Caffeine Half-Life Calculator: How Much Is Left at Bedtime, and Why the Half-Life You Pick Matters More Than the Coffee

First-order elimination from up to three doses at different times of day, to a bedtime you set, with the half-life as an explicit input rather than a hidden constant — because the half-life is almost the whole answer. The commonly quoted figure of about 5 hours sits inside a measured adult range of 2.3 to 9.9 hours, and on the same three cups those two ends differ by more than the headline itself. Named presets carry the published figures for smoking, oral contraceptives, pregnancy and cirrhosis, each with its source. Published daily intakes are reported with the body that published them and are not framed as a safe amount: the FDA’s 400 mg a day for most adults, and ACOG’s statement about less than 200 mg a day in pregnancy, which is a finding about miscarriage and preterm birth and not a limit.

Last updated 6 Oct 2026

These estimate an exposure — from the weather, or from what you have taken in — using published models, not a measurement of you. Individual tolerance varies a great deal, so treat the figure as a guide to the conditions rather than a verdict on your own safety. Where a page flags a dangerous level, that is one to act on now rather than later.

Three doses, one half-life you choose, and the amount still on board at bedtime

doses, times taken, bedtime and a chosen half-life -> caffeine still in you at bedtime, and when it drops below your threshold

Dose 1: caffeine in milligramsThe default is 150 mg, near the low end of the FDA’s figure of 113 to 247 mg for a 12-fluid-ounce cup of regular brewed coffee. That range is the reason this field asks for milligrams and not “cups”: the same drink name covers a more than twofold spread, and a South Indian filter coffee, an espresso and a mug of instant are not interchangeable. The table further down lists the published figures. Enter 0 for a dose you did not have.

Dose 1: time taken, on the 24 hour clockAs a decimal: 8 is 08:00, 8.5 is 08:30, 17.25 is 17:15. If a dose time is later on the clock than your bedtime, it is read as the most recent occurrence at or before your bedtime — that is, yesterday’s — so a 22:30 espresso with a 01:00 bedtime gives 2.5 hours, not a negative number. The page says so under the result whenever that happens.

Dose 2: caffeine in milligramsThe default is 71 mg, the FDA’s figure for a 12-fluid-ounce cup of black tea. Watch what happens to the rows under the result: this smaller, later dose contributes MORE caffeine at bedtime than the larger morning coffee does. That is the whole point of modelling the times rather than the total.

Dose 2: time taken, on the 24 hour clockDecimal hours again. Enter any time; a dose of 0 mg makes its time irrelevant.

Dose 3: caffeine in milligramsLeft at 0 by default. Set it to try the late one — the 4 pm cup or the evening cola — against the same bedtime. The field accepts up to 2000 mg so that an energy-drink day or a caffeine tablet can be modelled, which is not the same as suggesting it.

Dose 3: time taken, on the 24 hour clockDecimal hours. 17 is 17:00.

Your bedtime, on the 24 hour clockDecimal hours: 23 is 23:00, 0.5 is 00:30, 1.25 is 01:15. Everything on this page is reported at this moment and forwards from it, because every dose is taken as being at or before it.

Half-life to use, and whose figure it isThis is the most important control on the page and it is deliberately not hidden. Every figure here is a published measurement with its source named, not a guess, and the list is not a menu of equally likely options — it is the spread of the evidence. Picking between the 2.3 and 9.9 hour ends changes the answer by more than the headline itself, which is why the rows under the result always show both. Choose “type my own” to use the field below instead.

Your own half-life, in hoursUsed only when the selector above is set to “type my own”. Otherwise this field is locked and shows the preset’s figure. The floor of 0.25 hours is there because a half-life of zero is not a half-life, and the ceiling of 200 hours covers the slowest figure anyone has published for caffeine, which is from severe liver disease.

Threshold you want the level to fall below, in milligramsThere is no published threshold for this and the page will not invent one. It is your own number, and the page computes when the model crosses it and nothing more. 25 mg is used as a default only because it is roughly a cup of green tea, which makes it a recognisable yardstick rather than a recommendation. No published body states an amount of circulating caffeine above which sleep is affected.

