Caffeine on the Blocks, How Much Makes Sprinters Faster
A deep dive into one paper: what was known before, what the researchers did, what the numbers say and what they don't.
The paper: Mei, Li and Yang (2026), PLOS One: Acute doses of 3 and 6 mg/kg caffeine in male 100-m sprinters
Chapters
Transcript
The script the two AI voices read from. The times are worked out from the script, not heard, so a line can start a few seconds early or late.
Read the whole episode
Twenty-one sprinters and a capsule
Host 2Twenty-one trained sprinters swallowed a capsule an hour before running sixty metres, and on a strong dose of caffeine every single one of them ran faster than on a capsule of plain flour. The average gain was about seventeen hundredths of a second, which in sprinting is not nothing.
Host 1And the smaller dose, the kind you might get from a couple of strong coffees, did that help too?
Host 2It did: twenty of the twenty-one were faster on it, by about twelve hundredths of a second on average. The interesting part is that the bigger dose helped more for some things and not at all more for others.
Host 1Welcome to the Hastrup Review of Science, where artificial intelligence reads real science, so you don't have to pretend you understood the paper.
Why caffeine and sport go together
Host 1Before we get to the track, why do sports scientists spend so much time on caffeine in the first place?
Host 2Because it is one of the most widely used performance aids in competitive sport, and it is legal and cheap. The paper sums up how it works: caffeine crosses into the brain and blocks receptors for adenosine, a chemical that normally dampens nerve activity, so you get more alertness, more drive to the muscles, and less sense of pain and effort.
Host 1So adenosine is a sort of brake, and caffeine stops the brake from being pressed?
Host 2That is a fair picture. In the wider field, it helps that caffeine was taken off the list of banned substances for competition in two thousand and four, so athletes can use it openly, and researchers have had two decades of reasons to test it.
Host 1How much caffeine are we actually talking about when researchers say a dose that helps?
Host 2The usual recommendation, which the paper cites from a sports nutrition position statement, is three to six milligrams per kilogram of body weight, taken about an hour before exercise. These sprinters weighed seventy-four kilos on average, so three milligrams per kilo is roughly two hundred and twenty milligrams, and six is about four hundred and forty.
Host 1And in cups of coffee, what does that come to for an ordinary person?
Host 2As a rough rule from the wider literature, a mug of filter coffee carries somewhere around eighty to a hundred and twenty milligrams, so the low dose is two or three mugs and the high dose four or five, all at once. Food safety agencies generally treat about four hundred milligrams spread over a whole day as fine for healthy adults, so the high dose is a lot to take in one go.
What was known before
Host 1What was already known about caffeine and sprinting before this study?
Host 2Quite a lot for endurance, strength and jumping, and the evidence there is fairly consistent that caffeine helps. For sprinting it is messier, because most studies used footballers, basketball players or rugby players doing repeated short sprints with little breaks in between.
Host 1Why does that matter, if a sprint is a sprint?
Host 2Because repeated sprints test how well you recover between efforts, and that leans on your aerobic system, which a single all-out hundred metres barely uses. Team-sport athletes also move differently from specialist sprinters, so the authors argue you cannot simply carry those results over to the track.
Host 1Has anyone tested caffeine on proper track sprinters running a single race?
Host 2A little. One earlier field study gave six milligrams per kilo to sprinters over a hundred metres and found they ran faster, with most of the gain appearing in the first sixty metres, and another found improvements throughout the acceleration phase. But both used just one dose.
The open question
Host 1So nobody knew whether the smaller dose would have been enough on its own?
Host 2Exactly. And there were hints both ways: in some tasks three milligrams gives as much benefit as six, while in jumping, throwing and weight-lifting style tasks the higher dose has sometimes done more. So the question was whether doubling the dose buys anything for a sprinter, and in which part of the sprint.
Host 1When you say part of the sprint, what are the parts? I always thought you just run as fast as you can.
Host 2You do, but you are not at a constant speed. There is the reaction to the gun, pushing off the blocks, the acceleration over the first thirty to fifty metres, reaching top speed, and then holding on as you slow, and caffeine could in principle help one of those and not the others.
