Skip to content
Join free
Back

Biohacker-curious

Why one study, or one mouse, is not enough

A mouse is a clue. A single human trial is a start. A claim worth acting on needs a whole body of evidence behind it.

Checked against the sources ยท Updated 17 Nov 2026

In short

  • One study, or one animal study, is enough to settle a health claim
  • This page explains how evidence is weighed, not a health claim, so it carries no cranes. GRADE's high certainty is a body of trials, and in one review only about a third of highly cited animal findings were later confirmed in a human trial.

Why one study, or one mouse, is not enough

The message usually arrives in the evening, from a Tita who means well: a screenshot, a headline, and the words "a new study shows." This time it is a supplement, or a fruit, or something mice did in a lab. Someone asks where to buy it.

You want the headline to be true. One study is a clue, not a conclusion. Not a mouse study, and not one good study in people. A claim worth rearranging your life for needs a whole set of evidence behind it.

What happened to the famous animal findings

Daniel Hackam and Donald Redelmeier asked what happens to exciting animal findings once someone tests them in people. They followed highly cited animal studies, the ones other scientists kept referring to, and not one of them was a negative result. If any mouse research was going to hold up in humans, it was this batch.

About 37% were later confirmed in a trial in people. About 45% had never been tested in people at all. The rest were contradicted when people were finally studied. The authors' own line is the one to keep: patients and doctors should stay cautious about trusting even prominent animal research. This was the famous, positive work. An ordinary animal paper, the kind behind most headlines, may do worse.

One study in people is still one study

The people who review health evidence for a living do not stop at a single paper. They look at the whole set of studies on one question. The picture they trust most is several careful trials in people that agree. A doctor's write-up of what happened to a few patients can be a hint. It is not an answer.

That trust drops when the studies were run loosely, when the results disagree, when the studies do not quite fit the question, when they are too small to be sure, or when the unflattering studies never got published. A single human trial, even a good one, is the beginning of the question.

The essay everyone half-quotes

John Ioannidis's essay "Why Most Published Research Findings Are False" usually gets waved as proof that you can ignore the research. It is a model, not a head count. He did not tally every paper ever published. He laid out when a research claim is likely to be true, and the answer depends: on whether the study was big enough to see what it was looking for, on bias, on how many relationships were tested, and on how many of those relationships were real to begin with.

In that model, findings are less likely to hold when studies are small, when the effects being chased are tiny, when researchers have a lot of freedom in the analysis, when money or reputation is riding on the result, and when many teams are racing to be first. That is not an argument that science is broken. It is an argument that some corners deserve more skepticism. He also wrote that a well-run trial, big enough for the question and started for a genuine reason, can be right most of the time.

A later correction to the essay was not read again, so the account above follows the original text.

Honest limits

None of this means animal research is useless. A lot of human medicine started there. A mouse result is a starting clue, and the leap to people fails more often than the headlines suggest.

The 37% figure comes from an unusual sample, the most famous and most positive animal work of its time. It is the right sample for asking whether even the best translates. It is not the rate for all animal research.

A single excellent trial is not meaningless. Sometimes one big, careful study matters. The myth is the version that says one study, or one animal study, is enough to settle a health claim.

How a claim gets read here

Guidelines and large reviews come first, because they gather many studies. Trials in people come next. Studies that watch people over time come after that. Lab and animal work comes last, as a clue about where to look, not as proof of what works in you.

Three questions for the next "new study"

  • People or animals? If it is mice, it is too early to act on.
  • One paper or many? Look for reviews and guidelines that gather the whole set.
  • Has anyone repeated it? A finding nobody has tried to repeat is still waiting for its real test.

---

Sources

  • Hackam DG, Redelmeier DA. Translation of research evidence from animals to humans. JAMA. 2006. https://jamanetwork.com/journals/jama/fullarticle/203577
  • Cochrane Handbook, chapter 14 https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-14
  • Ioannidis JPA. Why most published research findings are false. PLOS Medicine. 2005. https://doi.org/10.1371/journal.pmed.0020124

General guidance, not medical advice or a diagnosis. Talk with your doctor about your own health.

How strong is the evidence?

Not rated yet

What the research says: Hackam and Redelmeier followed 76 highly cited animal studies. About 37% were later replicated in a human randomized trial, 18% were contradicted, and 45% were still untested. GRADE rates certainty for a body of evidence. Several low-bias randomized trials are the high-certainty example. Case reports sit at very low certainty.

What it doesn't say: This does not mean animal research is useless, or that a single excellent trial is meaningless. The animal sample was famous positive work, so ordinary animal papers may translate even less often. Ioannidis did not count the fraction of all papers that are false. He showed how a model makes false claims likely when studies are small, effects are tiny, or many ideas are tested at once.

Hackam DG, Redelmeier DA. Translation of research evidence from animals to humans. JAMA. 2006.

How we rate: the Flamel Method

Try this

When a post says 'a new study shows,' ask three things: people or animals, one paper or many, and whether anyone has repeated it.

That counts.

Get delightful living tips directly in your email

Join free

Sources

  1. Hackam DG, Redelmeier DA. Translation of research evidence from animals to humans. JAMA. 2006.
  2. Cochrane Handbook, chapter 14
  3. Ioannidis JPA. Why most published research findings are false. PLOS Medicine. 2005.

Last reviewed by Flamel Staff on 17 Nov 2026

Flamel is not medical advice and does not diagnose. If you have symptoms or readings in the red zone, see a doctor.