Caffeine and Sleep: What the Dose-and-Timing Trial Actually Found

"No caffeine after 2pm" is common advice, but it's a blanket rule standing in for something more specific: caffeine's half-life, your dose, and how close to bedtime you actually take it. Two people who both "stop caffeine by early afternoon" can have completely different sleep outcomes if one is having a single cup and the other is stacking coffee, an energy drink, and a pre-workout supplement in the same window. A recent randomized, placebo-controlled crossover trial tested the dose-and-timing question properly instead of guessing, and a 2023 meta-analysis of the wider evidence base backs up what it found — with a third study adding a wrinkle about whether being a habitual coffee drinker actually protects you.

The trial

23 healthy young men (mean age 25.3) with low-to-moderate habitual caffeine intake were given either 100mg (a typical cup of coffee) or 400mg (a high dose, roughly 3-4 cups) of caffeine at 12, 8, or 4 hours before bedtime, with polysomnography — the gold-standard, electrode-based sleep study used in sleep labs — measuring what actually happened to their sleep that night, compared with a placebo condition in the same participants.¹ Because it's a crossover design, each person acted as their own control across doses and timings, which is what lets the researchers attribute the differences to the caffeine itself rather than to who happened to sleep well that week. The all-male, young, healthy sample is the limitation to carry forward: it says nothing directly about women, older adults, shift workers, or people whose caffeine metabolism runs unusually fast or slow — a genuine source of variation we've covered in why caffeine hits some people harder than others.

What the doses actually did

The 100mg dose caused no significant sleep disruption at any timing tested, including just 4 hours before bed. That's worth sitting with: a normal cup of coffee, taken in the mid-afternoon, did not show up as a measurable problem on polysomnography in this trial. The 400mg dose was a different story: taken 4 hours before bed, it cut total sleep time by 50.6 minutes (p < 0.001), reduced sleep efficiency by 9.5% (p < 0.001), delayed sleep onset by 14.2 minutes (p = 0.013), and cut deep (slow-wave) sleep by 29.7 minutes (p < 0.001).

Now the part we got wrong first time round, and the part most write-ups of this trial get wrong too. At 8 and 12 hours before bed, 400mg of caffeine did not significantly reduce total sleep time. The point estimates looked meaningful — 28.7 minutes less sleep at 8 hours, 30.0 minutes less at 12 hours — but the p-values were 0.076 and 0.060 respectively, both above the conventional threshold. A 30-minute point estimate that fails to reach significance in 23 people is a result you report as inconclusive, not as a finding. What did reach significance further out was narrower and more specific: sleep efficiency was down 6.9% at 8 hours (p = 0.001), and deep sleep was reduced at both 8 hours (−15.3 min, p = 0.016) and 12 hours (−20.6 min, p = 0.001). So a large dose with breakfast may well be costing you slow-wave sleep. Whether it is costing you total sleep time, this trial genuinely does not establish. Deep sleep is the stage most of the marketing targets, and what does and doesn't move it is worth reading in the science of deep sleep.

The pattern is consistent: it's the dose that determines whether caffeine reaches your sleep at all, and the timing determines how much of that dose is still active when you lie down. Neither variable on its own tells the full story — a small dose taken very late and a large dose taken relatively early can land in roughly the same place, which is why "just watch the clock" and "just watch the milligrams" both miss half the picture.

How much of a buffer do you actually need?

A single trial is useful but narrow — 23 young men, one lab, one protocol. A 2023 systematic review and meta-analysis in Sleep Medicine Reviews pooled 24 studies and put a number on the buffer question directly, rather than testing one dose at a few fixed timepoints.² Across the pooled data, caffeine reduced total sleep time by 45.3 minutes (95% CI, 29.0–61.5) and sleep efficiency by 7.0% (95% CI, 4.0–10.0), lengthened sleep onset by 9.1 minutes (95% CI, 3.8–14.4), added 11.8 minutes of wake after sleep onset (95% CI, 2.5–21.0), and shifted architecture toward light sleep: N1 up 6.1 minutes, N3 (deep) sleep down 11.4 minutes.

Be careful how much weight that carries, because it is more studies rather than a lot more people: those 24 studies total only a few hundred participants between them, depending on which outcome you look at. The authors are explicit about the rest. Five of the included studies were rated high risk of bias — three lacked adequate washout periods, two had incomplete sleep staging. Several used wrist activity monitors rather than polysomnography, and actigraphy tends to overestimate sleep duration and underestimate wake time. Only five studies dosed caffeine across multiple days, so this says little about sustained intake. The meta-regression behind the cut-off times used only the amount and timing of the final dose, which probably understates what several doses across a day do. And genetic and metabolic differences in caffeine sensitivity were not accounted for anywhere in the review. The authors also state plainly that the findings may not generalise to adolescents, adults over 65, caffeine-naive people, or heavy habitual consumers — which, awkwardly, is a large share of the people who go looking for this advice.

That gap between the two cutoffs is the same dose-dependence the crossover trial found, just expressed as a buffer window instead of a fixed clock time. A modest coffee needs less runway than a stacked pre-workout or energy-drink dose — which is exactly why "cut off by 2pm" works for some people and fails for others drinking the same amount of caffeine but at a higher concentration or as a later top-up. It also explains why generic sleep-hygiene checklists that just say "avoid caffeine in the afternoon" tend to underperform for anyone whose real intake is closer to 400mg than 100mg: the checklist advice was calibrated for the wrong dose.