39.1mgExample

150 mg of coffee at 08:00 and 71 mg of tea at 14:00, no third dose, a bedtime of 23:00, a half-life of 5 hours and a threshold of 25 mg

One exponential per dose, added, and one logarithm to run it backwards

At your bedtime: remaining = ∑ dose_i × 0.5^{(h_i / t½)}, where h_i is the hours from dose i to the bedtime · Hours from the clock: h_i = (bedtime − time_i) mod 24, so a dose later on the clock than the bedtime is taken as yesterday’s · After bedtime the sum is a single exponential, because every term shares the same half-life: level(x) = remaining × 0.5^{(x / t½)} · So the crossing time inverts exactly: x = t½ × log₂(remaining / threshold) = t½ × ln(remaining / threshold) / ln 2

t½: the elimination half-life in hours, and the whole argument of this page. Measured adult values run from 2.3 to 9.9 hours in one study of ten people; reviews give 2.5 to 5 hours as typical. It is an input here, not a constant, because treating it as a constant is what makes other calculators look more certain than they are.
0.5^(h/t½): the fraction surviving. Equivalent forms you will see elsewhere: e^−kh with k = ln2/t½ = 0.693/t½, or 2^−h/t½. They are the same number; this page uses the base-½ form because it is the one you can check in your head at whole half-lives.
∑: addition, and the reason it is legitimate is superposition: a first-order system is linear, so the contribution of each dose can be computed alone and the results added. This stops being true if elimination saturates, which for caffeine happens well above dietary intakes.
mod 24: implemented as h = d − 24·floor(d/24) with d = bedtime − dose time, which puts h in [0, 24) for any pair of clock times and removes the need for a rule about midnight.
log₂: the base-2 logarithm, computed as ln(·)/ln 2. Reading it the other way round is a useful sanity check: the number it returns is how many halvings separate the bedtime amount from your threshold, and multiplying by the half-life turns halvings into hours.

Worked example

150 mg of coffee at 08:00 and 71 mg of tea at 14:00, no third dose, a bedtime of 23:00, a half-life of 5 hours and a threshold of 25 mg
Hours to bedtime, per dose. 23 − 8 = 15 hours for the coffee; 23 − 14 = 9 hours for the tea. Neither needs the midnight rule.
The coffee. 15 hours is exactly three half-lives at 5 hours, so the surviving fraction is 0.5³ = 0.125 and what is left is 150 × 0.125 = 18.75 mg.
The tea. 9 hours is 1.8 half-lives, so the surviving fraction is 0.5^1.8 = 0.2872 and what is left is 71 × 0.2872 = 20.39 mg.
And here is the point of the page. The 71 mg cup of tea taken at 14:00 leaves MORE caffeine at bedtime than the 150 mg coffee taken at 08:00 — 20.39 against 18.75. Twice the dose, six hours earlier, and it loses. Anyone reasoning from the day's total of 221 mg would have no way to see that.
Add them. 18.75 + 20.39 = 39.14 mg, which the headline rounds to 39.1 mg: about 17.7 % of the 221 mg taken.
When does it pass 25 mg? x = 5 × log₂(39.14 / 25) = 5 × log₂(1.5656) = 5 × 0.6466 = 3.23 hours after 23:00, which is about 02:14.
Now the honest part. The same two drinks on the fastest half-life measured in Blanchard and Sawers' ten adults, 2.3 hours, leave 6.3 mg. On the slowest, 9.9 hours, they leave 90.3 mg. The spread is 84 mg, more than twice the headline. The headline is not wrong; it is one point inside a range that the half-life, not the coffee, decides.

