Methodology

What these pages are

Each page collects passages from public documents: FDA Complete Response Letters, published in openFDA's Complete Response Letter dataset, and companies' filings with the SEC. The passages come from the EphemerisX knowledge graph, which records every sentence with the document, page and sentence id it came from.

Rules

Every factual statement on a page is a quoted passage, word for word, with a link to its source and its sentence id. The pages' own words are headings and descriptions of what a page contains; they add no facts.

Links to the source

Every passage links to the document it was read from, not to a dataset. A Complete Response Letter opens as the letter's own PDF, published by FDA through openFDA, at the page the passage is on; that page is found by reading the PDF itself, with text recognition for scanned pages. A passage from an SEC filing opens the filing on sec.gov scrolled to the passage, highlighted in browsers that support it. A passage whose document cannot be found is not shown.

A passage is left out when it is an incomplete sentence, when it is the closing text that every Complete Response Letter repeats, when it describes the approval process in general rather than the company's case, or when it only matched a topic by wording. Every left-out passage is recorded with its reason in the page's build record.

Gaps and breaks inside quoted text are as they appear in the source. Complete Response Letters are published with redactions.

Drug pages

A drug page quotes FDA's own documents for one application: its reviews, approval letters and labels, as published on Drugs@FDA. Each passage opens the FDA PDF at the page it is on. The documents were read again from FDA's copies, each checked against the fingerprint recorded when it was first read.

The decision section quotes only reviews and the original approval letter. Meeting minutes, form letters' standard wording and blank checklists are left out. Each passage is also found on its page in the PDF: a passage that does not follow the end of a sentence there is a fragment and is left out.

The key facts are held to the same rule. The status is shown only as far as a quoted document supports it, with links to those passages, and the sponsor is the company FDA addressed its most recent letter to, linked to that letter. A sponsor with no such letter is shown as recorded in the knowledge graph, without a link.

Updates

Pages are regenerated as companies are added to the knowledge graph. Each page shows when it was built and from how many companies.

Not advice

These pages report what documents say. They are not investment, legal or medical advice.