Start with who was studied, what was compared, and what was measured. Then ask how large the effect was—and where the evidence stops.
01 / Relative vs. absolute
Same relative effect. Different real-world scale.
Hypothetical event rates, not a treatment forecast. Change the starting rate to see what happens.
2.5 fewer events per 100: a 2.5 percentage-point decrease, not a 25 percentage-point decrease.
Model: comparison rate = starting rate × risk ratio, for the same outcome and follow-up. Fractional rates describe averages, not partial people. No uncertainty or harms are modeled; a lower rate is not always a better outcome.
02 / Put a finding in context
Read the question. Not just the conclusion.
Examples illustrate different questions, not a ranking of treatments. Publication dates and source-check dates are different.
A precise estimate is not the whole story
A confidence interval describes statistical uncertainty under the analysis assumptions. It does not remove bias, measure every harm, or predict an individual's response.
A biomarker is not a clinical outcome
Check the actual endpoint and follow-up. A short-term laboratory change does not, by itself, show that people live longer or avoid disease.
A new paper is a starting point
Read the population, comparator, effect size, limitations, funding, and relevant existing evidence. “New” is a publication date—not a quality grade.
0 curated records - verified by recency, rigor, and conflict checks.
Use this as a starting map, not an endorsement list. Each topic card includes an evidence
grade and a risk layer so you can prioritize what to study or test next.
Research Coverage Map
Topic presence is not the same as evidence maturity.
Explore topic coverage and research gaps
Research Desk
Filter the ledger by design, year, and source language.
This view is for discovery and editorial triage. A filter result is not a ranking of clinical quality;
read the study design, interpretation, and safety boundary in the linked dossier.
Publication dates describe the source; check dates describe recorded work on this library. The 90-day recheck interval is an editorial reminder, not an expiry date for science. These statuses do not run a live literature search or establish medical review.
Editorial update feed
What changed in the encyclopedia
Evidence Landscape
A visual map of maturity, risk, recency, and coverage.