All topics
Reading scientific evidence
Evidence is not a simple ranking. Ask about design, population, comparator, primary outcome, between-group difference and limitations. A human randomized trial can still provide limited evidence because of sample size, bias or outcome selection.
In-depth articles
All articles →A map of related concepts; see the entries and sources for detail.
Reference entries
How to read the evidence labels
Foundation means established nutrition knowledge; human studies need population and endpoint checks; observational findings show associations; mechanisms are not dosing advice; inference marks our judgment.
Mechanistic, observational and randomized evidence
Mechanistic studies explain possibilities; observational studies find associations; randomization reduces confounding. Quality and relevance still matter.
The comparator shapes interpretationProject inference
No-treatment, placebo and nutrient-matched comparators answer different questions. Subjective outcomes can be influenced by knowing one receives the product.
Start with the primary endpoint and group difference
“Better than last month” is not “better than control.” Selecting one significant result from many measures increases error risk.
Effect size, confidence interval and P valueProject inference
A P value does not tell you whether benefit is worthwhile or universal. Examine magnitude, uncertainty and real-life relevance.
Can ingredient evidence transfer?
Match identity, chemical form or strain, exposure, duration, population and endpoint. A mismatch can weaken transfer.
Funding and replicationProject inference
Industry funding does not automatically invalidate a study. Examine design, transparency and independent replication. A brand page is primary evidence of its claims, not independent efficacy.
Preprints, papers and corrections
Check first online date and article type; the year inside a DOI need not be the publication year. A correction is not a second independent trial.
Read the primary outcome and comparator
Check the prespecified primary outcome and the between-group difference before exploratory subgroups or secondary outcomes. A within-group change does not replace a comparison with the control.
A label is product information, not efficacy evidence
Labels and specifications help establish identity and amount. Whether human evidence applies also depends on preparation, population, dose, use and outcomes.