How the trial worked
Host 1Who were the runners in this study?
Host 2Twenty-one male hundred-metre sprinters from the training programme at Beijing Sport University, aged eighteen to twenty-five, with at least three years of systematic sprint training. Their personal bests averaged eleven point zero seconds, which meets the Chinese national second-class standard.
Host 1Is eleven seconds fast, or just fast for a university?
Host 2Good club or university level, well short of the elite: world-class men run under ten. So these are serious, trained athletes, not Olympic finalists, and that matters for how far the result travels.
Host 1And what did each of them actually swallow?
Host 2On three separate test days, each man got one of three treatments: a placebo capsule filled with white flour, caffeine at three milligrams per kilo, or caffeine at six. The capsules all looked the same and were white inside, and the person who prepared and coded them was not involved in the testing.
Host 1So every runner tried all three, rather than splitting them into three groups?
Host 2Yes, that is what a crossover trial means: each person acts as his own comparison, which cuts out the huge differences between individuals. It also means you need far fewer people than if you compared three separate groups.
Host 1Does the order not matter, though? Maybe you just get better at the test by the third time.
Host 2It can, and that is why the order was randomized into three rotating sequences, with seven men in each, so every treatment appeared equally often on the first, second and third visit. They also had a separate practice visit beforehand, and the statistical model explicitly allowed for which visit it was.
Host 1What stops caffeine from the first visit still being in someone's body at the next one?
Host 2The visits were at least seven days apart, while caffeine in adults typically halves in about four to six hours. They also asked everyone to avoid coffee, tea, cola, energy drinks and chocolate for forty-eight hours before each test, and to repeat the same meals before each visit.
Host 1And double-blind means neither the runners nor the people timing them knew which capsule was which?
Host 2Right, the participants and all the staff giving capsules and running the tests were kept unaware. One honest gap, which we will come back to, is that nobody checked afterwards whether the runners could guess what they had taken.
Host 1Walk me through a test day: what happened after the capsule went down?
Host 2They swallowed it with two hundred millilitres of water, warmed up gently from thirty minutes after, and at sixty minutes ran an all-out sixty-metre sprint from starting blocks, twice, five minutes apart, keeping the faster one. Ten minutes after that came two kinds of vertical jump, and then a ten-second all-out effort on a stationary bike called a Wingate test.
Host 1Why sixty metres and not the full hundred?
Host 2Sixty metres covers the start and the whole acceleration phase, which is where the earlier study saw most of the gain. The cost is that this study can say nothing about the last forty metres, where a sprinter holds speed or fades.
Host 1How do you measure something as small as a few hundredths of a second reliably?
Host 2Electronic timing gates recorded the time at ten, thirty and sixty metres. The starting blocks were instrumented force blocks that measured reaction time to the start signal and how quickly force built up against each block, and a row of light sensors along the first ten metres measured step length and step frequency.
Host 1How quickly force built up, is that the same as how hard they pushed?
Host 2Not quite: it is the rate of force development, how steeply the push rises in the first instant, rather than the peak push itself. In a sprint start you only have a fraction of a second on the blocks, so how fast you can get force on is often more useful than how much you could eventually produce.
Host 1And how did they turn all those numbers into a verdict?
Host 2With a mixed-effects model, which compares each man to himself across his three visits while adjusting for visit order and sequence group. For each outcome they then made three comparisons, low dose against placebo, high dose against placebo, and high against low, with a correction for making several comparisons at once.
A word from our sponsor
Host 1This episode is brought to you by Crocs, the foam clog, now tested in a randomized, double-blind crossover against bare feet.
Host 2Crocs have not paid for this, have not been asked, and as far as we can tell are unaware that the trial took place.
Host 1Common side effects include a start reaction that arrives seventeen milliseconds before the starting signal, and a mild urge to crouch in starting blocks at bus stops.
Host 2Mine reached the shop before I had decided to go. I followed after a seven-day washout.
