We don't predict approvals. We show you the facts.
Updated ·next FDA decision on the calendar
pdufa.bio is built on one principle: give biotech catalyst traders the verifiable facts and the primary sources, not a black-box guess at whether a drug will be approved.
What we will never do
Per-drug approval probabilitiesNever
Buy / sell / position-size callsNever
Composite "bullish/bearish" scoresNever
Fabricated or unsourced dataNever
Some tools will tell you a drug has a "73% chance of approval" from a machine-learning model. We think that's exactly the kind of false precision that gets retail traders burned. Approval is binary and idiosyncratic; a single number hides far more than it reveals.
Outcome archiveSource-verified, with primary links
When we say a cohort of similar-cap PDUFAs moved a median ±X% on decision day, that's history, labeled as history, not a probability for the specific drug in front of you. When we mark a past decision "verified," it links to the FDA / SEC / company filing so you can check it yourself. When a label is still price-only, we say so.
How a date gets its precision
Every event row carries a date_precision of day, month, quarter or year, and the row means exactly that much. A day-precision row is a calendar day a sponsor or the FDA stated in a filing, a release or an FDA notice, and we link the document. A coarser row is the window the sponsor gave (“third quarter of 2026”) or the ClinicalTrials.gov primary-completion estimate, and we say which. Three rules decide what happens when two statements about the same event disagree:
1. The most recent stated day winsA sponsor that said September 26 in May and November 22 in August moved the date. The later day is the date; the earlier one stays on the row as date_history, with the filing that moved it.
2. A quarter is never rounded to a day“Q3 2026” is published as Q3 2026, in the calendar and in the API, with date: null. We do not turn a quarter into September 30 or a month into the 15th, and a row we cannot source to a stated day is downgraded to the window we can source, not annotated.
3. A later window does not override an earlier dayA 10-Q that says “third quarter” after an 8-K that said September 26 is the same date at lower resolution, not a change. The day stands. If the later window does not contain the day, the two statements conflict and a person settles it against the filings before anything is published.
The rule runs inside the harvester, where two statements first meet, and again as a build guard on the dataset the API ships. The API exposes date_precision, source_url and date_history on every row so you can hold us to it.