Quitting caffeine

Most people do not quit caffeine; they quit quitting, on day two, when the headache arrives on schedule. Knowing the schedule is most of the fight.

01What withdrawal is

Daily caffeine leaves the brain compensating: it upregulates the adenosine receptors caffeine blocks, as covered in tolerance and cycling. Remove the caffeine and the compensation is suddenly unopposed: adenosine signalling surges through an enlarged receptor field, blood vessels in the head dilate, and the system that was propped up daily sags below its true baseline. That is the headache, the fog and the flatness. It is not weakness of character. It is receptor arithmetic reversing.

02The timeline

The definitive review of over 60 studies maps the course precisely: onset 12 to 24 hours after the last dose, peak intensity at 20 to 51 hours, total duration two to nine days.¹ Headache affects about half of regular users; fatigue, difficulty concentrating and irritability are the other core symptoms, and the syndrome is documented from habitual intakes as low as 100 mg a day. In practice: day one is deceptively fine, day two is the wall, day four is visibly downhill, and by the second week it is over for almost everyone.

03Taper beats cold turkey

Withdrawal severity tracks the size of the drop, which is why the evidence favours reducing gradually. In the one controlled trial of structured caffeine reduction, 77 percent of participants following a progressive taper hit their reduction target, against 23 percent of controls.² A taper trades one hard week for a few mild ones, and it keeps you functional at work while it happens, which is usually the difference between quitting and quitting quitting.

04A working taper

Trials have not compared taper schedules, so the structure below is reasonable practice built on the pharmacology, and we label it as such. Cut the latest dose first, since it does the most sleep damage per milligram, per the timing protocol. Then reduce total intake by roughly a quarter to a half each week, using real numbers rather than feelings: an espresso is about 80 mg, a 200 ml filter coffee about 90 mg, black tea about 50 mg, an energy drink about 80 mg per 250 ml.³ Swapping cups for half-decaf or tea keeps the ritual while the dose falls, and the ritual is half the habit. Expect the mild version of the symptoms at each step down, on the same schedule as above, just quieter.

05What you get back

The return is concrete. Regular caffeine measurably degrades sleep: the meta-analysis puts the average cost at 45 minutes of total sleep and a 7 percent hit to efficiency,⁴ and laboratory work shows even daytime-only regular dosing reshapes sleep architecture.⁵ That comes back. So does an honest baseline: part of what a daily dose delivers is relief of the previous day's withdrawal, so the morning lift you lose was partly a refund you were paying for.⁶ After the reset, caffeine works properly again at lower doses, and whether you return at all is covered in tolerance and cycling. One medical note: pregnancy has its own limit, 200 mg a day per the European safety assessment.³

06The research horizon

The taper logic this protocol uses is now being tested formally. A randomised trial at Johns Hopkins put 100 adults through a six week guided gradual reduction, delivered almost entirely by manual with under five minutes of staff contact, tracking caffeine intake alongside anxiety, sleep, and withdrawal symptoms.⁷ The mechanism behind tapering also keeps firming up: chronic caffeine upregulates adenosine receptors, which is why abrupt stopping hits hard and stepped reduction gives the receptor population time to renormalise. And the newest dosing data adds a useful floor: the 2025 polysomnography work found 100 mg to be essentially sleep-neutral even four hours before bed,⁸ which means the last step of a taper is standing on measured ground rather than superstition. Most people quitting caffeine are running a protocol from a 2004 memory. The 2020s version is gentler, slower, and better instrumented, and it works with the receptors instead of against them.

07Where the system sits

How we handle evidence. The timeline and taper evidence are cited; the taper schedule itself is labelled as practice, because schedule trials do not exist. Educational content, not medical advice.