Two databases answer this question and they do not sound alike. One is a pool of controlled trials in which creatine and placebo produce almost identical complaint rates. The other is a government file containing hospitalisations and deaths. Both are real. Only one of them can establish cause. This is research journalism and contains no usage guidance, per our editorial standards.
Stream one: the controlled record
The largest assembly of it was published in 2025 and covers 685 human clinical trials of creatine supplementation, with 12,839 participants taking creatine and 13,452 taking placebo. Ninety-five per cent of the studies used creatine monohydrate, at an average of about 0.166 g/kg/day — roughly 12.5 g a day — for a mean of 64.7 days, in studies running up to fourteen years (Antonio, Kreider et al., J Int Soc Sports Nutr 2025, PMID 40198156).
| Measure | Placebo | Creatine | p |
|---|---|---|---|
| Studies reporting any side effect | 13.2% | 13.7% | 0.776 |
| Studies reporting GI issues | 4.3% | 4.9% | <0.001 |
| Studies reporting cramping / muscle pain | 0.9% | 2.9% | 0.008 |
| Participants reporting GI issues | 4.05% | 5.51% | 0.820 |
| Participants reporting cramping / muscle pain | 0.07% | 0.52% | 0.085 |
| Total frequency across participants | 4.21% | 4.60% | 0.828 |
The middle two rows are the interesting ones, and they show a distinction that decides how this literature should be read. Counting studies that mention a complaint, gastrointestinal issues and cramping both separate from placebo. Counting participants who actually reported one, neither does. A study-level count is a count of whether a symptom appeared in a paper at all; a participant-level count is a count of how many people had it. When the two disagree, the second is the one about people.
Across all 49 side effects evaluated there was no significant multivariate difference (p = 0.340), no significant difference in markers of renal function, and differences between groups generally within ±0.5%.
Stream two: the reports
The FDA's Center for Food Safety and Applied Nutrition Adverse Event Reporting System collects voluntary reports about foods, dietary supplements and cosmetics. It is public. Queried on 2026-09-11 for products whose brand name contains "creatine", it returns 210 reports.
The outcomes attached to them:
| Outcome | Reports |
|---|---|
| Other serious or important medical event | 117 |
| Hospitalisation | 76 |
| Visited emergency room | 55 |
| Visited a health care provider | 48 |
| Life threatening | 19 |
| Disability | 15 |
| Death | 10 |
And the most-reported reactions: nausea (16), vomiting (14), raised blood creatine phosphokinase (12), headache (12), raised heart rate (11), raised blood pressure (10), dyspnoea (10), rhabdomyolysis (10).
Printed like that, it reads as a safety signal. It is not one, and the reason is visible inside the same query.
What the report database cannot do
It has no denominator. Two hundred and ten reports out of how many people taking creatine? The database does not know, and neither does anyone else. A rate cannot be computed from a numerator.
It performs no causality assessment. A CAERS record is somebody's account that something happened while they were taking something. Nobody adjudicates whether the product caused the event.
It lists every product, not the cause. This is the decisive one, and the query shows it directly. Asking the same 210 reports which products they name returns, alongside the creatine products: fish oil (8 reports), multivitamin (6), glutamine (3). Those are not creatine products. They are there because a report captures everything the person was taking, and any of them could match a search as easily as creatine does.
A published analysis of 28.4 million adverse event reports quantified exactly this problem for creatine: mentions were rare at 0.00072% of all reports, 46.3% of the creatine-associated CAERS records had no creatine in the products listed at all, and 63% of those that did involved other forms of creatine or co-ingestion with other supplements or drugs. Our own query reproduces the shape of that finding independently.
And the comparison nobody runs. The same database, queried the same way on the same day for products whose brand name contains "whey", returns 293 reports — more than creatine. Whey protein is not the subject of a safety controversy. The number is a measure of how many people use a product and how visible its name is, not of how dangerous it is.
The kidney question, and why a blood test misleads
The most persistent specific worry has a chemical explanation that is worth stating precisely.
