Caffeine half-life math

"No caffeine after 2pm" is decent advice built on a number almost nobody checks: the half-life. This page does the arithmetic — what actually remains in your bloodstream at 11pm — and explains why your curve may look nothing like the population average. The endpoint is not a rule you obey; it's a two-week experiment that finds your own curfew.

The Rule Everyone Quotes — and Where It Comes From

A drug's half-life is the time it takes your body to clear half of it. Caffeine's is usually quoted as five to six hours, and that figure comes from pharmacokinetic studies in healthy adults: after one half-life, 50% remains; after two, 25%; after three, 12.5%. The parent saboteurs topic covers what caffeine does (it blocks adenosine, the molecule that builds your sleep pressure). This page is about the numbers underneath that mechanism — because the numbers are what decide whether your afternoon cup is still working against you at midnight.

The cleanest way to feel the math is a worked example. Say you have 200 mg of caffeine — roughly a large brewed coffee — at 2pm, and your half-life is the typical 5.5 hours. At 10pm, about a third of the dose (~70 mg) is still circulating. At 2am, roughly a fifth (~45 mg) remains. That late dose is not a rounding error: it is comparable to a full shot of espresso, arriving exactly when your brain is trying to build deep sleep.

Same Dose, Different Curve

The population average hides the actual distribution. The chart below shows how much of a 2pm dose is still in the blood at 11pm, by metabolizer type:

Your Half-Life Is Not the Population Average

Caffeine clearance is dominated by one liver enzyme, CYP1A2, and that enzyme's activity is all over the map. The main drivers:

• 🧬 Genetics. Variants in and around the CYP1A2 gene shift enzyme activity in both directions, and variants in adenosine-receptor genes may change how strongly a given blood level affects you. Studies find real differences on average — but for any one person, genetic test results have a mixed record at predicting what actually happens to sleep.
• 🚬 Smoking. Polycyclic hydrocarbons in cigarette smoke induce CYP1A2, roughly halving caffeine's half-life in smokers — one reason smokers self-report needing more caffeine.
• 💊 Oral contraceptives. Estrogen-containing pills slow CYP1A2 markedly; half-lives roughly double in many users.
• 🤰 Pregnancy. Clearance slows progressively, especially in the third trimester, when half-lives can exceed ten hours — a strong argument for near-total caffeine reduction late in pregnancy.
• 🫁 Liver function. Anything that reduces hepatic capacity — disease, some medications, older age to a mild degree — lengthens the half-life. Caffeine clearance is literally used as a clinical test of liver function.

The practical consequence: any cutoff time you read is an average built for someone else. A 3pm coffee that a fast metabolizer clears to 12% by 11pm leaves a slow metabolizer at 50%. Both people "feel fine" at dinner.

Feeling Fine Is Not the Test

The most important study in this literature is unglamorous. Researchers gave healthy people 400 mg of caffeine — a large but common daily amount — at bedtime, three hours before bed, or six hours before bed (Journal of Clinical Sleep Medicine, 2013). Every timing condition cut total sleep time by roughly an hour and worsened subjective sleep, with effects detectable even at six hours out. The part people remember: participants frequently did not perceive the damage. Sleep was objectively shorter; mornings were not dramatically different to them.

EEG studies tell the same story under the hood. A moderate caffeine dose before sleep reduces slow-wave activity — the electrical signature of deep sleep — even when people fall asleep normally (Neuropsychopharmacology, 1995). Caffeine does not need to keep you awake to hurt you; it can let you fall asleep and then quietly thin out the deep stages. That is why "I can have an espresso after dinner and still sleep fine" is usually a claim about feeling fine, not sleeping fine — a distinction the deep sleep topic makes central.

Tolerance, Withdrawal, and the Decaf Footnote

Two confusions trip people up. First, tolerance is not the same as metabolism. With daily use, your brain adapts so that a given dose feels less stimulating — but the molecule still blocks adenosine receptors, and the sleep architecture costs remain measurable in habitual users. Tolerance can also backfire through withdrawal: overnight abstinence produces morning headaches and sluggishness that people attribute to poor sleep, then treat with the morning coffee that perpetuates the cycle.

Second, decaf is not zero — a cup carries roughly 3–5 mg, and "half-caf" drinks hide a real dose. In someone with a long half-life who drinks a large decaf at 9pm, that residue is minor; in a fast metabolizer it is negligible. The lesson is not decaf paranoia but arithmetic hygiene: count the total daily load, not the individual cup, because it is the tail of the day's caffeine — the part still uncleared at 11pm — that your brain experiences. The daily-ceiling and source-by-source numbers live on the parent topic.

Questions, Answered Briefly

• 🧬 Should I get a caffeine-metabolism genetic test? They classify CYP1A2 variants, but the correspondence between that result and your actual sleep is loose. A two-week curfew experiment is cheaper, faster, and measures the outcome you care about.
• 🍫 Do tea and chocolate really count? Yes — a mug of black tea is roughly 45 mg and 30 g of dark chocolate roughly 20–30 mg. Total daily load, not beverage identity, is what survives to midnight.
• ⏰ Is there a dose that's safe any time of day? The honest answer is that even modest evening doses leave measurable caffeine in the brain at lights-out for most people. The timing of your last dose does the work that reducing it cannot.
• 🛌 Does exercise or food speed clearance? Not meaningfully. Grapefruit slows CYP1A2 slightly, but the dominant variables are genetics, smoking, hormones, and liver function — none of which respond to willpower.

Related Topics