Adverse Event¶
Use adverse-event commands for safety surveillance across four source-backed paths:
- OpenFDA FAERS drug adverse-event reports,
- CDC WONDER VAERS aggregate vaccine adverse-event summaries,
- recall notices,
- device events.
Vaccine searches default to combined OpenFDA FAERS + CDC VAERS when the query resolves to a vaccine and the active filters are VAERS-compatible. Non-vaccine searches stay FAERS-only in practice. --source vaers is aggregate-only, accepts only the vaccine query and --limit, and rejects --reaction, --outcome, --serious, --date-from, --date-to, --suspect-only, --sex, --age-min, --age-max, --reporter, --count, --offset, recall classification, and device fields.
Search FAERS reports¶
By drug:
Serious reports only:
Reaction-focused filter:
The reaction summary labels each percentage as a share of returned reports: the numerator is the number of returned-page FAERS reports containing that reaction and the denominator is the number of reports on that returned page. The page is only a sample of all matching reports. These spontaneous-report shares are not incidence in treated or exposed patients and do not establish causality.
biomcp drug adverse-events <name> instead uses an aggregate OpenFDA count and
labels the percentage as a share of matching reports. Its denominator is
all matching FAERS reports. JSON identifies these meanings in
summary.percentage_context, including the exact divisor and explicit false
values for incidence and causality.
FAERS counts are FAERS-only and therefore require --source faers; count operations require offset zero:
For --count, Percent of Shown Count divides each bucket by the sum of the
shown bucket counts. That sum is not a count of distinct reports because one
report can contribute to multiple buckets. Count-mode JSON remains the raw
bucket contract and does not contain calculated percentages or
percentage_context.
Search vaccine events with VAERS¶
Combined FAERS + VAERS for vaccine queries:
VAERS-only aggregate summary:
VAERS summaries are aggregate-only. They surface the matched vaccine display name, CDC WONDER code, CVX code(s) when the query resolves through the CDC CVX/MVX bridge, serious vs non-serious counts, age distribution, and top reaction counts. --limit bounds the number of top reaction rows.
--source only applies to --type faers; recall and device searches keep
their existing source-specific paths.
Search recall notices¶
Classification filter:
Search device events (MAUDE)¶
Manufacturer filter:
Product-code filter:
--manufacturer and --product-code are valid only with --type device.
Device seriousness is typed: bare --serious and --serious any select Death or Injury and render as death_or_injury, --serious death selects Death only, and --serious injury selects Injury only. FAERS seriousness values such as hospitalization are rejected on the device route.
Get a report by ID¶
Report resolution is source-aware and returns the corresponding markdown format.
Request report sections¶
| Section | Description |
|---|---|
reactions |
Adverse reactions reported |
outcomes |
Reaction outcomes (death, hospitalization, etc.) |
concomitant |
Concomitant medications |
guidance |
Safety guidance and labeling |
all |
Include all sections |
Helper commands¶
There is no direct adverse-event <helper> family. Use
biomcp drug adverse-events <name> when you want the inbound drug pivot into
this safety surface.
JSON mode¶
Combined-source outcomes¶
JSON from search adverse-event <query> --source all includes independent section_outcomes.faers and .vaers entries while preserving vaers.status. One branch can be unavailable without erasing a healthy result from the other; empty means that source completed successfully with no matching evidence. A successful VAERS branch credits both CDC CVX identity resolution and CDC VAERS evidence. When --source all includes FAERS-only filters or a nonzero offset, vaers.status explains the skip and its section outcome remains not_requested because no VAERS fetch ran. Device-search JSON includes the effective query summary, including the selected seriousness meaning.
Practical tips¶
- Include drug generic names for better FAERS recall.
- Use plain vaccine names, family names, or common brand names when you want the VAERS path.
- Treat FAERS rows and VAERS aggregate counts as signal, not incidence estimates.
- Validate serious findings through full source documents when needed.