How long does caffeine work? Half-life, sleep and withdrawal

We have been roasting, brewing, training, and selling coffee for years, but surprisingly know very little about the substance that makes it interesting for most people: caffeine. I wanted to change that and spoke with Dr. Carolin Reichert, deputy head of the Center for Chronobiology in Basel. She has been researching for almost twenty years how caffeine and adenosine control our sleep, and also works as a psychotherapist.

Her first answer to almost every question was: "that's individual." Her second: "there's no evidence for that." So what caffeine actually does to us, what might be perceived truths, and where Carolin Reichert attests a certain recklessness to us all, I clarify in this article.

What caffeine does in the brain: blocking adenosine, not providing energy

Caffeine doesn't give us energy; it blocks the signal that tells us we're tired.

Adenosine is responsible for this. Reichert clarifies a common misconception: "Adenosine is not a hormone, we call it a neuromodulator or sleep factor." It accumulates throughout the day and docks onto receptors. Caffeine sits on the same receptors.

So the tiredness is still there, it's just not reported to us. This also explains why it comes back once the caffeine is broken down.

In research, two receptor subtypes are relevant, and they do different things. The A2A receptor is responsible for the waking effect: in mice without the A2A receptor, caffeine no longer had a waking effect, but in mice without the A1 receptor it did. The A1 receptor, on the other hand, is responsible for building up and breaking down what is called sleep pressure. At normal doses, caffeine only partially occupies adenosine receptors, estimated at 25 to 50 percent, and not selectively.

Half-life: on average five hours, individually two to ten

Caffeine is absorbed through the gastrointestinal tract. About 99 percent is absorbed within 45 minutes, with peak levels occurring between 15 and 120 minutes depending on the person.

After that, the breakdown begins, which also varies individually. The average among healthy adults is about five hours, with a range of 1.5 to 9.5 hours. An independent measurement series found half-lives between 2.27 and 9.87 hours at 5 mg per kilogram of body weight.

Half-life means: after this time, half is gone. With an espresso containing 70 mg at 4 p.m., with a five-hour half-life there are still 35 mg in the blood at 9 p.m., and around 17 mg at 2 a.m. In someone who breaks it down slowly, the same espresso can still be at over 40 mg at 2 a.m.

Institute of Medicine 2001; Temple et al. 2017; Evans et al., StatPearls 2024; Blanchard & Sawers 1983.

Who breaks it down quickly and who slowly: a gene is involved

An enzyme called CYP1A2 handles the breakdown almost entirely, but at different speeds. Between individual people, a sixty-fold difference in breakdown speed has been measured.

A gene variant called rs762551 is partly responsible. Those with two A variants break it down quickly, those with one or two C variants break it down slowly. In the population, it's distributed roughly like this: 40 percent fast, 50 percent mixed, 10 percent slow.

This is why experience reports in comment sections vary so much. Reichert: "Some people barely react to caffeine at all and others very strongly." And further: "The mechanism of action, that receptors are blocked from this adenosine system, will be the same in all people. But as soon as we look at the effect and see how strongly it affects sleep, performance, mood, motivation, it's very different."

And what about age?

Here we were unable to confirm a widespread claim. "Caffeine sensitivity increases with age" appears on many health portals without study references. However, this has been refuted: a comparative study with young and older men concluded that "most aspects of the pharmacokinetic behavior of caffeine are very similar in young and elderly men". And the only age-comparative study on sleep EEG we could find shows the opposite: after 200 mg of caffeine, changes in REM sleep were only significant in young men, not in middle-aged ones.

How long does caffeine affect sleep? The numbers

Here we have hard values. A 2023 meta-analysis has calculated what caffeine does to sleep:

Gardiner, Weakley, Burke et al. 2023, Sleep Medicine Reviews 69:101764.

Interesting is the associated calculation: for each additional hour of distance to bedtime, sleep loss decreases by 2.8 minutes; for each milligram of dose it increases by 0.2 minutes.

Deep sleep is the most important value. 45 minutes less sleep is noticeable, eleven minutes less deep sleep is not. Both are measurable, but only one is felt.

The last coffee: dose beats time of day

The question of how many hours before sleep the last coffee should be is the most common on this topic. The best answer comes from a controlled study from 2025, in which 23 healthy men received 100 mg and 400 mg of caffeine respectively 12, 8, and 4 hours before bedtime, double-blind and placebo-controlled, with sleep measurement at home.

Gardiner, Weakley, Burke et al. 2025, Sleep 48(4):zsae230. 23 healthy men, 25.3 ± 5.0 years.

The study's original conclusion: "A 100 mg dose of caffeine can be consumed up to 4 hours prior to bedtime, but 400 mg may negatively impact sleep when consumed within 12 hours of bedtime."

In translation, this means: a single coffee in the late afternoon is different from four spread throughout the day. Time of day alone tells us little; the amount decides. And 400 mg twelve hours before sleep means that high morning consumption affects the night.

