A reference that stays out of the way

Why it exists

The tools that hold this information are usually a browser tab or a separate application. Both make you leave what you were doing, and leaving what you were doing mid-note is how mistakes get made. So the design constraint came first: it must never take the screen away from you.

That is why it is borderless, always on top, semi-transparent, and collapses to a 54-pixel pill placed exactly where the collapse button was — so closing and reopening are the same click in the same spot, with no mouse travel at all.

Who it is for

Students, trainees and researchers, working under supervision, and the licensed clinicians who supervise them. The curated tier is written around chairside practice: antibiotics and prophylaxis, analgesics, local anaesthetics with and without vasoconstrictor, sedation agents, antifungals and antivirals, and the anticoagulants and bisphosphonates that change what you can safely do.

It is a reference aid. It does not diagnose, prescribe, or practise medicine, and it is not a substitute for the label or for your own judgement.

How the data is put together

Four sources, layered, and each one labelled in the app wherever its data appears. You should never have to guess whether what you are reading was reviewed by a person or lifted verbatim from a label.

• The graded tier — severity gradings for the drugs prescribed chairside, ranked first. Every entry and every interaction is grounded in the FDA label for the same drug and names the section that supports it, with a link to the current label; the prose shown is the label’s own words. An interaction the label does not support is not shown at all. What is editorial is the grading — major, moderate or minor — which labels do not provide.
• openFDA labels — the FDA's own corpus, public domain. Deduplicated on the FDA's UNII ingredient codes rather than on drug names, which is the difference between filing acetaminophen once and filing it 301 times. RxNorm then folds salt and hydrate forms of the same ingredient together, and where two of those differ clinically the entry names both rather than letting one product's dosing stand for the other.
• The FDA's CYP and transporter tables — public domain, and the basis for predicted pharmacokinetic interactions. These live in a section of their own, each labelled a prediction and each naming its pathway. Only a strong modifier acting on a sensitive substrate is graded major; over-grading a prediction teaches people to ignore the app.

A useful sign that the inference is sound: the mechanism graph, built only from the FDA's tables, independently rediscovers interactions the curated tier lists by hand — clarithromycin with triazolam, clarithromycin with simvastatin — and grades both major through CYP3A. That agreement is a test in the suite, not an anecdote.

What it deliberately does not do

Adverse-event report co-occurrence would be easy to mine and would produce something that looks like an interaction list. It would actually be a measure of how often two drugs are prescribed to the same kind of patient. In a tool someone prescribes from, that is worse than having nothing.

Nothing here invents a severity. Where a source grades an interaction, the grade is shown and attributed; where it does not, the pair is shown ungraded rather than given a colour it has not earned.

Label sections are stored whole. An earlier version truncated them on ingest and cut 58% of boxed warnings mid-sentence — a half-read boxed warning is more dangerous than an absent one.

The two editions

Academic Edition is free and available now — for students, trainees, researchers and academic staff, used under supervision. It is the whole application, with nothing held back.

Professional Edition is for clinical practice. It is in development: there is no price and no date yet, and what it becomes depends partly on what people tell us they need. See pricing, or tell us what you would need.

There is no third, cut-down free edition. If you want to try Rxference without installing anything, the demo on the front page is the real interface running on the real data, in your browser.

Both editions carry the same data. They did not always. The Academic Edition once bundled a licensed third-party interaction database that a commercial product could not legally include, and that difference is gone: every source in the application is now a work of the US federal government, in the public domain, so there is nothing the paid edition has to leave out. What separates the editions is who may use them and what we commit to, not what they know.

The application tells you which edition you are running: the footer reads Academic Edition or Professional Edition, taken from your license.

Register an academic license if you are using the Academic Edition.

Attribution

• openFDA drug label API and bulk downloads — U.S. Food and Drug Administration. Works of the U.S. federal government, public domain.
• RxNorm — U.S. National Library of Medicine.
• Drug Development and Drug Interactions: Table of Substrates, Inhibitors and Inducers — U.S. Food and Drug Administration. Public domain.

Rxference is not affiliated with, endorsed by, or sponsored by any of the above, nor by any electronic health record vendor.

Privacy

This section describes how the application behaves. The privacy policy is the formal notice covering the website and the licence service as well, including your rights and how to exercise them.

Nothing you search for ever leaves the machine. The drug database is a local file, the application has no account, and no drug name, search term or card you open is transmitted, logged or stored anywhere but on your own computer. That is the promise, and it has no exceptions.

There is no telemetry and no analytics. We considered adding an anonymous usage counter and decided against it. It would have bought us a rough count of who is running which version; it would have cost the one unqualified privacy claim this kind of software almost never gets to make. That was a bad trade.

The application does make one network request. Once a day at most, it fetches a small signed file from this website to see whether a newer version exists. It is an ordinary request for a public file: it carries no identifier, no version number, no query string and no body, so the only thing disclosed is that some computer asked for a file anyone can ask for. Nothing is downloaded or installed without you being told what it is and agreeing to it, and every update is checked against a cryptographic signature and a checksum before it runs.

The other network operation is the data pipeline fetching a fresh openFDA export, which only happens when you ask it to.

This website is static. Anything you type into the academic license form is used to record and support your license, and for nothing else.

The website counts visits, and it does so without cookies. That is why there is no consent banner here: the rule that produces those banners is about storing or reading something on your device, and nothing is stored on yours. No cookie is set, nothing is written to local storage, and no identifier is sent.

What is counted, per page and per day, is the number of views, how long the page was visible, and how many people started a download. Those are totals. There is no record of a visit, only of how many there were — which means we cannot tell a returning visitor from a new one, or follow anybody from one page to the next. Doing either would need an identifier, an identifier would need your consent, and we would rather have the simpler answer and no dialog.

Our web host keeps standard server logs for 90 days, as every web server does. Those include IP addresses, and they are used for aggregate counts and to investigate abuse. If your browser sends Do Not Track or Global Privacy Control, the counting script does not run at all.

Who is behind it

Rxference™ is published by Rxference Foundation, a Delaware corporation. It is a small project rather than a large organisation, and the Academic Edition is distributed free; if it saves you time and you would like to help keep it going, there is a donation page.

The full terms are on the disclaimer page, and they are the same text the app shows under its own Disclaimer link.