Trial Eligibility Matcher
Free-text criteria → structured filters → eligible-patient list with risk scores.
Demo guide — how to read this screen
- 1. Pick an anchor-trial chip below — each demonstrates a different PANDA value proposition, and a "PANDA Fit" brief appears on the left with who matches in the live cohort.
- 2. Click Find eligible patients to run that trial's real inclusion/exclusion text against the live 14,999-patient cohort.
- 3. Scan Parsed criteria: green = applied as a database filter; amber = recognized but not in PANDA's schema yet (RAS mutation, ECOG, organ function).
- 4. Edit any criterion and re-run. Change "metastatic" to "resectable" and the count moves — the filter actually changes. Editing an amber line does nothing, by design.
- 5. Read the two headline numbers: the green count is the structured pre-screen (an upper bound); the Estimated real-world eligible card discounts it for the criteria we can't filter yet — the honest funnel toward enrollable patients.
PANDA Fit — High Impact
Revolution MedicinesPrimary endpoint: PFS and OS in RAS G12-mutant population; PFS by BICR per RECIST 1.1
Primary endpoint is BICR per RECIST 1.1 — Patient Review + EUS AI engine. Automated lesion segmentation, κ-tracked inter-reader agreement (0.81), disease-burden trajectory by arm, audit-ready discordance flags. Cohort Builder pre-screens by RAS mutation + measurable disease + ECOG 0–1.
In the live cohort this matches metastatic PDAC with measurable disease — PANDA risk groups I/J. RAS mutation status, ECOG, and organ function stay amber: in production those come straight from the molecular report and the EMR feed, so a coordinator sees exactly what still needs checking before screening.
Largest mPDAC survival win ever; co-develop PANDA as the central read engine for follow-on combo and earlier-line studies
Trial criteria
Parser is deterministic keyword/regex extraction for the demo. Production routes through an LLM with the same structured-filter contract and cites back to each extracted field.