Caffeine and sleep: what 24 studies found
A 2023 review of 24 controlled studies of caffeine and the night after. What it found on sleep and timing, and why its averages are not your own.
By Lomba 4 min read
In 2023 a review in Sleep Medicine Reviews reviewed 24 controlled studies of caffeine and the night that follows. In each study the same adults slept after caffeine and without it, and the two nights were compared.
Here is what it found, what it says it cannot tell you, and the one idea we think matters most.
What the study did
- Who
- Healthy adults aged 18 to 65, mostly low to moderate caffeine users. Twelve of the 24 studies enrolled only men.
- What was compared
- Each person's night after a measured dose of caffeine against their own night without it. Where the studies reported it, the last dose came anywhere from bedtime to about 16 hours before.
- What was measured
- Total sleep time, time to fall asleep, time awake during the night and sleep stages. Eighteen studies used polysomnography, the sleep lab's standard, three used activity monitors and three used diaries or questionnaires.
- Who paid
- The paper names no funder. Its authors declare no conflicts of interest.
Four takeaways
Sleep was about 45 minutes shorter.
Across 340 participants, total sleep time was 45.3 minutes lower after caffeine (95% interval 29.0 to 61.5). Sleep efficiency, the share of time in bed spent asleep, was 7.0 points lower (4.0 to 10.0). Time in bed was similar in all but four studies, so the authors put this down to disrupted sleep.
Falling asleep took longer and sleep was lighter.
Time to fall asleep was 9.1 minutes longer (3.8 to 14.4) and time awake during the night 11.8 minutes longer (2.5 to 21.0). Deep sleep was 11.4 minutes shorter (4.3 to 18.5) and the lightest stage 6.1 minutes longer (2.3 to 9.9). REM sleep showed no clear change.
Dose and timing showed up for total sleep only.
Each extra hour between the last dose and bedtime went with 2.8 minutes less sleep lost (0.4 to 5.2). From that model the authors estimate a cut-off of 8.8 hours before bedtime for a cup of coffee (107 mg) and 13.2 hours for a serving of pre-workout supplement (217.5 mg). They found none for a cup of black tea (47 mg). A cut-off here is the point where the model's sleep loss stops being statistically significant.
The paper says where it stops.
Its results may not carry to adolescents, adults over 65, people who never have caffeine or heavy users. Most studies gave caffeine for a single day, and the two that gave it for more than a week found no significant impairment to sleep. The authors say regular use may blunt the effect, though tolerance has not been clearly established in humans. Five of the 24 were rated at high risk of bias.
pooled change95% intervalinterval crosses zero
Each row sets the night after caffeine against the same people’s night without it, with no change at 0. Rows in minutes share one scale, so their lengths compare. Sleep efficiency is in points, on a scale of its own.
- Half-lives from 2 to 10 hours, the range the review cites for healthy adults.
- The review’s estimates: 8.8 hours for its coffee dose (107 mg) and 13.2 hours for its pre-workout dose (217.5 mg).
8.8 hours after the cup, a half-life of 5.0 hours leaves 30% of the dose. A half-life of 2 hours leaves 5%, and one of 10 hours 54%. That gap is the review’s estimate for its coffee dose (107 mg). Calculated from one assumption, that the amount left halves once every half-life. Not a measurement of any person.
The main takeaway
The 45 minutes and the 8.8 hours come from group averages. People clear caffeine at very different speeds.
The authors cite a half-life of two to 10 hours across healthy adults. They call their cut-off times "a starting point", because "it is likely that the effect of caffeine on sleep will differ across individuals". A separate analysis of 141 published caffeine studies found that smoking consistently went with faster clearance and oral contraceptive use with slower.
So a pooled cut-off describes a group. The timing that matters for one person is that person's own, and only their own nights can show it.
What Lomba does with that
Lomba has Coffee as a habit you log with a tap. It records the grams of ground coffee for the day and shows an estimate of the caffeine. Then it lines up the mornings after a day with coffee against your own mornings after a day without, on hours asleep, sleep performance, recovery, heart rate variability and resting heart rate from a wearable. Each comparison gets an answer, such as Held up, Set aside, Still collecting or No clear gap.
How it works follows two examples through the checks behind those answers.
It has limits, and we would rather say them:
- The day, not the hour. Lomba compares a day's amount, not the time of the last cup, so it cannot check the review's cut-off times. A member can add an item of their own, such as a late coffee, and it is compared the same way.
- Every day leaves nothing to compare. If coffee is logged every day, Lomba has no days without it, and says so.
- A pattern, not a cause. The review itself notes that people often have caffeine after a short night. Lomba counts the morning before, and what it finds is still a pattern in your own days, not proof of cause, and not medical advice.
Sources
- Gardiner C and others. The effect of caffeine on subsequent sleep: A systematic review and meta-analysis. Sleep Medicine Reviews, 2023.
- Grzegorzewski J and others. Pharmacokinetics of Caffeine: A Systematic Analysis of Reported Data for Application in Metabolic Phenotyping and Liver Function Testing. Frontiers in Pharmacology, 2022.
Both papers are open access.