The source hierarchy
Not all evidence carries the same weight, and a page that cites a single small study as though it settled something is doing its reader a disservice. We work down this order.
1. Systematic reviews and clinical guidelines
Systematic reviews and meta-analyses that pool the available trials, and clinical practice guidelines from major professional bodies and national agencies. These represent the considered position of a field rather than one result, and they are our default starting point for any claim about what works.
2. Primary research
Peer-reviewed original studies, weighted by design: randomised controlled trials above cohort studies, cohort studies above case-control, and all of them above cross-sectional surveys. We note sample size, duration, population and funding when they affect how much the result should be trusted — and we say when a finding is in animals or cells rather than people.
3. Expert consensus and professional statements
Consensus statements and position papers, used where trial evidence is thin or absent. We label them as consensus rather than as evidence, because that is what they are.
What sits outside the hierarchy
Press releases, conference abstracts that never became papers, preprints that have not been peer reviewed, single anecdotes and commercial white papers are not used to support factual claims. Where we mention a preprint because it is genuinely newsworthy, we say explicitly that it has not been peer reviewed.
Which organisations we cite
The public-health, clinical and research organisations we lean on most are listed in full, with what each is good for, in our trusted organisations directory. In summary they include the CDC, the NIH and its institutes, MedlinePlus, the FDA, the US Preventive Services Task Force, the major disease-specific professional bodies, the WHO, the NHS and Cochrane.
Inclusion is not endorsement of every position an organisation holds, and these bodies do disagree with one another. Where they do, we present the disagreement and name the parties rather than picking one quietly.
We link to the primary source
Every source in a Sources list links directly to the original material — the guideline, the study, the agency page — not to a news article about it and not to another summary. External links open in a new tab and carry rel="noopener nofollow".
Where a study sits behind a paywall we link to it anyway, and where a free full text or an agency summary of the same work exists we link to that too, so you are not required to pay to check us.
The two rules we do not bend
We never cite a study we have not read
Not the abstract alone where the full text is available, and never a citation copied from another article's reference list. Abstracts routinely overstate what a paper found, and inherited citations propagate errors — including citations to papers that say the opposite of what they are cited for.
We never invent a source
No fabricated citations, no invented statistics, no attributing a claim to an organisation that never made it. This rule is stated explicitly because AI-assisted drafting makes plausible-looking fake references trivially easy to produce. Every reference on this site is located, opened and read by a person before it appears. See the AI section of our editorial policy.
When a source goes out of date
Guidelines are revised, drug labels change, and screening recommendations move. When a cited source is superseded, the page is updated to the current version and the change is dated. If a source is withdrawn or a study is retracted, we remove it and correct anything that relied on it, following our corrections policy.
If you find a source on this site that has been superseded, retracted or misrepresented, tell us at corrections@healthcaptain.com.