Published caffeine half-lives, with who measured them and in whom

These are measurements, not recommendations, and that is why they are reproduced as numbers with their sources rather than resolved into one figure. Three honest gaps. There is no clean trimester-by-trimester table of caffeine half-life in the literature this page could find: the direction and the end point are well established, the intermediate values are not. The smoking entry is a rate statement, not a measured pair of half-lives, so the “roughly half” is a derivation and is labelled as one. And the oral contraceptive studies disagree with each other — 1.5-fold, 1.7-fold and about 2-fold — so “roughly doubles”, which is what is usually said, is the top of the published range rather than the middle of it.
Group	Half-life	Source
Healthy adults, typical	2.5 to 5 hours	Arnaud 1993, as quoted in Nehlig’s 2018 review
Ten healthy adults, individually	2.3 to 9.9 hours	Blanchard and Sawers 1983, as quoted in the same review
Adult men and non-pregnant women	mean 3.4 hours, range 2 to 5, n = 25	Knutti, Rothweiler and Schlatter 1981, saliva
Cigarette smokers	roughly half the non-smoking figure	Nehlig 2018: smoking “almost doubles the rate of caffeine metabolism” by inducing CYP1A2
Three weeks after stopping smoking	caffeine concentrations at 203 % of baseline on unchanged intake	Nehlig 2018, citing Faber and Fuhr 2004 and Swanson et al. 1997
Women not using an oral contraceptive	6.2 hours	Pair reported in Nehlig 2018
Women using an oral contraceptive	10.7 hours	Same pair — an increase of about 1.7-fold
Control women, separate study	5.37 hours	Abernethy and Todd 1985, 162 mg oral dose
Oral contraceptive users, same study	7.88 hours	Abernethy and Todd 1985 — an increase of about 1.5-fold, with clearance 1.05 against 1.75 mL/min/kg
Oral contraceptive users, residence time	about 2-fold longer, within the first cycle	Rietveld et al. 1984, 9 women
Pregnancy, across gestation	mean 8.3 hours, range 3 to 16, n = 57	Knutti 1981; values returned to normal within a month of delivery
Pregnancy, towards the end	11.5 to 18 hours	Nehlig 2018; prolonged mainly in the third trimester
Hepatic alcoholic cirrhosis	50 to 160 hours	Nehlig 2018, citing Scott et al. 1988

Caffeine in common drinks, as the United States Food and Drug Administration publishes it

From the FDA’s consumer page, a United States Government work. The figure to carry away is the first row: 113 to 247 mg for the same description of the same drink, a 2.2-fold spread that depends on the bean, the grind, the brew method and the strength. An espresso, a South Indian filter coffee made with a strong decoction, and a mug of instant are not interchangeable and should not be entered as the same number. **What this page could not get:** USDA FoodData Central is the better source for per-100-g figures and would have given espresso and instant separately, but its API would not serve this build, so the table above is the FDA’s and no figure for South Indian filter coffee is quoted here at all, because no United States Government source publishes one. If you know the strength of what you drink, enter milligrams; if you do not, enter the range and run the page twice.
Drink	Serving	Caffeine
Regular brewed coffee, non-specialty	12 fl oz, about 355 mL	113 to 247 mg
Decaffeinated coffee	8 fl oz, about 237 mL	2 to 15 mg
Black tea	12 fl oz	71 mg
Green tea	12 fl oz	37 mg
Caffeinated soft drink	12 fl oz	23 to 83 mg
Energy drink	12 fl oz	41 to 246 mg

The two published daily intake figures on this page, and exactly what each one says

Reported with attribution and in the published wording, because the wording is the substance. Two things worth noticing. The 200 mg figure is ACOG’s, not the FDA’s — the FDA’s own page says only that anyone pregnant, trying to conceive or breastfeeding should talk to their health care provider about whether to limit caffeine. And neither figure is about the amount *circulating* in you, which is what this calculator computes; they are about the amount taken in over a day. The total row under the result is the one to compare against them.
Body	Figure	The wording, and what it does not say
US Food and Drug Administration	400 mg a day	“For most adults, the FDA has cited 400 milligrams a day — that’s about two to three 12-fluid-ounce cups of coffee — as an amount not generally associated with negative effects.” An amount not generally associated with negative effects in most adults. Not a maximum and not a safety limit.
US Food and Drug Administration	about 1,200 mg	The amount at which the FDA estimates toxic effects such as seizures can be observed with rapid consumption. This is a toxicity estimate, not an intake figure.
American College of Obstetricians and Gynecologists	less than 200 mg a day	Committee Opinion 462, reaffirmed 2023: “moderate caffeine consumption (less than 200 mg per day) does not appear to be a major contributing factor in miscarriage or preterm birth”. A finding about two outcomes below a stated amount. ACOG adds that the relationship of caffeine to growth restriction remains undetermined.

