Before sharing a health claim, find the source behind it and check whether the headline describes the same people, intervention and outcome. A reassuring logo or an impressive percentage is not enough. That is how to evaluate health news in practice: ask what was measured, how it was compared and what remains unknown.
Reported from: MedlinePlus guidance on evaluating health information, NCCIH’s Know the Science resources and ClinicalTrials.gov’s description of its registry.
Find the claim’s actual source
MedlinePlus recommends checking who runs a health website, why it exists, where its information comes from and whether it is current. A product seller may accurately report a study, but its commercial interest remains relevant. A clinician’s title also cannot substitute for a source that supports the particular claim.
Follow the link. Does it reach the paper, guideline or regulator’s decision being described, or just an organization’s homepage? Look for a date, a named issuing body and an account of how the recommendation applies. A real citation can still be attached to a sentence it does not support.
Read the methods before treating the result as an answer
NCCIH’s guide to journal articles directs readers through methods, results and discussion. Ask who participated, how many people were included, what comparison was made and how long follow-up lasted. Laboratory findings do not directly demonstrate benefits for patients. An observational association can have explanations other than the exposure being studied.
Random assignment can strengthen a causal comparison, but it does not erase missing data, bias or uncertainty. A review is only as useful as its methods and included evidence. Check whether the measured outcome was something people experience, such as symptoms or hospitalizations, or an intermediate marker. Our chia-seed article illustrates why a change in a measurement is not automatically proof of disease prevention.
Make the size of the effect understandable
Here is a hypothetical example, not a study result: a risk falling from two people in 100 to one in 100 is a 50% relative reduction and a one-percentage-point absolute reduction. Both descriptions can be mathematically correct; the second makes the starting risk visible. Ask for the time period and the number of people as well as the percentage.
Benefits also need a comparison with harms and uncertainty. A point estimate does not tell the whole story. Our PCOS fertility report uses actual live-birth counts and explains why the group result is not an individual forecast.
A registered trial is not a proven treatment
ClinicalTrials.gov explains that the U.S. government does not approve the safety and science of every listed study. A registration describes research; it is not a seal of effectiveness. A completed status does not, by itself, tell you the results. Look for posted results or a matching publication, and check whether the outcomes reported match the question being asked.
Our Trial Tracker methodology explains the limits of the publication’s own registry snapshot. A locally displayed record can lag behind the source registry.
What to do before sharing
Keep the qualification attached to the claim: for example, “small study,” “observational association” or “registration without results.” If you cannot trace the evidence, say that it remains unverified rather than repeating the headline as fact. For a question about your own symptoms, medication or screening, take the source to a qualified clinician. News can help frame that conversation; it cannot make the decision for you. Pausing to evaluate health news before sharing it protects the people who trust you.
