How We Review Evidence
The goal
Medicine changes because evidence changes. Our job is to translate medical research into understandable, cautious explanations — without overstating certainty.
What we look for first
When evaluating a topic, we typically start with:
- Clinical practice guidelines (and who authored them)
- Systematic reviews and meta-analyses
- High-quality trials and replication
- Large, well-designed observational studies where trials aren't feasible
Types of evidence (and what they're good for)
Randomized controlled trials (RCTs)
Good for testing whether an intervention causes an outcome — but still limited by who was studied, how long, and what outcomes were measured.
Observational studies
Useful for real-world patterns and long-term outcomes, but vulnerable to confounding (people who do X may differ from people who don't).
Systematic reviews and meta-analyses
Often strongest overall — if the included studies are high quality and comparable. A meta-analysis can amplify bias if it combines biased studies.
Mechanistic / early research
Helpful for understanding "how it might work," but rarely enough to justify confident claims about real-world outcomes.
Common interpretation pitfalls
We try to protect readers from:
- Relative risk without absolute risk ("doubles risk" can still be tiny)
- Surrogate outcomes (a lab marker improves but patient outcomes don't)
- Short follow-up (benefits and harms can change over time)
- Subgroup over-interpretation (small slices of data are noisy)
- Single-study hype (replication matters)
How we communicate uncertainty
We use language intentionally:
- May = plausible but uncertain
- Likely = evidence leans this way, but not definitive
- Strong evidence = consistent results across high-quality studies
- Unclear / mixed = meaningful disagreement or inconsistent findings
What triggers an update
We revisit pages when:
- Major guidelines change
- Strong new evidence emerges (especially replicated)
- Safety signals become clearer
- Readers flag issues that reveal ambiguity or error
A practical checklist for readers
When you see a health claim, ask:
- What is the outcome — a lab marker or a real patient outcome?
- How big is the absolute risk change?
- Who was studied — does it apply to me?
- Was it replicated?
- Does guidance agree across credible bodies?