Caffeine for training and focus: timing, dose, tolerance and sleep

Caffeine is the most reliable legal performance aid we have, but the gains are modest and the trade-off is sleep. The sports-nutrition consensus is that 3 to 6 mg per kg of body mass, taken about 60 minutes before training, consistently improves endurance and gives smaller boosts to strength and power. Its half-life of roughly 5 to 6 hours is why an afternoon coffee can quietly cut your deep sleep that night, even if you fall asleep fine. How much it helps, and how badly it disrupts sleep, varies a lot between people, partly down to a single gene.

How caffeine actually works

As you stay awake, a molecule called adenosine builds up in the brain and pushes you toward sleep and fatigue. Caffeine is shaped enough like adenosine to sit on its receptors and block them, so the brain stops receiving the "you are tired" signal. The result is more arousal and, crucially for training, a lower perception of effort and pain at a given workload.¹² You are not adding energy; you are masking fatigue, and the sleep pressure that adenosine represents is only postponed, not erased. Caffeine reaches peak blood levels at roughly 30 to 120 minutes, the basis for the standard hour-before timing.¹ If you are building a wider routine, our health library and the stack builder show where stimulants fit.

The endurance evidence is the strongest

Aerobic endurance is where caffeine earns its reputation. The 2021 International Society of Sports Nutrition position stand, the most thorough review of the field, concludes that endurance shows the most consistent, moderate-to-large benefit, though the size differs between individuals.¹ A 2018 systematic review and meta-analysis of 46 studies found a small but clear improvement in endurance performance with moderate doses.² In cyclists specifically, a 2024 meta-analysis reported that 4 to 6 mg/kg significantly improved time-trial performance, while low doses under 3 mg/kg did not reach significance.⁴ The practical read: for a hard run, ride or row, caffeine is one of the few supplements with genuinely robust support.

Strength and power: real, but smaller

For lifting, the effect is present but you should not overstate it. A 2018 meta-analysis by Grgic and colleagues found caffeine improved maximal strength (effect size around 0.20) and power (around 0.17), small effects most visible in upper-body lifts and in movement velocity rather than a transformed one-rep max.³ Muscular endurance, doing more reps before failure, tends to respond a little more strongly. So caffeine can sharpen a session and help you grind out extra quality reps, but it will not add meaningful weight to the bar by itself. It pairs sensibly with creatine, which targets a different energy system.

Dose and timing

The position stand is precise: caffeine reliably works at 3 to 6 mg/kg of body mass, which is roughly 210 to 420 mg for a 70 kg person.¹ The minimum useful dose may be as low as 2 mg/kg, and going above 6 mg/kg buys little extra performance while sharply raising the risk of jitters, a racing heart, gut upset and anxiety.¹ Take it about 60 minutes before you start, though faster formats such as caffeinated gum act sooner. There is no need to chase the high end: start at the lower dose, see how you respond, and only nudge up if needed.

The half-life problem: why afternoon caffeine wrecks deep sleep

Caffeine's half-life in a healthy adult is about 5 to 6 hours, meaning half the dose is still circulating that long after you take it, and a fraction lingers far longer.⁷ A 3pm coffee can therefore leave a quarter of the caffeine in your system at bedtime, still blocking adenosine receptors. In a controlled 2013 trial, a 400 mg dose taken 6 hours before bed cut objectively measured total sleep time by more than an hour, and participants often did not notice the loss.⁵ The damage is not only to how long you sleep but to its depth: evening caffeine suppresses slow-wave (deep) sleep, the most physically restorative stage, blunting recovery from the very training the caffeine was meant to support. A sensible rule is no caffeine within about 8 to 10 hours of bed. See our piece on sleep architecture for why deep sleep matters.

Tolerance, cycling and the genetics of response

Regular use does build some tolerance: at least in lab settings, a daily 3 mg/kg habit can blunt some of caffeine's effects over time.¹⁰ But the practical picture is reassuring. The position stand notes that caffeine remains ergogenic in habitual users, and that short pre-event withdrawal offers little reliable benefit while risking headaches and irritability, so it is not recommended.¹ If you want a fresher response for an important session, the simplest lever is to keep your everyday intake modest rather than to abstain and rebound.

Individual variation is large, and one driver is the CYP1A2 gene, which makes the liver enzyme that clears most of your caffeine. People carrying the "AA" version tend to be faster metabolisers; carriers of the C-allele clear it more slowly and keep higher blood levels for longer. A 2018 study by Guest and colleagues in 101 cyclists found 2 and 4 mg/kg improved time-trial performance only in AA metabolisers, with no benefit in AC carriers and impaired performance in CC carriers at the higher dose.⁶ Be honest about the limits: other trials have failed to replicate a clear genotype effect, so the science is unsettled.¹ You do not need a gene test to act on this. Your own response to a 3pm coffee tells you most of what you need.

This article is educational and does not constitute medical advice, diagnosis, or a treatment recommendation. Medication uses described as “off-label” are not licensed for that purpose in the UK and should only be considered under qualified clinical supervision. Always speak to your GP, pharmacist, or a registered specialist before starting, stopping, or changing any treatment. If you have severe or alarm symptoms - unintentional weight loss, blood in your stool, difficulty swallowing, persistent vomiting, a fever, or severe pain - seek urgent medical care.