Free U.S. shipping on orders $45+ · Subscribe & save
Caffeine Sensitivity Check
Seven behavioural markers that track with fast and slow caffeine clearance, turned into a band, an estimated half-life range and a milligram ladder you can test against. No DNA kit, and no claim to be one.
Seven questions
Every answer is filled in already, so the worked example below is live before you touch anything. Nothing you enter leaves your browser.
Used only to turn the cutoff hours into clock times.
Estimated band
Typical
Estimated half-life 4.5 to 6.5 hours
Your answers sit around the middle of the published group means. Nothing you reported pulls hard in either direction, which is the most common result and the least interesting one.
This is a band estimated from behaviour. It is not a genotype. The CYP1A2 variant rs762551 is what splits people into the groups usually labelled fast and slow, and only a genetic test can tell you which one you carry. Behaviour and genotype do not line up neatly, so treat this as a starting hypothesis you then test against the ladder below.
What moved the estimate
- A late-afternoon coffee versus that night of sleepslower
- One ordinary cup, racing heart or shaky handsno effect
- Coffee after dinner and still sleepingslower
- How long one cup seems to lastslower
- Oestrogen-containing oral contraceptiveno effect
- Smoking tobaccono effect
A milligram ladder to test against
Start on the lowest rung for a week and note two things: whether your sleep changed, and whether anything felt off in your chest or hands. Only then consider the next rung. The cutoff column is how many hours before bed that dose should land, given your band, and the clock column applies your bedtime of 11:00pm.
| Rung | Hours before bed | Latest clock time |
|---|---|---|
| 80 mg | 6.0 to 8.7 h | 2:17pm to 4:58pm |
| 100 mg | 7.5 to 10.8 h | 12:12pm to 3:31pm |
| 150 mg | 10.1 to 14.6 h | 8:24am to 12:53pm |
Read the earlier clock time as the cautious one. Where a row says there is no clean window, the arithmetic has put the cutoff before you woke up, which is the honest reading: at that dose and that band, no time of day clears bed. Rungs below 107 mg sit below the lowest dose Gardiner 2023 reported a figure for, so those rows are an extrapolation and the least trustworthy on the table. None of these rungs is a performance dose: the International Society of Sports Nutrition position stand (Guest and colleagues, 2021) puts the ergogenic range at 3 to 6 mg/kg, about 210 to 420 mg for a 70 kg adult. Keep the total from every source, coffee, tea, cola, pre-workout and pouches together, under 400 mg a day, which is the ceiling in the European Food Safety Authority’s 2015 opinion for healthy adults.
What this computes
Three things. A band, either clears fast, typical or clears slowly. An estimated half-life range for that band. And a milligram ladder with a pre-bed cutoff for each rung, expressed both in hours and as a clock time. Six of the seven questions place the band. The seventh, habitual daily intake, deliberately does not touch the band at all. Habit changes how a dose feels, through tolerance, and there is no good reason to think it changes how fast your liver clears the molecule. It only decides which rungs the ladder starts on.
The arithmetic, in full
Gardiner and colleagues (Sleep Medicine Reviews, 2023) pooled 24 studies and reported that a 107 mg coffee should land at least 8.8 hours before bed, and a 217.5 mg serve of pre-workout at least 13.2 hours, if you want to avoid losing total sleep time. Read as half-lives, 8.8 hours is 1.76 half-lives for someone with a population-typical 5 hour half-life. Time to decay to any fixed fraction scales linearly with half-life, and each doubling of dose costs one more half-life. So the table above computes hours before bed as your half-life multiplied by 1.76 plus the base-two logarithm of the dose divided by 107.
That formula is checkable against Gardiner’s own second anchor. At 217.5 mg and a 5 hour half-life it gives 13.9 hours, against their published 13.2, so it reproduces their second reported point to within about 45 minutes. The table prints a range rather than a number because your half-life is a range, not a number.
Where the band edges come from
They bracket published group means rather than being invented. Parsons and Neims (Clinical Pharmacology and Therapeutics, 1978) measured a mean caffeine half-life of 3.5 hours in 13 smokers against 6.0 hours in 13 non-smokers. Abernethy and Todd (European Journal of Clinical Pharmacology, 1985) measured 7.88 hours in oral contraceptive users against 5.37 hours in nine non-smoking, drug-free matched controls. Those four figures span 3.5 to 7.9 hours, which is why the bands here run 3 to 4.5, 4.5 to 6.5 and 6.5 to 9 hours.
Grzegorzewski and colleagues (Frontiers in Pharmacology, 2021) curated 141 studies covering 4,714 individuals and found both effects reproduce across fifty years of pharmacokinetic work: smoking raises caffeine clearance, oral contraceptives lower it. The direction is well replicated. The exact hours are not, and the two studies the band edges lean on have a few dozen people between them.
What it cannot tell you
It cannot tell you your genotype. CYP1A2 rs762551 is the variant behind the fast and slow labels, and only a genetic test reads it. Behavioural markers and genotype correlate loosely at best, so a band here is a hypothesis to test on the ladder, not a finding.
It cannot tell you that you have caffeine intolerance, an allergy, or a disorder of any kind, and it does not try. If you told it that one cup gives you a racing heart, it does not interpret that. It prints one line pointing you at a doctor and stops. It will not tell you that is fine and it will not tell you it is not, because it cannot know.
It also cannot hand you a dose that is safe for anxiety, and no page should. Liu and colleagues (Frontiers in Psychology, 2024) pooled 14 comparisons from 8 articles covering 546 healthy participants and found that intake below 400 mg still moved anxiety scores, with a standardised mean difference of 0.61 (95 percent confidence interval 0.42 to 0.79). There is no known dose beneath which caffeine reliably leaves anxiety alone. Childs and colleagues (Neuropsychopharmacology, 2008) gave 102 low-consumption adults 0, 50, 150 and 450 mg in separate double-blind sessions and found the anxiety response at 150 mg tracked ADORA2A and DRD2 variants, which is to say people differ here in ways a seven-question form cannot see.
And it knows nothing about your liver, your pregnancy status, your other medications or anything else that changes the same arithmetic. Caffeine pharmacokinetics shift with all of those. If a sleep problem has lasted more than two weeks, it is worth raising with a clinician rather than adjusting a dose against a table on a website.