How We Research Drug–Supplement Interactions

Every reviewed pair on this site traces back to real, recorded sources. Here is the process behind that, in more detail than the methodology page covers at a glance.

Source hierarchy

Sources are prioritized in roughly this order:

1. FDA-approved labeling and FDA safety communications
2. DailyMed / National Library of Medicine drug labels
3. NIH Office of Dietary Supplements fact sheets
4. PubMed-indexed primary research and systematic reviews
5. Clinical practice guidelines from recognized professional bodies
6. MedlinePlus, for patient-facing reference context

We deliberately do not use commercial/proprietary interaction databases as a source -- everything traces to a source you could look up yourself.

How a pair moves from candidate to published

A pair is chosen for review, then researched by directly reading the primary sources for both substances -- not by trusting a summary or a snippet. Findings are authored with the exact source and the specific claim it supports. Content then moves through draft, verification, and publication stages -- see the editorial policy for that workflow in full. Nothing is published without at least one linked source; this is enforced at the database level, not just by editorial habit.

Catching false positives

A keyword match is not the same as a real interaction claim. During research, mentions that turned out to be a drug's own salt-form product name (like "Atorvastatin Calcium"), an inactive tablet-filler ingredient listing, or a bibliography citation unrelated to the actual prose of a label have all been caught and excluded before publishing -- each finding is verified in context, not just detected by text search.

How "no documented interaction" findings are handled

When sources genuinely do not document an interaction for a pair, that is published explicitly as its own result -- see what that result means -- rather than left as an unreviewed gap. This is what makes full coverage of a catalog meaningful: every pair has a real answer, not just the ones with an interesting finding.

Corrections and updates

Evidence changes over time, and mistakes are possible. Anyone can report a suspected error via corrections. Reports are reviewed before any change is made -- never auto-published.