An early alert has no recall number yet

A device recall record, read off openFDA, carries a recall number the moment it exists — the field this blog has already traced back to FDA’s Recall Enterprise System. Since November 2024, some of the highest-risk removals and corrections FDA knows about have had a public page before they had that number at all.

A pilot built to close a specific gap

On November 21, 2024, CDRH announced a pilot aimed at one problem: the time between FDA’s first awareness of a likely high-risk device removal or correction and the agency’s first public word about it. An Early Alert posts when FDA believes a company’s action is likely to become the most serious kind of recall, but before the agency has determined that the action meets the regulatory definition of one — the classification step every recall record otherwise waits on. The pilot launched scoped to five device areas: cardiovascular, gastrorenal, general hospital, obstetrics and gynecology, and urology. On September 29, 2025, CDRH expanded it to every device type it regulates.

What the page carries, and what it can’t yet

An Early Alert names the affected product, states the basis for the alert, describes what a firm or user should do, and reports any deaths or injuries associated with it. It also carries a status — FDA still collecting information, the action confirmed, or new information added since the alert first posted. What it does not carry, by construction, is a recall number or a classification. Both are outputs of the hazard-evaluation step FDA has not finished yet. The alert lives on its own page, in FDA’s Medical Device Recalls and Early Alerts database — not in an Enforcement Report, because an Enforcement Report entry does not exist for this action until classification closes that gap.

The alert is FDA saying, in public, that it is worried before it has finished the paperwork that would normally be the first public trace of that worry.

The feed a pipeline reads starts later

openFDA’s device enforcement data — sourced from the Recall Enterprise System, and the dataset FDA Radar’s own recall corpus is built on, per what we monitor and how often — picks up a recall once FDA has classified it, not before. An Early Alert precedes that record entirely; nothing about the alert obligates FDA to backfill it into the enforcement feed on any particular clock, only that the eventual classified entry will exist once the hazard evaluation is done. openFDA’s own documentation for the device recall data makes a related point directly: the data is not meant to be used as a method for issuing alerts to the public in the first place. FDA built a separate channel for that job on purpose. The API was never supposed to be the fast path.

The practice

For a product code, firm, or device class worth watching closely, the Early Alerts page is the earlier and, for a subset of the highest-risk actions, the only public signal for a period FDA does not commit to a length — it lasts exactly as long as the hazard evaluation takes. A monitoring pipeline keyed to the classified enforcement feed, FDA Radar’s included, will not show that alert until the classification lands and the record openFDA serves catches up. Whether a specific Early Alert changes how urgently a specific supplier or product code needs attention is a judgment for the regulatory or quality professional holding that file — this is regulatory intelligence and method, describing where a signal appears first, never regulatory advice about what any one alert means for a specific portfolio.

FDA Radar’s recall coverage runs on the enforcement endpoint, exactly as what we monitor and how often says: openFDA API, irregular cadence, roughly a week behind FDA. An Early Alert sits upstream of all of that — a gap worth naming here rather than leaving a reader to notice it only when a familiar feed stays quiet.