Sleep is the best-measured point of contact. The 2020 collection contains an original study: 6 mg per kilo taken in the evening before exercise clearly degrade the following night (efficiency down, more awakenings, longer time awake in bed) without affecting next-day performance, which does not prove that the lost night costs nothing elsewhere.

The meta-analysis by Gardiner and coauthors (Sleep Medicine Reviews, 2023), on 24 studies, quantifies the average effect: 45 minutes less total sleep, 7% less efficiency, 9 minutes more to fall asleep, 12 minutes more nighttime wakefulness. To stop measuring an effect, a coffee of 107 mg must be drunk at least 8.8 hours before bedtime, and a dose of 217 mg at least 13.2 hours before.

These are averages over populations mixing slow and fast metabolizers; everyone can make their own measurement by moving their last cup back by two hours.

MDPI 2020, pp. 42-45; Gardiner et al., The effect of caffeine on subsequent sleep: a systematic review and meta-analysis, Sleep Med. Rev. 69 (2023) 101764

The European Food Safety Authority published its opinion on caffeine on May 27, 2015. For the healthy adult, it retains 400 mg per day and 200 mg in a single intake, or 5.7 mg per kilo per day. These are not recommended doses but ceilings: below them, the agency sees no safety concern.

Two groups receive a lower threshold. The pregnant or breastfeeding woman is brought down to 200 mg per day, against 300 mg in the old value. We saw above why: the enzyme that eliminates caffeine loses a third of its activity in the first trimester, half in the second. Children and adolescents are held to 3 mg per kilo per day.

ANSES, in its opinion of January 31, 2014 on an enriched confectionery, takes up the 400 mg per day for the adult. For the child it retains the stricter dose of Health Canada, 2.5 mg per kilo. And it recalls who pays first for exceeding it: children, pregnant women and sensitive people. In them appear irritability, anxiety and sleep disturbances, and, at high doses, panic attacks.

On pregnancy itself, the 2020 collection is mixed. A Polish study of a hundred women consuming on average 68 mg per day finds no link with the weight or length of the newborn. Meta-analyses conducted at higher consumption levels see something else. They associate caffeine with an increased risk of miscarriage, of 7 to 19% per increment of 100 to 150 mg daily. Low birth weight is also associated, with results that vary by country.

There is a reason to be wary of these associations even more than of the others. First-trimester nausea is more frequent in pregnancies that go well, and it often turns women off coffee. The woman whose pregnancy holds firm therefore drinks less, which is enough to manufacture the association without any molecule having acted. Reverse causation is particularly hard to rule out here. What follows is a matter for a conversation with a health professional.

EFSA, Scientific Opinion on the safety of caffeine, EFSA Journal 13(5):4102 (2015); ANSES, opinion 2013-SA-0068 of 01/31/2014; MDPI 2020, pp. 229-233; LC ch. 19

For sport, the reference dose is called ergogenic, that is, capable of improving performance. It is 3 to 6 mg per kilo before exercise, or 210 to 420 mg for 70 kg. The International Olympic Committee recognizes it as such.

Going above brings nothing and costs something. At 9 to 11 mg per kilo, in athletes who are nonetheless habituated, the additional benefit is not clear. Palpitations, anxiety and digestive disturbances, on the other hand, appear in up to 92% of subjects. Severe effects begin between 9 and 13 mg per kilo in a single intake, or 630 to 910 mg at once for 70 kg.

Anti-doping regulation long sought a threshold, then gave up. The urinary threshold was 12 µg/mL in 1987, then more than 6 µg/mL, which corresponds to more than 9 mg per kilo ingested. The World Anti-Doping Agency removed caffeine from its prohibited list in 2004, and keeps it under monitoring.

There remains the figure of the non-responder, the person on whom caffeine would not act. The collection doubts it, for a reason of protocol. Most studies test each subject only once with caffeine and once without, whereas an athlete's performance already varies from one day to the next. A repeated-measures study finds that all subjects respond at least three times out of eight. In other words, what was taken for insensitivity may be only measurement noise.

One last figure circulates and says nothing about the coffee drinker. A toxicity study on pubescent rats works at 120-180 mg per kilo per day, or 8 to 13 g per day scaled to a human. It serves to identify the target organs of an intoxication, not to describe a cup, and cannot be extrapolated.

MDPI 2020, pp. 11-18, 90-93, 115-120; LC ch. 19