What a half-life actually removes, which is less than most people expect

The “five half-lives and it is gone” rule of thumb is where the 3 % comes from, and it is a statement about one dose. At a 5 hour half-life a morning coffee is down to 3 % by the early hours; at the slow end of the measured adult range the same coffee takes two days to get there, and a daily habit never clears between doses at all.
Half-lives elapsed	Fraction left	At a 5 hour half-life	At a 9.9 hour half-life
1	50 %	5 hours	9.9 hours
2	25 %	10 hours	19.8 hours
3	12.5 %	15 hours	29.7 hours
4	6.25 %	20 hours	39.6 hours
5	3.125 %	25 hours	49.5 hours
7	0.78 %	35 hours	69.3 hours
10	0.098 %	50 hours	99 hours

Why the half-life is the answer, and the four things that change it

Almost every caffeine calculator hides the one number that decides the result. They take your cups and your bedtime, apply a half-life of five hours without saying so, and print a figure to one decimal place. The arithmetic is trivial — dose times a half raised to hours over half-life, added up — and the uncertainty is enormous, and it is all in that constant. This page makes the half-life an input with its provenance attached, so that the shape of the answer is visible: a range whose width is set by which study you are, with one point marked inside it.

The measured spread is not subtle. Arnaud’s review gives 2.5 to 5 hours as the usual adult range. Blanchard and Sawers, dosing ten healthy adults in 1983, measured individual half-lives from 2.3 to 9.9 hours — a more than fourfold spread in ten people who were all healthy. Knutti’s 1981 saliva study of 25 men and non-pregnant women found a mean of 3.4 hours with a range of 2 to 5. Put those together and the honest statement is that a healthy adult’s caffeine half-life is somewhere between about 2 and 10 hours and nobody can tell you where without measuring you.

Four things move it, and three of them move it a long way.

Smoking. Caffeine is cleared mainly by CYP1A2, and the polycyclic aromatic hydrocarbons in tobacco smoke induce it. Nehlig’s 2018 review in Pharmacological Reviews states that cigarette smoking “almost doubles the rate of caffeine metabolism as a result of enzyme induction”, which is the same thing as roughly halving the half-life. The more useful half of that fact is what happens when someone stops: clearance falls by about 36 % quickly, and three weeks after quitting on unchanged coffee intake, caffeine concentrations have been reported at 203 % of baseline. Someone who quits smoking and keeps drinking four cups a day is, in caffeine terms, now drinking eight. Jitteriness, insomnia and a racing heart in the weeks after quitting are routinely put down to nicotine withdrawal when part of it is caffeine.

Oral contraceptives. Oestrogen-containing oral contraceptives impair caffeine clearance. Abernethy and Todd, in 1985, gave 162 mg of caffeine to nine users and nine matched non-smoking controls and measured half-lives of 7.88 against 5.37 hours, with plasma clearance of 1.05 against 1.75 mL/min/kg and no change in volume of distribution. Nehlig reports a different pair, 10.7 against 6.2 hours. Rietveld and colleagues in 1984 found caffeine residence time increased by a factor of two, detectable within the first cycle of starting. So the commonly repeated “doubles it” is the top of the published range: 1.5-fold, 1.7-fold and 2-fold are all in the literature, and this page offers the measured pairs rather than averaging them into a number nobody measured.

