Chart Review in Oncology: Surfacing Biomarkers and Inferring Eligibility from the Patient Record.
Four case studies on how Areti's PRIME engine reads pathology, surfaces molecular biomarkers, and infers ECOG and TNM from unstructured notes — with clinician verification on every inference.
- ✓
The PRIME engine architecture — how it ingests notes, labs, and pathology reports, then maps to inclusion/exclusion criteria - ✓
Biomarker-driven recruitment: APOL1 CKD (Elixia) — 8× site baseline, volume-driven outreach pause - ✓
Note-level interpretation: Early Alzheimer's (Gadolin) — 55% screen-fail vs. >95% baseline - ✓
Active oncology program: DLBCL (Genentech STARGLO) — real-time immunophenotype and progression tracking - ✓
Inference framework: how PRIME reads "mostly in bed" and returns ECOG 3 — with TNM, RECIST, and pathology examples
Free · 6 pages · PDF
Get the bundle.
Tell us where to send the PDF. You'll get instant access — no waiting on email.
By submitting, you consent to Areti Health storing your information and contacting you about clinical trial recruitment. We never sell your data. Privacy policy.
>90%
PRIME ↔ nurse alignment
3,500+
EMR leads chart-reviewed across cases
DLBCL
Active oncology program (STARGLO)
4
Showcases inside
Inside the brief
Four case studies. One engine: PRIME.
Three completed enrollments and one active oncology program — selected to demonstrate the precise capability oncology recruitment depends on: reading the chart, surfacing medical events, and inferring eligibility that keyword search misses.
01 Biomarker-driven · Personalized medicine
APOL1-Mediated CKD (Elixia)
A genotype-defined enrollment problem solved by record-level biomarker targeting. Structurally identical to a BRAF V600E or KRAS G12C stratified oncology trial.
163
responses · 8× baseline
130
visits in 10 days
1,000
EMR leads matched
02 Chart review · Automated outreach
Early Alzheimer's Disease (Gadolin Research)
A high-screen-fail study rescued by interpreting unstructured notes and lab narratives — the same problem oncology faces with pathology typing, ECOG, RECIST, and staging.
44
visits in 48 hrs
2 wks
vs. 4–6 mo forecast
55%
screen-fail vs. >95%
● ACTIVE 03 Oncology showcase
Challenging Lymphoma — DLBCL (Genentech STARGLO)
Real-time immunophenotype and biomarker monitoring to capture patients at the right line of therapy. Continuous re-reading of the record so the AI Coordinator reaches the patient inside the trial-eligible window.
CD19low
subtype-aware
Real-time
progression tracking
Community
cancer center routing
04 Natural-language inference framework
Inferring Disease Severity & Stage from Notes
How PRIME reads "mostly in bed" and returns ECOG 3 — with the same engine surfacing TNM stage, pathology subtype, and biomarker status, and a clinician-verifiable rationale on every call.
ECOG
inferred from narrative
TNM · RECIST
staging coverage
>80%
alignment with clinicians
The PRIME engine
Precise Record Retrieval, Ingestion & Matching Engine.
PRIME ingests the full longitudinal record — medical history, diagnosis, medications, procedures, labs — plus typed, dictated, and scanned narrative notes. An advanced AI layer with a clinical intelligence library surfaces biomarkers, infers disease severity and stage, and matches patients against protocol inclusion/exclusion criteria. Every inference carries a rationale verified by clinical staff, with >90% alignment between PRIME and reviewing nurses.
Want to run PRIME against two or three live oncology protocols?
A 30-minute working session is the fastest way to see biomarker surfacing, performance-status inference, and real-time progression tracking combine on a real indication.
Book a 30-min working session Browse other case studies