About Evidence Hound
An open, citable reference for the medicine of aging dogs, assembled from randomized controlled trial data and published in the open so it can be checked.
Why this exists
Owners of aging dogs are making real medical decisions under time pressure, usually by searching the internet between vet visits. What they find is dominated by supplement marketing, forum anecdote, and articles that treat a manufacturer's press release and a placebo-controlled trial as equivalent kinds of information. Meanwhile the actual trial literature is scattered across paywalled journals in language written for clinicians.
This site puts the trial evidence in one place, states plainly how strong it is, and shows its work.
How it is built
Every claim lives in a JSON file under data/interventions/ with its citations attached. A validator
enforces the schema, and the site is generated from those files by a dependency-free Node script. There is no CMS, no
database, and no editorial back room. The content and the code are the same repository, and the published
dataset is a single file anyone can download and reuse under CC BY-SA 4.0.
That structure is deliberate. It means a veterinarian who disagrees with a grade can open a pull request changing one field and stating why, and the disagreement becomes part of the public record instead of an argument in a comment section.
What this site will not do
- Tell you what to give your dog. It has never met your dog.
- Take affiliate revenue or product sponsorship. The moment a site earns money per supplement click, its grades stop being trustworthy.
- Hide negative findings. Grade D exists precisely so failures stay visible.
- Treat regulatory approval as proof of effectiveness, or absence of approval as proof of uselessness.
Colophon, for the curious
This site has zero dependencies. npm install downloads nothing, there is no lock file
to audit, and no transitive package can change what a grade says. For a project whose entire value is
being checkable, that felt like the right trade against convenience.
The stack, such as it is:
- Content
- One JSON file per intervention, validated against a JSON Schema by a hand-written ~90-line checker. The data is the CMS.
- Generator
- A single Node script using template literals. No React, no bundler, no framework.
- Front end
- One stylesheet and one vanilla JS file. The tracker is ~370 lines and stores to localStorage.
- Hosting
- Firebase Hosting on Google Cloud, deployed by GitHub Actions in about 35 seconds.
- Licence
- AGPL-3.0 for code, CC BY-SA 4.0 for the evidence.
The part worth stealing: some tests assert editorial policy rather than code correctness. An entry graded U for "untested" that cites a completed randomized trial fails CI, because it should have been regraded. An entry graded A or B that cites no trial fails too. The build refuses to publish data that contradicts the grading rules, which means the rules cannot quietly rot as the content grows.
That idea generalises well beyond dogs. Any project making graded claims from evidence can encode its own standards as tests and let CI enforce them.
The data is an API
Everything on this site is one JSON file, served with permissive CORS so you can build against it:
curl -s https://evidencehound.org/data.json | jq '.interventions[] | {name, tier}'It carries every intervention, grade, citation, funding disclosure, and the full monitoring framework. CC BY-SA 4.0, so build what you like as long as it stays open. If you make something, tell us.
Contributing
Corrections, new interventions, and challenges to existing grades are all welcome, particularly from veterinarians and researchers. See CONTRIBUTING.md. If you are citing a trial we have misread, that is the most valuable contribution you can make.