Pregnancy. This is the largest ordinary effect. Knutti’s 1981 study measured a mean half-life of 8.3 hours in 57 pregnant women against 3.4 hours in non-pregnant adults, with a range of 3 to 16 hours, and found that values returned to normal within a month of delivery. Nehlig gives the half-life as rising to 11.5 to 18 hours towards the end of pregnancy, prolonged mainly in the third trimester, against the habitual 2.5 to 5 hours. One thing this page could not establish, and says so plainly: a clean trimester-by-trimester table of half-life values. The direction is certain, the third-trimester end point is published, and the intermediate figures are not available in a form worth quoting. A 2016 study measured dose-normalised caffeine concentrations rising significantly between the first and third trimesters in 59 pregnant women, which confirms the gradient without giving half-lives for it. The presets therefore offer the measured end points and not an invented curve between them.

Liver disease, and genotype. In hepatic alcoholic cirrhosis the plasma half-life has been reported in the range of 50 to 160 hours — from two days to a week. At those rates a daily coffee accumulates rather than clearing, and the single-day model on this page understates the real level substantially; it would take a week of doses to reach steady state. Caffeine clearance is in fact used as a probe of liver function for exactly this reason. Separately, a common variant in CYP1A2 (the −163C>A substitution, rs762551) sorts people into faster and slower metabolisers, which is part of why ten healthy adults spread over fourfold.

What the page deliberately does not do. It does not tell you a safe amount, because that is not a thing this page can know and the bodies that publish figures are careful about what they are claiming. It reports the FDA’s 400 mg a day as “an amount not generally associated with negative effects” in most adults, which is its actual wording and is not a limit. It reports ACOG’s less-than-200-mg statement as a finding about miscarriage and preterm birth, which is what it is, and notes that ACOG itself says the relationship to growth restriction is undetermined. It makes no claim about sleep quality beyond arithmetic: it will tell you how many milligrams the model leaves at your bedtime, and it will not tell you that the figure is too high, because no published body states a circulating amount above which sleep is impaired. And it does not model absorption. Peak plasma concentration arrives roughly 30 to 120 minutes after a dose, so for the first hour or two the model is above reality; after that it is a reasonable description of the decay.

Related pages on this site. The heat index page is the other exposure page in this group, and the two interact more than you would think: caffeine is a mild diuretic, heat is a fluid problem, and a hot night with 100 mg still on board is a worse night than either alone. The alcohol units page takes the same approach of reporting each body’s published figures side by side rather than resolving them, and the pack-years page is the one to read alongside the smoking section above. For the general pharmacokinetics, the medical set has a drug half-life and steady-state page that does the accumulation arithmetic this page leaves out, and a theophylline converter — theophylline being one of caffeine’s own active metabolites, and the one that rises most in the third trimester of pregnancy.

Frequently asked questions

What half-life should I pick if I do not know anything about myself?

Use the 5 hour preset to get a figure, then read the two rows below it — the same doses at 2.3 and at 9.9 hours — as the actual answer. Those are the fastest and slowest individuals in a study of ten healthy adults, and the gap between them is usually larger than the headline. If you have one of the named conditions, pick its preset: the published figures for smoking, oral contraceptives, pregnancy and cirrhosis are all in the list with their sources. If you have no idea at all, the honest answer to “how much caffeine is still in me” is a range of a factor of ten, and a page that hands you a single number to one decimal place is hiding that from you.

I stopped smoking and coffee suddenly feels much stronger. Is that real?

Yes, and it is one of the better-documented drug interactions in ordinary life. Tobacco smoke induces CYP1A2, the enzyme that clears caffeine; Nehlig’s review states that smoking almost doubles the rate of caffeine metabolism. When you stop, the induction fades: caffeine clearance falls by about 36 % quickly, and three weeks after quitting with unchanged coffee intake, caffeine concentrations have been reported at 203 % of baseline. Set the selector to 2.5 hours to see your old numbers and to 5 hours to see your new ones on the same drinks. Some of what is attributed to nicotine withdrawal in the first weeks is caffeine excess, and halving the coffee is a reasonable thing to discuss with whoever is helping you quit.

Why does the page not just tell me when I can have my last coffee?

