A relative drop alone doesn’t say how much it helps. Ask: out of how many?
The claim
A 25% lower risk means it helps one in four people.
What the evidence says
A made-up example: a condition affects 4 in 1,000 people; with a made-up treatment, 3 in 1,000. That is a 25% relative drop and 1 fewer person in 1,000 in absolute terms. If 400 in 1,000 were affected, the same 25% would mean 100 fewer people. Same percentage, a hundred times the effect.
Why it matters: a Cochrane review of 35 studies (83 comparisons) found that relative risk reductions were perceived as larger and more persuasive than the same effect stated as absolute reductions, without better understanding. Natural frequencies (“1 in 1,000”) were understood better than probabilities. Researchers who study this recommend absolute risks and natural frequencies.
The numbers
- Akl 2011 (Cochrane, 35 studies, 83 comparisons): relative risk reductions were perceived as larger (SMD 0.41) and more persuasive (SMD 0.66) than absolute ones; understanding little or no different (SMD 0.02); natural frequencies understood better than probabilities (SMD 0.69).
- Gigerenzer 2007: recommends frequency statements, absolute risks and natural frequencies.
- Arithmetic example (made up): 4 → 3 in 1,000 is 25% relative and 1 in 1,000 absolute; 400 → 300 in 1,000 is 25% relative and 100 in 1,000 absolute.
Example numbers are made up.
Verdict: Not true
A relative drop alone doesn’t say how much it helps. Ask: out of how many?
Why only this far: This is arithmetic, not a health study: the same 25% can mean 1 fewer in 1,000 or 100 fewer in 1,000. Evidence level A refers to the Cochrane review on how people perceive the formats.
Evidence level A: Randomized trials or meta-analyses of randomized trials. How verdicts work.
What we don’t know
Whether clearer numbers lead to better decisions. The review didn’t test behavior.
Sources
Checked 15 Oct 2026. Every link opens the original paper or its PubMed record.
- Akl et al., Cochrane Database Syst Rev 2011Cochrane review (35 studies, 83 comparisons)DOI 10.1002/14651858.CD006776.pub2 · PubMed 21412897
- Gigerenzer et al., Psychol Sci Public Interest 2007ReviewDOI 10.1111/j.1539-6053.2008.00033.x · PubMed 26161749
Follow for the reel on Instagram
Educational content, not medical advice. This check rates a claim, not you or your health. Talk to a doctor about your own situation. Found a mistake? Write to hello@mostlyproven.com.
