Evidence
How claims on this site are graded, what counts as supporting evidence, and why the absence of evidence is reported as a finding rather than omitted.
The grading rule
A verdict is attached to a specific claim, never to a category. "Whole-body MRI detects cancer" and "whole-body MRI screening saves lives" are different claims with different evidence, and they get different verdicts even though both concern the same machine.
What counts
- Systematic reviews and meta-analyses of observational or randomised studies, with the heterogeneity reported.
- Randomised controlled trials with clinically meaningful endpoints, not just biomarker changes.
- Large cohort studies with enough follow-up to say something about outcomes rather than detection.
What does not count
- Detection rates presented as if they were survival benefits. Finding something earlier is not the same as changing what happens, and it can be actively harmful if it leads to treatment of findings that would never have caused harm.
- Before-and-after measurements without a control group.
- Biomarker changes with no demonstrated link to outcomes people care about.
- Provider-published success rates with no denominator, no follow-up period, and no definition of success.
How verdicts change
A verdict is revised when better evidence appears, and the article keeps the date of the sources it relies on. Where a claim is widely repeated but unsupported, it gets an explicit insufficient rather than being quietly left out.