Cancer Trial Finder

A patient-profile → clinical trial matching tool. You fill in a six-step profile (demographics, diagnosis, biomarkers, treatment history, labs, other health info), and it searches the public ClinicalTrials.gov registry for recruiting studies, scores each one for relevance against your profile, and lets you export a PDF discussion sheet to bring to your oncologist.

No backend, no build step, no API key required for the core app — it's static HTML/CSS/JS that calls ClinicalTrials.gov's free public API directly from the browser.

A deliberate scope decision

This project does not curate or feature a fixed list of "repurposed drugs" (things like fenbendazole or ivermectin) as a special matching category. A couple of those substances have no real oncology trial base and are mostly known from online cancer-cure misinformation — wrapping them in a scoring UI would hand them false algorithmic legitimacy to people who are often scared and searching for options. What's here instead searches all of ClinicalTrials.gov based on the patient's actual diagnosis and biomarkers, which will surface real drug-repurposing trials on their own merits (metformin, aspirin, propranolol, disulfiram, itraconazole, and statins all have genuine active oncology trials) without a thumb on the scale for any particular drug.

In the same spirit, the matching logic only hard-excludes a trial on structured fields the API actually provides (age range, sex). Everything else — biomarkers, prior therapy, proximity — is a transparent, additive "relevance score" with plain-language reasons attached, not a manufactured percentage. The UI says so directly, and a mandatory disclaimer/consent checkbox has to be checked before a search can run. See the comments at the top of js/scoring.js if you extend the logic — please keep that framing.

Project structure

index.html              wizard + results markup
css/styles.css           all styling
js/data.js                dropdown/option lists
js/wizard.js               step navigation
js/ctgov.js                  ClinicalTrials.gov API client + ZIP geocoding
js/scoring.js                 profile-vs-trial matching logic
js/pdf.js                      PDF discussion-guide export (jsPDF)
js/app.js                       glue: form → search → render
server/explain-proxy.example.js  optional, NOT wired in — see below

Running it locally

Any static file server works. From this folder:

python3 -m http.server 8000
# then open http://localhost:8000

Opening index.html directly via file:// will usually work too, but some browsers block cross-origin fetch() calls from file:// pages — if trial search doesn't load, use the command above (or just deploy to GitHub Pages, which serves over https:// and avoids the issue entirely).

Deploying to GitHub Pages

git init
git add .
git commit -m "Initial commit"
git branch -M main
git remote add origin <your-repo-url>
git push -u origin main

Then in the repo: Settings → Pages → Deploy from a branch → main → / (root). No build step, no environment variables needed for the core app.

Notes on the two external services this uses

• ClinicalTrials.gov API v2 (https://clinicaltrials.gov/api/v2/studies) — free, public, no key. Docs: https://clinicaltrials.gov/data-api/api
• OpenStreetMap Nominatim — used once per search to turn a ZIP/postal code into coordinates for distance sorting, so the app can do proximity scoring without a paid geocoding key. It's free but has a light usage policy (attribution, ~1 request/second) — see https://operations.osmfoundation.org/policies/nominatim/. If a person doesn't enter a ZIP, or geocoding fails, the app just falls back to searching by state/country text instead of failing.

Both are called directly from the browser. If either ever changes its CORS policy and browser requests start failing, the fix is a one-file serverless proxy (same pattern as the example below) rather than any change to the matching logic.

Optional: adding an AI-generated plain-language explainer

This isn't built in, on purpose — wiring up a "free AI" would mean either shipping an API key in client-side code (which anyone could copy out of your public repo and use on your bill) or standing up a backend, which is outside the scope of a static site. server/explain-proxy.example.js is a template for the second option if you want it later: a small serverless function (Vercel/Netlify-style) that keeps your Anthropic API key server-side and proxies one request at a time. Instructions are in that file's comments. There's no free tier that doesn't require you to hold an API key somewhere safe — Anthropic's current pricing and setup docs are at https://docs.claude.com.

Disclaimer

This tool is an informational research aid built on public registry data. It is not medical advice, does not determine trial eligibility, and is not affiliated with ClinicalTrials.gov or the U.S. National Library of Medicine. Anyone using it should confirm every detail with the trial site and their own treating physician before making any treatment decisions.