How to Read Health Claims and Evidence
Stephen McCarthy • Educational commentary
Narration: Microsoft Zira Desktop (generic synthetic voice, not the author’s voice).

How to Read Health Claims and Evidence

This video uses a generic synthetic narrator, not the author's voice. A health headline can sound decisive while leaving out the details that make a finding useful. The aim of this guide is to help you ask better questions about research. It does not recommend a treatment or replace a conversation with a qualified health professional. Start by translating the headline into a clear question: who was studied, what was compared, which outcome was measured, and over what period? Keep those four details beside you as you read.

Sources:
https://www.nccih.nih.gov/health/know-science/how-to-make-sense-of-a-scientific-journal-article/overview

Identify the design

Next, identify the study design. In an observational study, researchers examine what happens without assigning the exposure or intervention being compared. In a randomized trial, chance determines assignment to study groups. Randomization can make a treatment comparison more informative about cause and effect, but the conduct of the trial still matters. An association means two things vary together. It does not, by itself, show that one caused the other. People who choose a certain activity may also differ in sleep, income, health, or access to care. Researchers may address some of those differences, but a headline rarely tells you how well.

Sources:
https://www.nih.gov/about-nih/science-health-public-trust/tools/understanding-clinical-studies
https://grants.nih.gov/policy-and-compliance/policy-topics/clinical-trials/clinical-trial-besh-or-observational-study-involving-humans

Ask what the numbers mean

Now look at the size of the result. Imagine a completely hypothetical comparison: twenty people out of one thousand experience an outcome in one group, and ten out of one thousand in another, over the same period. The relative reduction is fifty percent. The absolute difference is ten people per thousand, or one percentage point. Both descriptions are mathematically correct. They answer different questions. The example does not describe a real treatment. It shows why a percentage needs a starting risk, a denominator, and a time period. Ask for the numbers in each group, then look for harms and burdens as well as benefits.

Sources:
https://www.cancer.gov/about-cancer/screening/patient-screening-overview-pdq

Keep uncertainty visible

A study reports an estimate, not a guarantee. A confidence interval helps communicate statistical precision under the study's assumptions. A wide interval may leave room for importantly different interpretations. A small p-value does not tell you how large an effect is, whether it matters in daily life, or the probability that the claim is true. Also read the limitations. Were many participants missing from the final analysis? Were outcomes measured consistently? Did the researchers test many questions and highlight only a few results? You do not need to solve every statistical problem yourself. Recognizing an unanswered question is already better than treating a single number as certainty.

Sources:
https://doi.org/10.1080/00031305.2016.1154108
https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-14

Check the outcome and the wider evidence

Check what the researchers actually measured. A change in a laboratory marker is not automatically the same as an improvement in symptoms, function, or survival. Some surrogate measures have strong support in particular settings; others remain uncertain. Then ask whether the participants and follow-up period fit the question you care about. Finally, look beyond one paper. A systematic review brings together studies using an explicit method. It is still important to examine the quality and relevance of the evidence it contains. Agreement across sound studies can be informative. Repeating a weak claim across many websites does not create independent evidence.

Sources:
https://www.fda.gov/drugs/regulatory-science-action/public-posting-comprehensive-surrogate-endpoint-table-cder-and-cber-regulated-products
https://www.nia.nih.gov/health/clinical-trials-and-studies/what-are-clinical-trials-and-studies
https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-14

Turn a headline into useful questions

Before sharing a claim, check who published it, who paid for the work, and whether the source supplies a date and a link to the research. A study's presence in a registry is not a government endorsement of its safety or conclusions. Keep a brief reading note: the question, the design, the result, the main limitation, and what remains unknown. That note is more useful than a saved headline alone. Bring questions about personal care to an appropriate health professional. The companion article supplies the source links and a worked example. The goal is a more careful conversation about evidence, with uncertainty stated plainly.

Sources:
https://www.medlineplus.gov/evaluatinghealthinformation.html
https://clinicaltrials.gov/