From the 2023 meta-analysis, rough guidelines can be derived for a bedtime of 10 p.m.: black tea with 47 mg needs no time limit, coffee with 107 mg at least 8.8 hours away, a pre-workout product with 217.5 mg at least 13.2 hours.

Can you fall asleep with caffeine? Yes

Reichert warns against dramatization. In studies where participants had to stay awake for 40 hours and were given caffeine, "the participants can still sleep afterwards".

A 2024 study backed this with numbers. 41 people stayed awake for 38 hours, some receiving about 350 mg of caffeine twice, with the last dose 6.5 hours before recovery sleep. Sleep came in both groups; in both conditions sleep time was over 500 minutes. Caffeine reduced the recovery effect by about 30 minutes and deep sleep by about 36 minutes. Sleep pressure wins, but sleep becomes shallower.

Caffeine withdrawal: headaches in half, from 100 mg per day

This is the part where I'm personally affected. Once I drank only decaffeinated coffee for an entire December, my Decaf December, and the headaches of the first week were real. They also happen when I don't have coffee in the morning and then notice a slight headache a few hours later.

The reference work on caffeine withdrawal is from 2004 and provides clear numbers:

Juliano & Griffiths 2004, Psychopharmacology 176(1):1–29.

The 100 mg is the most important number in this table. One cup of coffee a day is enough to trigger withdrawal symptoms if stopped. In addition to the headache, documented are: fatigue, reduced energy, reduced alertness, sleepiness, depressed mood, concentration difficulties, irritability, and a feeling of not being clear-headed.

Reichert names the same key symptoms and puts them in context: "This withdrawal syndrome exists, and it also shows that we have adapted to this chronic consumption that may have happened before, that the receptor system has adapted to it."

Is there caffeine dependence? Not according to WHO

We often hear that coffee is addictive. We would say that coffee is a habit. And the technical definition of addiction is quite different.

The ICD-11 of the WHO has a separate block for caffeine, 6C48 "Disorders due to use of caffeine". It includes: harmful use as an episode, harmful pattern of use, caffeine intoxication, caffeine withdrawal, and caffeine-induced disorders such as an anxiety disorder. What is not in there is a code for dependence. Alcohol, opioids, and stimulants each have one; caffeine does not.

Reichert explains why: for dependence, there must be, among other things, loss of control, a strong craving for the substance, and reduction of family, hobbies, and work. "All of that doesn't happen with caffeine." And further: "But what does happen is that you might consume more even though you know it has negative consequences. Harmful use exists for caffeine, and these physiological adaptations happen too."

Tolerance: completely after 18 days, but not for sleep

The body gets used to caffeine, and measurably so. In a study with three times 300 mg daily over 18 consecutive days, caffeine produced no subjective effects at the end in the caffeine group, but it did in the placebo group.

This doesn't apply to sleep in the same way. A lab study with three times 150 mg over ten days found no differences in sleep duration, sleep architecture, and sleep efficiency between caffeine, withdrawal, and placebo. However, a large analysis with data from the UK Biobank and home sleep measurements found that four or more caffeine-containing drinks per day shortened sleep time, with a range of 11 to 229 minutes depending on the model. This range is too wide to cite as a single value, but the direction is clear.

There's another less comfortable interpretation, and it's well documented. In habitual users who were briefly abstinent, the wakefulness after caffeine administration exceeded neither that of placebo-treated non-consumers nor that of former users. In plain language: the morning coffee doesn't make many people more awake than normal; it restores normalcy.

The cortisol myth: waiting 90 minutes does nothing

I confronted Reichert with a rule or rather idea that I hear repeatedly at least in coffee and social media circles: wait 60 to 90 minutes after waking up because otherwise the cortisol rise will be disrupted and an afternoon slump will follow.

What was actually measured even flips the story around. A study with 96 people gave 300 or 600 mg of caffeine daily over five days and then tested the cortisol response. After abstinence, caffeine triggered a clear cortisol rise. After five days of regular use, the response to the morning dose was completely gone, in the original "abolished". In the afternoon, a residual response remained.

So whoever drinks coffee daily has no cortisol response in the morning that caffeine could disturb. The rule addresses a group in which the described mechanism doesn't exist.

Reichert has a second observation on this, and I think it's the more important one: "Of course it's clear that the later I consume caffeine, the longer it stays in the body and the later it's broken down." That might be the whole mechanism, which then doesn't need a cortisol story.

Placebo and nocebo work both ways

Some of caffeine's effect happens in the mind, and demonstrably in both directions. I had asked whether expectation only enhances the waking effect. Reichert's answer: "No, also nocebo effects."

Specifically: "If we expect that we can't sleep because we've just had caffeine, that can indeed keep us a bit more awake." She puts it psychologically: "The more tense we are, the more threatened we feel, the less likely we are to fall asleep."

This explains a good part of the experience reports where a single espresso after lunch supposedly ruined the night. And it has a practical consequence: sleep hygiene rules can become a problem in themselves. Reichert on this: "As soon as you focus too much on these rules and structure your whole life around them, it becomes a stressor."

