How we choose sources, verify every citation, describe the strength of evidence and test our tools.
We cite peer-reviewed research first: randomized trials, systematic reviews and meta-analyses where they exist, then large cohort studies, then smaller or earlier work. Clinical guidelines (for example from the Endocrine Society, KDIGO or the US Preventive Services Task Force) are cited for recommendations. For regulatory events, such as an FDA decision, we link the official notice or a named news report.
We do not cite secondary summaries, product pages or opinion pieces as evidence for a health claim.
Every reference is matched to its PubMed record and linked by PMID and DOI, so you can open the source in one click.
When we write or re-verify an article, each specific claim is read against the abstract of the paper it cites, and against the full text when the abstract is silent. A number in our text must appear in the source; if a claim and its source disagree, we change the claim, replace the source or remove the claim.
Automated checks run on every update: each citation number must point to a listed reference, references are screened for pointing at the wrong paper, and pages cannot claim more citations than they list. We are working through every article claim by claim; changes that alter a fact, a figure, a recommendation or a source are recorded in our corrections log.
We say what kind of study supports a claim. A randomized trial in people can show cause and effect; an observational study shows an association, which may reflect other differences between groups; animal and cell studies are labelled as such and are not presented as proven human benefit; mechanisms and expert opinion are described as hypotheses.
Where possible we give absolute numbers and the size of the effect, not just whether a result was statistically significant, and we report when a trial found no benefit.
Calculators that implement a published model are tested against a reference implementation of that model; for example, our PhenoAge calculator must reproduce the formula from Levine et al. (2018) exactly before any update goes live. Questionnaires that are not validated models are labelled as estimates. The supplement interaction checker lists only interactions with published evidence, each linked to its source.
IQ Healthspan is educational. It is not medical advice and cannot account for your health history or medicines. Please talk to a clinician before changing treatment, and see our medical disclaimer.