Host 1Less common: the ventilation holes develop their own rate of force development, and the clogs complete the sixty metres without you.
Host 1Rarely, the left clog is allocated to placebo, the right one is not, and the wearer arrives at work in two different years. Crocs. Large effect, wearer lost to follow-up.
What they found
Host 1How much faster did caffeine actually make them over sixty metres?
Host 2On the low dose, sixty-metre times dropped by about twelve hundredths of a second compared with placebo, and on the high dose by about seventeen and a half hundredths. That works out to roughly one point seven and two point four percent of their placebo time.
Host 1Is two percent big? It sounds tiny.
Host 2For trained sprinters it is not tiny. The authors point to earlier work showing that the same sprinter's sixty-metre time varies by about one point four percent from day to day, so both average gains are larger than that normal wobble, and the high dose clearly so.
Host 1You said every runner improved, is that really all twenty-one?
Host 2On the high dose, yes, all twenty-one ran faster than on placebo; on the low dose, twenty out of twenty-one. That consistency is in some ways more convincing than the averages.
Host 1Where in the sprint did the time come from, the start or later on?
Host 2Already over the first ten metres the low dose was about five and a half hundredths faster and the high dose about seven and a half. The gains then kept building at thirty and at sixty metres, so the time was won along the whole distance measured, starting right at the gun.
Host 1And did doubling the dose help in every stretch?
Host 2Not clearly. Over cumulative thirty and sixty metres the high dose beat the low dose, by about three hundredths and five hundredths of a second, but at ten metres the extra two hundredths was not statistically convincing. And within each separate stretch, ten to thirty and thirty to sixty, the two doses did not differ significantly, so the extra advantage seems to pile up gradually rather than come from one moment.
Host 1What happened to reaction time at the gun?
Host 2Both doses shortened it by about seventeen to nineteen milliseconds, and the gap between the doses was only one and a half milliseconds, which is nothing. So for reacting to the signal, the low dose already got the whole benefit.
Host 1Seventeen milliseconds, is that something a sprinter would care about?
Host 2In context, yes: it is a meaningful chunk of a reaction that in trained sprinters usually lasts somewhere around a seventh of a second. Keep in mind it was measured as movement onset in the blocks, so it is one of the cleaner numbers in the paper.
Host 1And the push against the blocks, the rate of force development?
Host 2Here the pattern is different: all three measures, the overall horizontal push and the front and rear block separately, went up on the low dose and went up further on the high dose. So for how fast force builds in the blocks, more caffeine kept giving more.
Host 1Did they find that runners took longer strides or quicker steps?
Host 2No: over the first ten metres, average step length and step frequency did not change on either dose. The authors say that only rules out the simple story of longer or quicker steps; it does not mean the running itself was unchanged, because they only had averages over those first ten metres.
Host 1What about the jumps and the bike test after the sprint?
Host 2Jump height rose about five percent on the low dose and seven to eight percent on the high dose, for both the jump with a dip first and the jump from a held squat. On the bike, both doses raised peak and average power by roughly fifty to a hundred watts, but the high dose added nothing beyond the low one.
Host 1Hang on, the bigger dose was no better on the bike?
Host 2Correct, and peak power was even slightly lower on the high dose, about nineteen watts, though that difference was borderline and fell away in a second, more cautious analysis. The authors read it as no extra benefit from the higher dose, not as the low dose being superior.
Host 1Did the runners feel the sprint was easier on caffeine?
Host 2Yes: on the standard six to twenty effort scale, the typical rating was fourteen on placebo and on the low dose, but the spread shifted down, and it was thirteen on the high dose, and the odds of rating the sprint as harder were roughly four to six times lower on caffeine. Again, the high dose did not beat the low one here.
Host 1If I picture an athlete who jumped a lot higher on caffeine, did he also gain the most in the sprint?
Host 2No, and that is a telling result. Within individuals, how much someone's sixty-metre time improved was not linked to how much his jump, bike power, reaction time or block push improved, so no single measure explains the sprint gain.
What it means and for whom
Host 1So what is the main idea I should take from all of this?
