Clinical Trials SoCal

Physician Search Guide

Retrieve cohorts, then review the evidence

The search is a conservative protocol-review aid. It proposes source-grounded patient facts, waits for physician confirmation, and retrieves plausible cohorts. It does not determine eligibility or recommend treatment.

Back to Search Trials

De-identify before typing

Do not enter a name, MRN, date of birth, address, phone number, email, or another direct identifier. The raw narrative is not deliberately stored in local or session storage, but de-identification remains the physician's responsibility.

Confirmation-first workflow

  1. Enter a concise, de-identified case.
  2. Inspect each extracted fact and exact source phrase.
  3. Review the facts you can verify. Provisional trials appear immediately; missing or unreviewed facts remain unknown.
  4. Resolve contradictions and inspect unmodeled text.
  5. Run retrieval and review the current protocol/site.

Recommended structure

[age/sex] + [cancer/setting] + [histology/risk] + [therapies/dates] + [progression] + [biomarkers] + [ECOG/labs]

Exact values, units, assays, and dates are more useful than broad summaries.

Example clinical description

72-year-old man with mCRPC. Progressed on enzalutamide; last dose 2026-06-01. No prior docetaxel. BRCA2 pathogenic alteration. PSMA-positive PET. ECOG 1. Labs on 2026-08-01: hemoglobin 9.5 g/dL, ANC 1.5 x 10^9/L, platelets 125 K/uL, CrCl 55 mL/min.
Received ≠ progressed on
State the event that is actually documented.
Planned ≠ prior
Considering pembrolizumab is not exposure.
Generic labs do not pass thresholds
Include numeric results, units, and collection date.

Context that cannot become a positive fact

  • Family history of prostate cancer
  • Cannot rule out brain metastases
  • Possible recurrence
  • Considering nivolumab

Possible, conditional, hypothetical, and non-patient assertions trigger abstention or a clarification request.

Criterion states

  • SATISFIED: confirmed evidence satisfies the predicate.
  • VIOLATED: confirmed evidence contradicts it.
  • UNKNOWN: patient evidence is missing, uncertain, stale, or contradictory.
  • NOT_MODELED: protocol text cannot be represented safely.
  • NOT_APPLICABLE: the branch does not apply.

How to read the result queues

Priority for protocol review
Positive disease evidence, reviewed critical data, no modeled critical conflict, and current local status evidence. Not an eligibility verdict.
Potentially relevant
Plausible disease context, but important patient evidence remains unknown or uncertain after review.
Manual review — trial data incomplete
Cohort boundaries or critical trial criteria are incomplete, ambiguous, or unreviewed.
Modeled conflict
A represented criterion conflicts with confirmed facts; the record remains visible for audit.
Disease-context retrieval only
Same cancer context without enough positive evidence for a more specific queue.

Each queue is ranked and shown in batches of 20. Use the “Show next” control to inspect every retained cohort; batching does not remove candidates from the audit set.

Bladder / urothelial

BCG-unresponsive NMIBC with CIS after adequate BCG; recurrence documented 2026-05-15.

Kidney / RCC

Metastatic clear-cell RCC, IMDC intermediate risk. Progressed on nivolumab plus cabozantinib.

Testicular / GCT

NSGCT after first-line BEP with residual mass 2.4 cm, markers normal, no prior HDCT.

Protocol and site verification are mandatory

One NCT record can contain different arms and cohorts. Until those boundaries are curated, the result is explicitly sent to manual review. Study-level recruiting status is not proof that a particular local site or cohort is open. Review the current protocol and confirm directly with the site before referral.