How much caffeine is in it?

The numbers vary more widely than most tables admit. This especially applies to espresso.

Conversions from EFSA 2015, Table 1 (mg/L). Espresso measurement values: Crozier, Stalmach & Lean 2012, Food & Function 3(1):30–33. Capsule values: Desbrow, Hall & Irwin 2019, Nutrition and Health 25(1):3–7. Instant coffee: USDA Legacy 2018.

The espresso row is the most interesting in the whole article. Among 20 examined establishments, the highest shot was more than six times higher than the lowest, and three of 20 espressos exceeded the daily recommendation for pregnant women of 200 mg alone. Bean, grind size, brew ratio, and cup volume decide more than the drink category. Anyone wanting to know how much caffeine is in their own espresso will find the calculation in the article on the double espresso.

Regarding energy drinks, Reichert made a correction during our conversation. First, on the amount: "In terms of caffeine, it contains less caffeine, often about eighty milligrams for a small can." Red Bull's manufacturer gives 80 mg for the 250 ml can as well. A regular cup of coffee is above that. Second, on the target group: "It's not that teenagers consume more than adults. Worldwide or in Europe at least, teenagers consume roughly half as much as adults."

How much is too much? The EFSA values

EFSA NDA Panel 2015, Scientific Opinion on the safety of caffeine, EFSA Journal 13(5):4102. The BfR confirms the same values in its statement No. 037/2026.

For children and adolescents, there's a caveat that's important. The value of 3 mg per kilogram is not from a child study but a transfer from adult data. The EFSA itself writes that "the information available is insufficient to derive a safe caffeine intake". Reichert says the same in different words: "We actually have so little data to even establish a guideline on what is tolerable and what is not."

On caffeine intoxication, because that's often asked. It is a diagnosis in both classification systems; the DSM-5 cites typical trigger doses well over 250 mg plus at least five of twelve symptoms, the ICD-11 cites doses over one gram per day. Reichert says: "But I've never heard of a caffeine-induced death." In forensic literature, 92 deaths are documented over 59 years, with life-threatening amounts from about 10 to 14 grams. All documented cases involve tablets, powder, or shots, not coffee. 10 grams equals about 100 cups at 100 mg per cup in a short time.

Reichert explicitly names one risk: "The problem is more in combination with alcohol, because there are simply interactions there that are unfavorable."

Caffeine and sports: 3 to 6 mg per kilogram, 60 minutes before

Here the evidence base is better than for almost anything else. The 2021 position paper of the International Society of Sports Nutrition cites specific values: caffeine improves performance "consistently" at 3 to 6 mg per kilogram of body weight. The minimally effective dose may be 2 mg per kilogram. Very high doses around 9 mg per kilogram provide no additional benefit, just more side effects. The usual intake time is 60 minutes before exertion.

At 75 kilograms body weight, 3 to 6 mg per kilogram is 225 to 450 mg. That's a lot of coffee, and the upper half exceeds the EFSA daily recommendation.

The effects are real but not spectacular. In time trial performance, an analysis of 46 randomized studies showed a plus of 2.22 percent; in strength, a standardized effect size of 0.18; in muscular endurance, 0.30. Caffeine is most effective for aerobic endurance. However, two of the 46 studies showed worsened time trial performance, and five showed lower average power.

Caffeine is not forbidden in sports. It has been on the World Anti-Doping Agency's monitoring program for years, and for 2026 it continues: "The substances listed in the monitoring program are fundamentally not prohibited but are monitored."

A question from our daily lives

During cupping we slurp coffee and spit it out, a hundred times with the spoon. Do you absorb caffeine that way? Reichert: through the oral mucosa it's fundamentally possible, "for example through caffeine chewing gum". But: "When we drink coffee, most of it is certainly absorbed through the gastrointestinal tract." She knows of no studies on dwell times in the mouth. In chewing gum research, it takes 44 to 80 minutes of active chewing to reach effective levels. With spitting out, absorption is likely minimal, and the tired feeling after a long cupping has other reasons.

And if caffeine is the problem?

Reichert doesn't consider caffeine dangerous: "I don't think it's some kind of devil substance. We would know if it really had serious effects." Her wish is different.

That's our position too, and we have an answer to it on the shelf. For those who want the coffee and don't need the caffeine, today there's a real choice. According to our numbers, a cup of decaffeinated coffee contains 1.4 to 3.6 mg of caffeine, compared to 80 to 138 mg in filter coffee. The 2025 study showed that 100 mg four hours before sleep had no measurable effect. At three milligrams, you don't need to think about timing anymore.

How decaffeinated coffee is made, what four processes exist, and what remains measurably in the cup, is explained in the article on decaffeination processes. How it tastes, why it ages faster and extracts differently, in the main article on decaffeinated coffee. You'll find our Decafs in the Decaf collection.

By the way: caffeine itself is a bitter substance. If you want to know where bitterness in coffee comes from and what part of it isn't tied to caffeine, you'll find that in our article on bitterness in coffee.

To listen