Creatine supplementation raises serum creatinine, the waste product of creatine metabolism. Serum creatinine is also the number laboratories use to estimate glomerular filtration rate — the standard measure of kidney function. So a person taking creatine can produce a blood test that looks like declining kidney function while kidney function has not changed. Hultman's 1996 loading study observed the mechanism directly: when supplementation stopped and muscle creatine declined, urinary creatinine excretion rose correspondingly (PMID 8828669).
The controlled evidence on actual renal function is consistent: no significant differences in renal markers across the 685-trial analysis, and renal indices unchanged in the 2026 meta-analysis of seven trials in 608 postmenopausal women.
This does not extend to people with existing kidney disease, who are systematically excluded from these trials and about whom the literature therefore says nothing.
The other specific claims
Hair loss. No trial that examined hair has found an effect. The claim traces to one 2009 hormone study that looked at no hair at all, and the 2025 trial that did look found nothing — the full chain is traced here.
Bloating and water retention. Creatine draws water into muscle cells; that is part of the mechanism, and early weight gain is real and measured. Subjective bloating is a separate claim and is one of the 49 side effects showing no significant excess over placebo. The literature on it is tiny: a PubMed search for creatine with bloating or fluid retention returns 39 records of any kind.
Dehydration and cramping. Cramping is the one complaint that separates at the study level and fails to at the participant level. That is the most honest thing that can be said about it: a weak, inconsistent signal that does not survive being counted properly.
Who assembled this evidence
The two largest safety syntheses on this page are authored by researchers who disclose substantial creatine-industry relationships — scientific advisory roles with named creatine brands, industry-sponsored research, creatine donations for research use, paid consultancy for supplement manufacturers, and in one case the chief executive of the society that publishes both journals.
That is not a reason to discard them. The 685 trials they pool were run by other people and are individually retrievable, and a null result is not what a sponsor commissions. It is a reason to say it out loud, which is the same standard this site applies to the efficacy literature and to every compound it covers.
The irony is worth noting: the controlled evidence here is far stronger than the evidence behind any of the research compounds sold for muscle growth, several of which have no human safety trial at all. The substance with 685 trials and a live government report file is the well-documented one. Documentation is what produces numbers to be alarmed by.
The honest summary
- In 685 controlled trials, creatine produced no excess of side effects over placebo on any of 49 measures.
- FDA's report database holds 210 creatine-associated reports including ten deaths, and cannot attribute any of them to creatine.
- Nearly half of those reports contain no creatine product at all, and the same file holds more reports for whey.
- The kidney worry rests on a blood marker that creatine raises without kidney function changing.
- People with existing kidney disease are excluded from these trials, so the evidence does not cover them.
Sources and dates
FDA CAERS queried live via the openFDA food/event endpoint on 2026-09-11 (searches on products.name_brand; counts on reactions.exact, outcomes.exact, products.name_brand.exact). PubMed searches the same day via E-utilities.
| Source | What it supplied |
|---|---|
| Antonio, Kreider et al., J Int Soc Sports Nutr 2025, PMID 40198156 | 685 trials, participant counts, the 49 side effects, the 28.4 million-report analysis |
| openFDA CAERS, 2026-09-11 | 210 creatine reports, outcome and reaction breakdowns, the co-listed products, the whey control (293) |
| Postmenopausal meta-analysis, J Int Soc Sports Nutr 2026, PMID 42141930 | Renal indices unchanged; adverse events mild and similar to placebo |
| Hultman et al., J Appl Physiol 1996;81(1):232–7, PMID 8828669 | Urinary creatinine excretion rising as muscle creatine declines |
| Antonio et al., J Int Soc Sports Nutr 2021;18(1):13, PMID 33557850 | The twelve standing questions and their disclosure statement |
| PubMed count, creatine + bloating/fluid retention | 39 records of any kind, 2026-09-11 |
Figures are reported as the sources state them. Nothing here is a recommendation or an instruction.