Host 2That a dose of caffeine is not a single dial that turns everything up evenly. Reaction to the gun, feeling of effort and bike power topped out at three milligrams per kilo, while block push, jump height and overall sprint time kept improving at six.
Host 1If I were coaching a sprinter, would I just go for the higher dose then?
Host 2The paper would say: three milligrams per kilo was already enough to improve almost everything they measured in these men, and six added something for force and overall time. But they did not measure side effects, and from the wider literature we know higher doses bring more jitters, palpitations and poor sleep, so the extra few hundredths have a price the study cannot weigh.
Host 1And does any of this apply to someone who goes jogging on a Saturday?
Host 2Not directly: these were young, trained male sprinters, tested in a lab-like setting over sixty metres. What carries over is the general principle that caffeine can help short explosive efforts, but the size of the effect in an untrained person, or a woman, or an older runner, was not tested here.
The honest caveats
Host 1So what can we not conclude from this study?
Host 2Several things. We cannot say anything about the last forty metres of a race, nor about women, elite runners, or real competition with nerves and heats.
Host 1You mentioned blinding earlier. Could the runners have just known they had caffeine?
Host 2Possibly, and the authors acknowledge it. Six milligrams per kilo is a dose many people can feel, and they never asked runners to guess their condition, so expectation could have played a part, especially in the effort ratings; the hard numbers like times and block forces moved too, which helps, but does not rule it out.
Host 1Were there other problems with how it was set up?
Host 2A few. The jumps and bike test always came after the sprint, so they were measured on tired legs, and faster sprints on caffeine might have tired people differently; they also did not measure caffeine in the blood, or check whether regular coffee drinkers responded differently.
Host 1And is twenty-one people enough to trust the results?
Host 2For the big effects, the consistency across nearly every runner is reassuring. But there was no sample size calculation in advance; it was simply the number of eligible sprinters available, and the authors say it limits precision for small differences, especially between the two doses.
Host 1Was anything done in a way that should make a careful listener a little wary?
Host 2The trial was registered only after it had been run, which the authors put down to not realizing at the start that it counted as a clinical trial. Registering after the fact gives less protection against picking outcomes later, although the full data and analysis code were published with the paper, which counts in its favour.
Host 1Is there anything about the design that worried you statistically?
Host 2One technical point: with only three rotating orders rather than all six possible orders, they could not fully separate a lingering effect from the previous visit from the treatment itself. The week-long gaps, the lack of any order effects, and a check using only each runner's first visit all point the same way, but the authors say plainly that this reduces the concern rather than proving it away.
What to watch next
Host 1What would you want to see next from this line of research?
Host 2First, the full hundred metres, to see whether caffeine helps a sprinter hold top speed or only get there faster. Then step-by-step force measurements across the whole run, so you can see exactly how the push is changing rather than just the averages.
Host 1Would it help to test women and faster athletes as well?
Host 2Very much: everything here comes from young men around the eleven-second level, and both sex and training level could change the response. A proper look at side effects and sleep at the higher dose would also let coaches weigh the gain against the cost.
Sign-off
Host 1If you had to put the whole paper in one sentence, what would it be?
Host 2Caffeine made trained male sprinters faster over sixty metres, three milligrams per kilo did most of the work, and six added a little more for force and time but nothing for reaction or effort.
Host 1That is the Hastrup Review of Science deep dive. Every weekend this same feed carries the roundup, the week's new papers in one short episode, for those who prefer their science in smaller doses.
More from this week
- Sprint Caffeine, Lost Lean Mass and a Too-Eager Atlantic CollapseEpisode 14 · 10 min
- Weight-Loss Drugs and the Lean Mass That Goes With the FatDeep dive · 21 min
- The Waist the Scale Misses in Young PeopleDeep dive · 20 min
- Which Cholesterol Drugs Prevent a First Heart AttackDeep dive · 16 min
- Do Climate Models Give Up on the Atlantic Too Easily?Deep dive · 21 min
- Four Layers of Nickel, Superconducting Without the SqueezeDeep dive · 20 min