Creatine for cognition: what the 2024 meta-analysis actually shows.
Fitness feeds keep claiming creatine does more than build muscle. They say it sharpens your mind, lifts your mood, and fixes your sleep. Most of that noise traces back to one 2024 meta-analysis. The study pooled sixteen randomized trials of creatine versus placebo on cognitive tests. The headline went everywhere. The methods went nowhere.
I spent an evening with the supplementary material. Here is what the data shows.
What the paper says
Creatine supplementation produced a small but statistically significant improvement in memory tasks (g = 0.31, 95% CI 0.18–0.44), with larger effects in older adults and under cognitive stress.
The result is real. The confidence interval clears zero. The funnel plot looks symmetric, so publication bias is not doing the heavy lifting. The paper is fine. The retelling is the problem.
The thing that gets lost
An effect size of g = 0.31 is small. In the framework psychologists use, 0.31 sits at the lower end of what a careful experiment can even detect. For scale, one good night of sleep moves next-day cognition by g = 0.6 to 0.9, depending on the task. Pooled creatine is about a third of one good night.
The subgroups are where the loud claims actually live. In people over 65, the effect roughly doubles. In sleep-deprived young adults, it doubles again. In healthy twenty-somethings sleeping eight hours, the effect is statistically a rounding error. The "small average" hides a real signal in stressed populations and almost nothing in everyone else.
| Subgroup | Effect size (g) | What I would tell someone |
|---|---|---|
| Healthy young adults, well rested | 0.08 (n.s.) | Skip it for the brain. |
| Sleep-deprived adults | 0.42 | Plausible for shift work, jetlag, new parents. |
| Adults 65+ | 0.55 | Biggest signal. The intervention is cheap. |
| Vegetarians / vegans | 0.48 | Lower baseline; topping up does more. |
Three things that bug me
The doses were all over the place. Some trials used 3 g/day for two weeks. Others ran a 20 g loading phase, then 5 g/day for three months. The meta-analysis pools them as one intervention. They are not. Muscle saturates at about 28 days on 5 g/day. Brain creatine saturates more slowly, and may not saturate at all at low doses. Sub-saturating studies dragged the pooled estimate down.
"Memory" did a lot of work. The bucket includes digit span, delayed word-list recall, and logical-reasoning subtests. Those are different systems, lumped together. Convenient. It also hides the more interesting question: which kind of cognition actually budges.
Blinding was thin. Creatine has a gritty texture. A third of trials used inactive cellulose for placebo. The rest used dextrose or maltodextrin. Neither feels like creatine. Almost no trial checked whether participants could tell. Some could.
If you want to see how I gut-check papers, run the numbers through the paper sanity check: p < 0.001, n = 492, d = 0.31.
The mechanism, briefly
Creatine tops up the brain's phosphocreatine pool. That pool buffers ATP when demand outruns supply. The brain uses about 20% of resting energy, on 2% of body mass. When supply gets squeezed by age, sleep loss, altitude, or depression, the buffer matters. When supply is plentiful, a bigger buffer sits there doing nothing measurable.
That explains why subgroup effects dwarf the average. Creatine helps under stress. Pool the stressed and the unstressed together and you blur both into a small grey number that describes nobody.
Bottom line
Good study. Told straight, it says: a small average effect, a real benefit in stressed populations, almost nothing for healthy rested young people. Somewhere between the journal and your feed, "small and conditional" became "limitless." It usually does.