No caffeine after 2 p.m.

What is behind it

Caffeine acts as an antagonist at adenosine receptors. Adenosine is the key substance of homeostatic sleep pressure: it accumulates during wakefulness and produces tiredness. Caffeine occupies the receptors at which adenosine would have to dock and cancels the tiredness signal — without eliminating the underlying lack of sleep.

The relationship has been quantified in humans: in a randomized, double-blind crossover study, 21 healthy young men took 160 mg of caffeine or placebo at their usual bedtime and slept 4.5 hours later. Caffeine dose-dependently reduced delta activity in the EEG and heart rate.

Why the interval counts and not the time of day

Caffeine is broken down almost entirely in the liver, with CYP1A2 activity accounting for about 95 % of caffeine clearance. EFSA gives a plasma half-life of about 4 hours with a range of about 2 to 8 hours — breakdown can therefore differ by a factor of four between two people.

This explains why a fixed time of day is the wrong unit. The supported figures are intervals before going to bed: at least 8.8 hours for coffee with 107 mg per 250 ml, at least 13.2 hours for 217.5 mg, at least 6 hours as the lower limit. Which time of day this translates into depends on when someone goes to bed.

How much sleep it actually costs

The larger meta-analysis of 24 studies arrives at a total sleep time shortened by 45 min, a sleep efficiency reduced by 7 % and a sleep onset latency lengthened by 9 min. In addition, the structure shifts: light sleep increased by 6.1 min, deep sleep decreased by 11.4 min.

An independent, more recent meta-analysis of 22 polysomnographic crossover studies with 956 participants confirms this with somewhat smaller values. No significant effect was seen on the REM share.

What the studies show

Meta-analysis with the interval rule

Systematic review and meta-analysis of 24 included studies. Pooled effects: total sleep time −45 min, sleep efficiency −7 %, sleep onset latency +9 min, wake time after sleep onset +12 min, light sleep +6.1 min, deep sleep −11.4 min. The actual contribution is the derived interval rule: coffee with 107 mg per 250 ml at least 8.8 hours before going to bed, a supplement with 217.5 mg at least 13.2 hours beforehand.

Polysomnography only, independent confirmation

22 controlled crossover studies with a combined 956 participants, overnight polysomnography after acute administration of caffeine or placebo in healthy adults. In the random-effects model: total sleep time −34.67 min, sleep efficiency −4.74 %, deep sleep share −1.01 %, sleep onset latency +8.35 min, no significant effect on the REM share. Meta-regressions, however, found no linear relationship with age or dose, and the differences between subgroups were not significant.

Even six hours beforehand it still disrupts sleep

A fixed dose of 400 mg of caffeine, administered 0, 3 and 6 hours before the usual bedtime, was compared with placebo; sleep at home was recorded by self-report and objectively with a validated portable sleep monitor. All three time points led to significant sleep disruption. From this, the authors derive the recommendation to refrain from substantial amounts at least 6 hours before going to bed. For context: 400 mg at once is a strong stimulus.

The counter-finding in habitual consumers

20 young male habitual consumers went through a double-blind, randomized crossover study with three ten-day conditions: caffeine at 3 × 150 mg daily, withdrawal, and placebo throughout. After nine days, an eight-hour sleep episode was recorded by EEG, starting 8 hours after the last intake. Neither total sleep time nor sleep onset latency nor sleep architecture nor subjective sleep quality differed between the conditions; only the power density in the sigma band of 12–16 Hz was reduced.

Where the data stop

Two widely cited numbers do not stand up to scrutiny. According to the official figure, the half-life is about 4 hours with a range of about 2 to 8 hours — the often-cited 5 to 6 hours lie within this range, but above the central value. And the time of 2 p.m. appears in no study: the only quantitative figure is an interval before going to bed, and a fixed time of day presupposes a fixed bedtime.

The most important open point is individuality. In habitual consumers with intake in the morning and afternoon, no impairment of sleep structure could be demonstrated at all — but only in 20 young men, all good sleepers; for poor sleepers, women, older people and occasional consumers, this has not been tested. Genetics is also less clear-cut than often portrayed: the documented modulator of sleep sensitivity is ADORA2A, while CYP1A2 determines the rate of breakdown. In a study of 6,112 caffeine-consuming children, neither ADORA2A nor CYP1A variants were associated with sleep quality or sleep duration.

How to do it

What can be derived from the studies is an interval instead of a time of day: at least 8.8 hours before going to bed for coffee with 107 mg per 250 ml, at least 13.2 hours for a pre-workout serving with 217.5 mg, and 6 hours as the minimum recommendation. For this, you need to know your own dose: the BfR gives 90 mg for 200 ml of filter coffee and 80 mg for 60 ml of espresso. The condition under which nothing showed up: 3 × 150 mg daily over ten days, last intake 8 hours before the sleep episode. Anyone who wants to cut down should taper off gradually.

Safety

EFSA considers single doses of up to 200 mg of caffeine from all sources and daily amounts of up to 400 mg safe for adults, with the exception of pregnant women, for whom up to 200 mg per day applies; the BfR gives the same values, also for breastfeeding women. In pregnancy, the half-life lengthens to 6 to 16 hours; oral contraceptives, some medications and severe liver disease also slow down breakdown. For children and adolescents, according to EFSA, the information is not sufficient to derive a safe intake level. Abrupt abstinence in heavy habitual consumers is not a good idea: withdrawal was accompanied by increased sleepiness and more lapses of attention. Anyone with cardiac arrhythmias, an anxiety disorder, high blood pressure or reflux complaints should discuss the amount of caffeine with a physician.

Sources

• Gardiner et al., Sleep Med Rev 2023
• Chang et al., Sleep Med 2025
• Drake et al., J Clin Sleep Med 2013
• Weibel et al., Sci Rep 2021
• Rétey et al., Clin Pharmacol Ther 2007
• Jessel et al., Genes 2023
• EFSA NDA Panel, Scientific Opinion on the safety of caffeine, EFSA Journal 2015
• BfR, questions and answers on caffeine and caffeinated foods (German)