01 • Bottom line
A study answers one narrow question.
A human study is not simply “positive” or “negative.” Its answer depends on the people enrolled, the comparison, the outcome, the study length, and the way uncertainty and missing data were handled.
Start with: “What exactly happened in this group, under these conditions, for this amount of time?”
02 • The five questions
Turn the headline back into a research question.
Who was studied?
Count the participants, then check age, sex, health, training status, diagnosis, and location. A finding in one narrow group may not apply to everyone else.
What was compared?
Look for the actual intervention, dose or exposure, control group, randomization, and masking. A before-and-after change without a useful comparison is harder to interpret.
What did they measure?
Find the prespecified primary outcome. A lab marker, scale score, symptom report, performance test, and real-world health event are not interchangeable.
How much changed?
Read the numbers for every group—not only the percentage in the headline. Ask whether the difference was large enough to matter, not merely detectable by a statistical test.
Where does the answer stop?
Check study length, dropouts, missing data, adverse events, funding, conflicts, and the authors’ limitations. One short study rarely settles long-term benefit or safety.
03 • Read the numbers
“Significant” does not tell you whether the change was important.
A p-value describes how compatible the observed data are with a specified statistical model. It does not tell you the probability that a result is true, or whether a difference is large, useful, durable, or worth a risk or cost.
- Read the starting and ending values for every group.
- Look for the absolute difference, not only a relative percentage.
- Check the confidence interval: a wide range means the size of the effect is less certain.
- Confirm whether the reported outcome was the prespecified primary outcome.
- Read adverse events and withdrawals beside the benefit result.
04 • Warning signs
Slow down when the claim is larger than the study.
- A cell or animal experiment is described as proof of a human benefit.
- A tiny pilot study is treated as a final effectiveness or safety answer.
- The headline gives a dramatic percentage but no group totals or absolute change.
- The study measured a marker, while the post promises a real-world outcome.
- Only the abstract is quoted and the limitations, harms, or full paper are missing.
- One study is presented as settled science without replication or a broader review.
05 • Mini example
Watch a strong headline become a modest answer.
Imagine a four-week study of 36 trained adults. A social post says “Compound X improved recovery by 20%.” The paper shows a small difference on a self-reported soreness scale, wide uncertainty, several dropouts, and no long-term follow-up.
The honest takeaway is not “it works” or “it does nothing.” It is: this small, short study found a possible signal on one subjective outcome in a specific group. The size, reliability, long-term meaning, and safety of that signal remain uncertain.
06 • The 60-second check
Use this before you repeat a research claim.
- People, animals, cells, or a computer model?
- How many people, and were they similar to the people in the claim?
- Randomized and controlled—or only observed before and after?
- What was the primary outcome, and how was it measured?
- How large was the difference in absolute terms?
- How long did the study last, and how many people dropped out?
- What harms, limitations, funding, and conflicts were reported?
07 • Sources
Open the reading guidance.
Sources last reviewed August 17, 2026. These resources explain general research appraisal; they do not rate a particular intervention.