Because that answer needs a threshold, and no published body states a circulating amount of caffeine above which sleep is affected. The page therefore asks you for the threshold and computes the crossing time for it, and labels the default as an arbitrary yardstick rather than a recommendation. You can use it the other way round, which is more useful: set the dose and the bedtime, and read how many milligrams are left. That figure is comparable with the size of a dose, and “a whole cup still on board at bedtime” is a statement you can reason about without anyone having to invent a limit.

Is the 400 mg a day figure a safety limit?

No, and the FDA’s wording is careful about this. What it says is that “for most adults, the FDA has cited 400 milligrams a day — that’s about two to three 12-fluid-ounce cups of coffee — as an amount not generally associated with negative effects”. That is a statement about most adults and about an absence of generally observed negative effects. It is not a maximum, it does not apply to children or adolescents, and the same FDA page stresses the wide variation in how sensitive people are and how fast they eliminate it. Separately the FDA estimates that toxic effects such as seizures can be observed with rapid consumption of around 1,200 mg. This page reports both figures with their source and draws no conclusion from either.

And the 200 mg figure in pregnancy?

That one is not the FDA’s at all, which is a common misattribution. It comes from the American College of Obstetricians and Gynecologists, Committee Opinion 462, reaffirmed in 2023: “moderate caffeine consumption (less than 200 mg per day) does not appear to be a major contributing factor in miscarriage or preterm birth”. Read what that sentence is: a finding about two named outcomes below a stated amount, not a limit, and ACOG’s own document adds that the relationship of caffeine to growth restriction remains undetermined. The FDA’s own page says only that anyone pregnant, trying to conceive or breastfeeding should talk to their health care provider about whether to limit caffeine. The pharmacokinetic point is separate and makes the figure easier to understand: in late pregnancy the half-life rises to something like 11.5 to 18 hours, so the same cup of coffee stays around three to five times as long.

Why does my small afternoon tea beat my big morning coffee?

Because the exponent does more work than the dose. At the default settings a 150 mg coffee at 08:00 has had 15 hours by a 23:00 bedtime, which is three half-lives at 5 hours, so 12.5 % of it is left: 18.75 mg. A 71 mg tea at 14:00 has had 9 hours, which is 1.8 half-lives, so 28.7 % of it is left: 20.39 mg. Less than half the caffeine, taken six hours later, and it wins. This is the single most useful thing the page shows, and it is invisible to anyone reasoning from the day’s total. Each dose’s contribution is a separate row under the result for exactly this reason.

Can I enter more than three doses?

Not directly, and there is a workaround that is exact rather than approximate. Because first-order kinetics is linear, two doses taken at the same time add: a 60 mg and a 90 mg drink both at 16:00 are identical to one 150 mg dose at 16:00, so doses can be merged by time. If you genuinely have more than three distinct times, run the page twice and add the two headline figures — that is also exact, for the same reason. What you must not do is add the milligrams and use an average time, because the decay is exponential and the average of two exponentials is not the exponential of the average.

How accurate is the model itself, leaving the half-life aside?

It is a one-compartment model with instantaneous complete absorption, and its main error is at the start. Real plasma concentrations peak roughly 30 to 120 minutes after a dose, so for the first hour or two after drinking, the real level is below what the model says and then crosses it; by three or four hours the difference is small. Caffeine’s elimination is close enough to first-order at dietary intakes for a single exponential to be reasonable. At very large doses and in overdose that stops being true. The model also knows nothing about your weight, which matters for the peak concentration but not for the half-life, and nothing about tolerance, which changes the effect of a given level and not the level itself.

Does a daily coffee habit accumulate?

At an ordinary half-life, hardly at all: 24 hours is nearly five half-lives at 5 hours, so about 3.6 % of yesterday survives to today and the day-to-day carry-over is negligible. That is why this page models one day. But the carry-over scales with the half-life, and at the slow ends of the published range it stops being negligible — at 18 hours, roughly 40 % of yesterday is still there, and at the 50 to 160 hours reported in alcoholic cirrhosis almost all of it is, so the level climbs for a week before levelling off. On those presets the figure this page shows is an underestimate, and the page says so under the result. The steady-state arithmetic itself is done properly by the drug half-life and steady-state page in the medical set.

