The library

How we read studies

Every study page here names the kind of study it describes. That word matters more than the result does, because it tells you what the study was capable of showing in the first place. This page explains each one in plain language.

Last updated 30 July 2026

We do not grade evidence, rank studies or award stars. A design is not good or bad. It is suited to some questions and not to others, and saying which is more useful to you than a score would be.

The designs, one at a time

Randomised controlled trial also called RCT

People are split into groups by chance, one group gets the thing being tested and the other does not, and both are followed for the same period.

Good for: Because the split is random, the groups tend to be similar in every way except the one being tested. That makes this the clearest way to see whether the thing itself caused the change.
Careful with: Trials are usually small, usually short, and usually run in people who agreed to be in a trial. A result over 12 weeks in 40 men may not hold over 12 years in the general population, and some things cannot ethically or practically be randomised at all.

Cohort study also called prospective or longitudinal study

A group of people is followed forward over time, and researchers record what they do and what happens to them.

Good for: It works over long periods and in real life, so it can pick up slow effects and rare outcomes that a short trial would miss.
Careful with: Nobody was assigned to anything, so the men who sleep well may differ from the men who do not in a hundred other ways. Researchers adjust for the differences they measured. They cannot adjust for the ones they did not.

Case-control study

Researchers start with people who already have an outcome, pair them with similar people who do not, and look backwards for differences.

Good for: It is efficient for uncommon outcomes, because you do not have to follow thousands of people and wait.
Careful with: It leans on records and on memory, and both are imperfect. It also cannot show which came first, only that the two things sit together.

Cross-sectional study also called a snapshot

Everything is measured at one point in time, across a group of people.

Good for: It is a good way to describe how common something is, and to spot a pattern worth investigating properly.
Careful with: A snapshot cannot tell you what came first. If men with more body fat have lower testosterone in a single measurement, that is a real observation, and it says nothing on its own about which one moved the other.

Systematic review and meta-analysis

Researchers search for every study on a question using a stated method, and a meta-analysis then combines their numbers into one pooled result.

Good for: It gives you the whole field rather than one lucky result, and it makes disagreement between studies visible.
Careful with: A pooled result is only as good as what went into it. If the underlying studies are small, inconsistent or all share the same flaw, the pooled number inherits that. Combining studies does not make a finding true, it makes it easier to see.

Narrative review

An expert summarizes a field in their own words, without a stated method for choosing which studies to include.

Good for: It is often the clearest explanation of how a mechanism is thought to work, and a good place to understand the shape of a subject.
Careful with: Because the selection is not systematic, it reflects the author’s reading and judgment. Treat it as an informed argument rather than as a count of the evidence.

Animal study

The work is done in animals, usually rodents, often to isolate a mechanism that cannot be isolated in people.

Good for: It can show how something works, at a level of control and detail that human research rarely allows. Human trials frequently follow animal work, so this is often where a question starts.
Careful with: Animals are not men. Doses, timescales and biology all differ. A missing human study means the human question is open, not that the finding has been disproved.

Laboratory study also called in vitro

The work is done on cells or tissue outside a living body.

Good for: It is the most direct way to watch a biological process happen and to test what changes it.
Careful with: A cell in a dish has no bloodstream, no hormones arriving from elsewhere and no immune system. What happens there is a starting point for a question, not an answer about a whole person.

Two words worth knowing

Association. Two things move together. Taller men own more large coats, and coat size does not cause height. Most research on diet, sleep and lifestyle in large groups is about association, which is genuinely useful and is not the same as cause.

Causation. One thing produces the change in the other. Showing it usually takes a randomised trial, or a large body of evidence pointing the same way from several directions at once.

What else to look at on a study page

Who was studied. Age, health and country change how much a result has to say about you. A finding in 65 year old men with diabetes is not a finding about a healthy 34 year old.
How many people. Small studies swing. A striking result in 18 men is a reason to watch the question, not to act on it.
How long it ran. Six weeks tells you about six weeks.
Who paid. Funding does not make a study wrong, and it is worth knowing. When a paper does not say, we write that it was not reported rather than guessing.
What the study did not find. A result of no difference is a real result. We report those in the same voice as the positive ones, because a library that only tells you what worked is not describing the evidence.

The honest summary

One study is one piece of evidence. It can be well run and still be wrong, and it can be modest and still be the best information available. What we can do is describe each one accurately, tell you what its design can support, and link you to the paper so you can look for yourself.