffeine advice is a single number: no coffee after 2 pm. The trials that actually timed caffeine against bedtime tell a messier story. One found that 400 mg taken six hours before bed still cut sleep. Another found that a single morning dose measurably changed sleep sixteen hours later. A third found that habitual drinkers who stopped eight hours before bed showed no change in sleep structure on polysomnography at all.

The answers disagree because “how late” is the wrong question. What matters is the dose, how fast the body clears it, and how sensitive the sleeper is. All three can be estimated, and the estimate is more useful than any fixed clock time.

What happened when trials moved the dose

The best-known timing study gave 400 mg of caffeine, roughly four cups’ worth, at bedtime, three hours before bed, or six hours before bed, against placebo. All three timings disturbed sleep on both self-report and a validated portable sleep monitor used at home (p < 0.05 for each), and the authors judged the loss of total sleep time in the six-hour condition large enough to matter. Their recommendation was to avoid substantial caffeine for at least six hours before bed. Six hours is a floor, not a cutoff: it was the earliest timing tested, and 400 mg, which the authors describe as a moderate dose, is still about four cups taken at once.

Push the dose earlier and the effect does not vanish. Nine healthy men took 200 mg at 7:10 in the morning. By 11 pm their saliva caffeine had fallen from a peak of 17 µmol/L to 3 µmol/L, yet they slept less, and less efficiently, than on placebo, and their non-REM sleep showed less power in the slowest (0.25–0.5 Hz) EEG band. A 200 mg morning dose, roughly two cups, was still detectable in that night’s sleep.

At realistic drink doses the picture is graded rather than all-or-nothing. In a randomised five-way crossover, thirty healthy volunteers drank tea (37.5 or 75 mg caffeine per serving), coffee (75 or 150 mg) or water four times a day, with the last serving at 11 pm, and sleep was scored by questionnaire and wrist actigraph. Falling asleep, total sleep and rated sleep quality all worsened in proportion to caffeine dose (p < 0.001 for each), and tea, with half the caffeine of the coffee servings, was less likely to disrupt sleep.

The pooled numbers

A 2023 meta-analysis in Sleep Medicine Reviews pooled 24 studies. On average, caffeine took 45 minutes off total sleep time and 7 percent off sleep efficiency, lengthened the time to fall asleep by 9 minutes, added 12 minutes of wake after sleep onset, and shifted sleep toward lighter stages: 6.1 minutes more stage-1 sleep and 11.4 minutes less deep sleep. The authors then estimated how far before bed a given dose would need to be finished to avoid losing total sleep time: about 8.8 hours for a 107 mg cup of coffee (250 mL) and about 13.2 hours for a 217.5 mg pre-workout serving. Those are pooled estimates, not cutoffs tested in any single trial.

An earlier systematic review of epidemiological studies and randomised trials reached the same qualitative conclusion: caffeine typically delayed sleep onset, shortened and fragmented sleep, and left people rating their sleep worse, with less slow-wave sleep and more stage-1 sleep, wakefulness and arousals. It added two caveats worth keeping: older people appear to react more strongly to caffeine at night than younger ones, and most of the evidence comes from male participants in Western countries.

Why half-life beats clock time

Caffeine’s effect on sleep depends on how much is still circulating at bedtime, and that depends on its elimination half-life, the time for blood levels to fall by half. That number is not the same for everyone.

• In a saliva study of adult men and non-pregnant women, the mean half-life was 3.4 hours, with a range of 2 to 5 hours. In 57 pregnant women it averaged 8.3 hours and ranged from 3 to 16, and in most cases returned to normal within a month of delivery.
• Smokers cleared caffeine markedly faster than non-smokers: a half-life of 3.5 hours versus 6.0, in 13 people per group.
• Women who had used low-dose oestrogen oral contraceptives for more than three months had a half-life of 7.9 hours versus 5.4 in nine matched controls.
• A pharmacology review points to CYP1A2, the enzyme that metabolises about 95 percent of ingested caffeine and varies genetically between people, and lists the factors known to shift clearance: age, sex, hormonal status, liver disease, obesity, smoking and what people eat.

