Caffeine and endurance performance

Caffeine is, along with carbohydrates, one of only two ergogenic aids whose effect in endurance is both established, reproducible and of useful magnitude. This is not a forum opinion: it's what emerges from several converging meta-analyses. It remains to know how much, when, and what you can reasonably expect from it.

The real gain: 2 to 3%

At moderate doses, the measured improvement over placebo is approximately 2.9% on average power and 2.3% on time-trial performance, with effect sizes around 0.22 to 0.28 — small, but consistent from study to study.

Two percent seems negligible. On a 3:30 marathon, that's four minutes. That's more than six weeks of additional training will bring you at that level. Few legal things do better for so little.

The dose: 3 to 6 mg per kilo

This is the useful window, and it's well delimited on both sides.

• Below 3 mg/kg, the effect is no longer significant: doses of 1 to 3 mg/kg do not show reliable time-trial improvement.
• Between 3 and 6 mg/kg, the effect is equivalent across the entire range — no need to go higher. The 4 to 6 mg/kg zone gives the most consistent effect.
• Beyond that, you improve nothing and accumulate side effects: tachycardia, tremors, digestive upset, destroyed sleep.

Concretely, for 70 kg: 210 to 420 mg. A double espresso contains about 150 mg, a caffeinated gel 50 to 100 mg. Weigh yourself and calculate rather than counting in cups — coffee content varies threefold depending on the machine and grind.

Timing: about 60 minutes before

Peak plasma levels arrive 45 to 60 minutes after ingestion in liquid or capsule form, and the effect is maintained for several hours. On a long effort, a second smaller dose mid-race prolongs the effect — this is the logic behind caffeinated gels placed in the second half.

On a short race (5 to 10 km), a single dose one hour before is more than enough.

Should you stop caffeine before a race?

This is the most asked question, and the answer has changed. The idea of "withdrawal" over several days to restore sensitivity doesn't hold up well: regular consumers benefit from the ergogenic effect, even without withdrawal. Abrupt cessation mainly brings headaches and a week of degraded training before your goal — a bad trade-off.

On the other hand, individual response varies greatly, partly for genetic reasons (rapid or slow metabolism). Hence the rule that applies to any race strategy: test in training, on a long run or quality session, never on race day.

Cases where it doesn't work

• If your stomach doesn't cooperate. Caffeine accelerates transit and increases acidity. Combined with pre-race stress and an ill-prepared nutritional strategy, it causes exactly the problem you were trying to avoid.
• If the race is in the evening. The half-life is 5 to 6 hours, and longer in slow metabolizers. A dose at 6 p.m. is paid back in sleep that night — which can cost more than the 2% gained, especially over a two-race weekend.
• If you're counting on it to compensate. Two percent on a well-built base is valuable. Two percent on sloppy training or three nights at five hours, it doesn't make up for anything.

Precautions

Caffeine is not recommended in case of cardiac arrhythmia, uncontrolled hypertension, marked anxiety disorder, and during pregnancy beyond intakes recommended by medical supervision. In case of medication, seek advice: interactions exist.

Put it in the right place

Caffeine is a fine-tuning adjustment, not a foundation. It's worth discussing once volume, intensity distribution, sleep and nutrition are in place — the order matters, because each of these elements delivers far more than 2%. Sports Coach AI programs this base from your actual data; caffeine is tested on a long run, noted in how you feel, and saved for the day when it will deliver those four minutes.

References

• Grgic J. et al., 2018-2019, Sports Medicine, 49(1) — acute caffeine effect on endurance performance, systematic review and meta-analysis.
• Journal of the International Society of Sports Nutrition, 2024 — caffeine effect on time-trial in cyclists, dose meta-analysis.
• Nutrients, 2026, 18(12), 1989 — dose-response relationship of oral caffeine on aerobic performance.