Why does a dose time later than my bedtime not produce an error?

Because the more useful reading is that it was yesterday’s. If you had an espresso at 22:30 and go to bed at 01:00, the elapsed time is 2.5 hours, and refusing that would make the page useless for exactly the people most likely to want it. The page computes hours as (bedtime minus dose time) modulo 24, which puts every dose at the most recent time it could have been taken at or before your bedtime, and it tells you under the result whenever that wrap has been applied. If the dose really is still in the future, the figure is wrong and you want to move the bedtime rather than the dose.

Related calculators

References

Nehlig A. Interindividual differences in caffeine metabolism and factors driving caffeine consumption. Pharmacological Reviews 2018;70(2):384–411. Source for the 2.5–5 h and 2.3–9.9 h adult ranges, the smoking statement and the 36 % and 203 % figures after cessation, the 6.2 against 10.7 h oral contraceptive pair, the 11.5–18 h late-pregnancy figure and the 50–160 h cirrhosis range.
Knutti R, Rothweiler H, Schlatter Ch. Effect of pregnancy on the pharmacokinetics of caffeine. European Journal of Clinical Pharmacology 1981;21:121–126. Mean 3.4 h (range 2–5, n = 25) in adult men and non-pregnant women against 8.3 h (range 3–16, n = 57) in pregnancy, returning to normal within a month of delivery.
Abernethy DR, Todd EL. Impairment of caffeine clearance by chronic use of low-dose oestrogen-containing oral contraceptives. European Journal of Clinical Pharmacology 1985;28(4):425–428. Half-life 7.88 h in nine users against 5.37 h in nine matched controls after 162 mg orally, with plasma clearance 1.05 against 1.75 mL/min/kg and no change in volume of distribution.
Rietveld EC, Broekman MMM, Houben JJG, Eskes TKAB, van Rossum JM. Rapid onset of an increase in caffeine residence time in young women due to oral contraceptive steroids. European Journal of Clinical Pharmacology 1984;26:371–373. Residence time increased by a factor of two, already present in the first cycle.
Blanchard J, Sawers SJA. The absolute bioavailability of caffeine in man. European Journal of Clinical Pharmacology 1983. The ten-adult study whose individual half-lives of 2.3 to 9.9 hours are used here as the fast and slow comparison rows; quoted via Nehlig 2018.
Arnaud MJ. Pharmacokinetics and metabolism of natural methylxanthines in animal and man. 1993. Source of the 2.5 to 5 hour typical adult range; quoted via Nehlig 2018.
Yu T, et al. Pregnancy-induced changes in the pharmacokinetics of caffeine and its metabolites. Journal of Clinical Pharmacology 2016;56(5):590–596. Dose-normalised caffeine and theophylline concentrations rising significantly between the first and third trimesters in 59 pregnant women — evidence of the gradient, without half-lives for it.
United States Food and Drug Administration. Spilling the Beans: How Much Caffeine is Too Much? fda.gov. The 400 mg a day wording, the approximately 1,200 mg toxicity estimate, and the caffeine content table reproduced on this page.
American College of Obstetricians and Gynecologists, Committee on Obstetric Practice. Moderate Caffeine Consumption During Pregnancy. Committee Opinion No. 462, August 2010, reaffirmed 2023. The less-than-200-mg statement about miscarriage and preterm birth, and the statement that the relationship to growth restriction remains undetermined.
United States Department of Agriculture. FoodData Central. fdc.nal.usda.gov. Named as the better source for per-serving caffeine figures; its API would not serve this build, so the FDA table is used instead and no figure is quoted for South Indian filter coffee.

CalcEngines health calculators are for education and for checking arithmetic that has already been decided elsewhere. They are not medical advice, they do not decide what to give, and they do not replace the judgement of a doctor, nurse, midwife or dietitian who knows the person in front of them. Every figure depends on the values you enter and on the assumptions stated on the page — check it against the prescription, the product label and your local policy before acting on it.