The arithmetic is simple. Take 200 mg at 3 pm and go to bed at 11 pm, eight hours later. With a 3-hour half-life, about 30 mg is left. With a 5-hour half-life, about 65 mg. With an 8-hour half-life, 100 mg, half the dose, is still on board. The same coffee at the same time is a different exposure in different people, which is why a blanket “2 pm rule” suits some and fails others.

Regular drinkers, sensitive sleepers and age

Daily use changes the picture. Twenty young men who habitually drank caffeine took 150 mg three times a day for 10 days, with each night’s sleep starting 8 hours after the last dose. On polysomnography, total sleep time, sleep latency and sleep architecture came out the same whether they were on caffeine, placebo or withdrawal, and so did their ratings of sleep quality. The one signal that moved was EEG power in the spindle range (12–16 Hz), lower in both the caffeine and the withdrawal conditions than on placebo, which the authors read as a possible early sign of overnight withdrawal. For good sleepers who consume caffeine daily and stop eight hours before bed, then, the measurable cost may be small. The participants were young, male and good sleepers, so this does not generalise to people who already sleep badly.

Sensitivity is partly genetic. In one study, a variant in the adenosine A2A receptor gene (ADORA2A c.1083T>C) was distributed differently between adults who called themselves caffeine-sensitive and those who did not, and the genotype shaped the objective, sleep-EEG response to caffeine as well as the subjective one. Habitual intake went with worse self-rated sleep quality in the self-described sensitive group but not in the insensitive group.

Age matters too. When 22 adults averaging 23 years old and 24 adults averaging 52 took 200 mg, 400 mg or placebo in a double-blind crossover, caffeine lengthened sleep latency and reduced total sleep and efficiency, and at 400 mg those effects were more pronounced in the middle-aged group. Both age groups lost slow-wave and REM sleep in proportion to dose.

Setting a personal cutoff

Putting the trials together gives a procedure rather than a rule.

1. Estimate the dose. The meta-analysis used 107 mg for a 250 mL coffee; the crossover trial used 37.5–75 mg per cup of tea and 75–150 mg per cup of coffee. The pre-workout serving in the meta-analysis was 217.5 mg, about twice a home cup.
2. Pick a half-life bracket. Roughly 3–6 hours for a non-smoking adult with no other factors; longer on oral contraceptives or in pregnancy; shorter in smokers. Liver disease, obesity and diet also shift it.
3. Start from the evidence-based floor. Six hours before bed is the minimum the 400 mg trial’s authors recommend. The meta-analysis estimate puts one coffee at about 9 hours and a pre-workout dose at about 13. With an 11 pm bedtime that is 2 pm for a single coffee and mid-morning for anything larger.
4. Move it and measure. One approach is to shift the cutoff two hours earlier for a couple of weeks and compare. Time to fall asleep, total sleep and rated sleep quality were the outcomes that moved in the timing trials, and all three can be tracked with a simple diary.

Caffeine timing sits alongside schedule, light and alcohol in the sleep hygiene basics, and the same dose-and-timing logic runs through The Sleep Optimization Protocol. Two related points: the L-theanine and caffeine trials looked at focus and alertness, not at how long caffeine stays in the body, so the pairing says nothing about a cutoff. And reaching for a sleep aid to override a late coffee is the wrong order of operations; the melatonin dosing guide covers why low doses and timing, not a bigger dose, are where melatonin’s evidence lies.

If you are pregnant, take medication, or have a diagnosed sleep, heart or liver condition, talk to a clinician about caffeine rather than experimenting with timing alone.

Bottom line

Caffeine taken six hours before bed still cut sleep in a placebo-controlled trial, a 200 mg morning dose altered that night’s sleep EEG, and a 24-study meta-analysis found an average of 45 minutes less sleep and 11 minutes less deep sleep, with an estimated cutoff of about 8.8 hours for one coffee. Half-life runs from about 2 hours to 8 or more between people, so the useful cutoff is personal: estimate the dose, estimate your clearance, use the trial authors’ six-hour minimum as the floor, and move earlier if sleep onset and night wakings improve. Habitual drinkers who stopped eight hours before bed showed little measurable change in one 10-day study; middle-aged and older adults, oral-contraceptive users, pregnant women and self-described caffeine-sensitive people are the groups the evidence suggests need a longer gap.