Does being a regular coffee drinker protect you?

It's a reasonable hypothesis: habitual users build tolerance to caffeine's jittery, alerting effects, so maybe their sleep is less sensitive to an evening dose too. The data don't fully support that. In a double-blind crossover study of 20 habitual male caffeine consumers (average intake around 478mg per day) who were given regular caffeine doses (3 x 150mg daily) across 10 days and compared against placebo and withdrawal conditions, caffeine still delayed the onset of REM sleep — REM latency stretched to roughly 79 minutes versus about 54 minutes on placebo — and subjective sleep quality was worse, even though total sleep time and overall sleep architecture did not significantly differ between the three conditions — including sleep efficiency and the proportions of N1, N2, N3 and REM.³ That null result belongs in the summary as much as the REM finding does: what shifted was the timing of REM sleep, not how much sleep anyone got. If the distinction between the stages isn't familiar, NREM versus REM sleep is the primer. In other words: regular intake shifted the timing of REM sleep in people who consume caffeine every day and describe themselves as unaffected by it — but it did not shorten their sleep. This study shares the same narrowness as the trial above: 20 young healthy men, one circadian phase, no control for adenosine-receptor gene variants, and no women at all.

Worth separating two things that get run together here. Tolerance to the felt effects of caffeine is real and well documented. Tolerance to its effects on sleep architecture is a different claim, and this study is one small piece of evidence against it rather than a settled answer — it is a single 20-person study, and it found no change in total sleep time, so nobody should read it as proof that habitual drinkers sleep less.

What this means practically

The honest answer isn't a single cutoff time — it's dose-dependent, and the newer pooled data narrows the range further. A normal coffee-sized dose (around 100mg) appears safe for sleep even taken in the afternoon, roughly 9 hours or more before bed. A genuinely high dose (400mg — several strong coffees, or a couple of pre-workout servings) measurably disrupts sleep architecture even taken 8-12 hours before bed, and may need something closer to a 13-hour buffer to clear fully. If you're tracking down a sleep problem and drink multiple coffees or a pre-workout in the afternoon, the dose is more likely the culprit than the exact clock time — and assuming your years of coffee-drinking have made you immune isn't well supported either. Supplements are a common blind spot in that total: Cognitive Support does contain caffeine, named on the label inside a 255mg proprietary blend, and because the per-ingredient amounts inside that blend are not disclosed we can't give you its exact milligram figure — we've set out what we can and can't say about it in how much caffeine is actually in Cognitive Support. It's a morning product for exactly this reason.

Practically, that suggests a more useful self-audit than "did I have coffee after 2pm": add up your actual milligrams across the whole afternoon and evening — coffee, tea, energy drinks, pre-workout, and any caffeinated supplements — and compare that total against roughly 100mg versus 400mg-plus, not just the clock. Someone having one afternoon coffee is working with a wide safety margin. Someone having an afternoon coffee, a pre-workout, and a diet soda by dinner has likely stacked well past 400mg, and no cutoff time in the early afternoon reliably undoes that.

A moderate morning coffee habit is unlikely to be your sleep problem. A habitual afternoon 400mg+ intake very plausibly is — regardless of exactly when you stop, and regardless of how used to caffeine you feel.

Bottom line

"Cut off caffeine by early afternoon" is reasonable general advice, but it undersells how much dose matters. The evidence — one controlled dose-and-timing trial in 23 young men, a meta-analysis of 24 studies totalling a few hundred people, and a 20-person habitual-user crossover study — points the same direction: at typical coffee doses, timing has a wide margin of safety; at high or stacked doses, that margin narrows, and tolerance doesn't obviously close the gap. It is a consistent signal from a thin evidence base, and both halves of that sentence are true.

If caffeine turns out not to be your problem, the things with better evidence behind them are elsewhere: CBT-I outperforms any supplement we sell, alcohol does more measurable damage to sleep architecture than an afternoon coffee does (what a nightcap actually does), and if the specific problem is waking in the small hours, that has its own guide in how to fall back asleep after waking at 3am.

This article is for informational purposes only and has not been evaluated by the FDA. It is not intended to diagnose, treat, cure, or prevent any disease. Speak with a healthcare provider before making significant changes to your caffeine intake or starting any new supplement, especially if you take medication or have an existing health condition.

References

1. Dose and timing effects of caffeine on subsequent sleep: a randomized clinical crossover trial. SLEEP, 2024. academic.oup.com
2. Gardiner C, Weakley J, Burke LM, et al. The effect of caffeine on subsequent sleep: A systematic review and meta-analysis. Sleep Medicine Reviews, 2023;69:101764. sciencedirect.com
3. Weibel J, Lin YS, Landolt HP, et al. Regular caffeine intake delays REM sleep promotion and attenuates sleep quality in healthy men. Journal of Biological Rhythms, 2021;36(4):384-394. journals.sagepub.com

We cover the rest of the sleep evidence hierarchy — light, timing, CBT-I and where supplementation fits — in our sleep evidence hierarchy guide.