The limits of the evidence on molecular hydrogen

By Hydrogen Studies Editorial Team · Updated November 3, 2026

This site lists 2,276 studies on molecular hydrogen. A number that large is easy to mistake for a settled body of evidence. It is not. This page shows, with figures computed from the database as of November 4, 2026, how little of that research is in people, how small and short the human trials are, how lopsided the published results are, where the work comes from, and how much of it has never been repeated. The prose was last reviewed on November 3, 2026.

Most of the research is not in people

348 of the 2,276 studies (15%) involved human participants. 1,070 (47%) were done in animals, 276 (12%) on cells or tissue in a laboratory, and 426 (19%) are reviews or commentaries that report no new data. Only 126 studies are randomized controlled trials, the design that can show whether hydrogen caused an effect in people. Results in animals and cells often do not carry over to people, so a benefit shown in a rat or a dish is a reason to run a human trial, not a finding about human health.

The human trials are small and short

Of the 348 human studies, 297 record how many people took part. The median is 23 participants: 169 studies (57%) had fewer than 30 people and 258 (87%) fewer than 100; 39 had 100 or more, the largest 2,364. A trial of 20 people can only detect large effects and will often produce a positive or negative result by chance.

Duration is recorded for 258 human studies. The median is 28 days: 43 lasted a single day or a single dose, 106 up to four weeks, 90 between one and six months, and 19 longer than six months. Nothing in this database tells you what years of daily hydrogen water does, because almost no one has studied it.

The results skew positive

For 1,884 studies the pipeline records the direction of the main result. 1,764 (94%) are positive; 120 (6%) are neutral, negative or mixed, 31 of them in people. A field in which 94% of published studies succeed is not a field in which hydrogen works 94% of the time; it is a field in which null results are less likely to be written up, submitted and published. We list the exceptions rather than bury them: null and negative results.

Where the research comes from

China accounts for 1,097 studies (48%), Japan accounts for 523 studies (23%) and United States accounts for 121 studies (5%). Concentration in a few research communities matters: trials recruit from one population, share methods and reviewers, and publish in a connected set of journals, so agreement among them is weaker evidence than agreement among independent groups would be. 94% of the studies were published since 2010, so this is a young literature.

• China: 1,097 studies (48%)
• Japan: 523 studies (23%)
• United States: 121 studies (5%)
• South Korea: 80 studies (4%)
• Taiwan: 72 studies (3%)
• Serbia: 55 studies (2%)

Who paid for it is mostly unrecorded

Funding is recorded for 166 studies and declared conflicts of interest for 187; extraction so far has covered human studies. For the rest, a blank field means not yet recorded, not "no conflict". Device makers fund some hydrogen research, and this site is itself funded by one (see the editorial policy). When funding is recorded, industry sources are shown as such, including any study funded by Echo Technologies. Only 41 of the 348 human studies mention a trial-registry identifier in their abstract, which is one rough sign of how often outcomes were pre-specified.

Few findings have been replicated

Randomized trials in people cover 228 health-condition tags in this database, but only 48 of those tags have two or more randomized trials. For most conditions the entire randomized evidence is a single trial, usually small and short, that no independent group has repeated. A result that has never been reproduced should be read as a lead, not a conclusion.

Doses and delivery vary, and often go unrecorded

Hydrogen was given as water, inhaled gas, saline, baths and tablets, at concentrations and for durations that differ by orders of magnitude between studies, and no trial has compared doses head to head. The hydrogen concentration is recorded for 550 studies (24%). Two studies that both report "hydrogen water" may have delivered very different amounts of hydrogen, so results do not add up across studies the way the headline count suggests.

What this means when you read a study here

• Check the evidence label first. Animal and laboratory results are not findings about people.
• Look at the number of participants and the duration. Fewer than 30 people, or a few weeks, cannot support a general claim.
• Ask whether anyone has repeated it. Most topics rest on one trial.
• Notice who paid, and that for most studies we do not yet know.
• Read our hub introductions, which grade the evidence per topic as "Low" or "Very low" for these reasons, and the methodology for how the numbers above are produced.
• If your question is simply whether hydrogen water is a scam, here is our answer.

Every figure on this page is recomputed from the database when the page is built (as of November 4, 2026); the text is reviewed when the method changes. Corrections: contact us.