Anecdote
A personal experience can generate a question. It cannot establish cause, frequency, or general safety.
DIRECT
Evidence map
A claim is not a result. A result is not a verdict.
Use the full chain—from question and protocol to data, interpretation, replication, and regulatory status.
Evidence ladder
Different evidence answers different questions.
Observational signal
An association can be informative, but self-selection, expectancy, reporting, and other differences may explain it.
Controlled experiment
Randomization, a credible comparator, prespecified outcomes, and transparent attrition strengthen causal inference.
Replication and synthesis
Independent studies and systematic review help show whether an effect persists across settings and methods.
Regulatory conclusion
Approval requires the applicable legal and evidentiary process. A trial, publication, or breakthrough designation is not approval.
Common interpretation traps
Five places a confident headline can outrun the record.
- Expectation. Participants may infer assignment from subjective effects.
- Self-selection. People who choose to participate or self-report may differ from others.
- Outcome switching. A positive secondary measure is not the same as a prespecified primary success.
- Short follow-up. A brief observation window cannot establish durable benefit or long-term safety.
- Generalization. Screened study participants and controlled protocols may not resemble unsupervised use.
Claim audit
Make the statement smaller until the evidence fits.
AskRecordDo not infer
Who was studied?Eligibility, exclusions, sample size, settingThat the result applies to everyone
What was tested?Substance, formulation, protocol, supervisionThat another product or context is equivalent
Against what?Placebo, active comparator, usual care, or noneThat change over time proves causation
Which outcome?Primary measure, timing, effect, uncertaintyThat statistical significance equals clinical importance
What happened next?Adverse events, attrition, follow-up, replicationThat missing or short-term data prove safety
Next tool
Put one study through the reader.
Capture what the record supports—and the question it leaves open.
Open the study reader