Not every correction reaches FDA

Run a firm through a recall feed and come up empty, and the read is tempting: this company hasn’t had a corrective action in years. FDA’s own rule for when a correction or removal has to reach FDA at all says something narrower than that. A correction that never appears in a public recall database is not proof nothing happened. It may be proof of exactly one thing — that whatever happened didn’t cross the line 21 CFR 806.10 draws between a report and a record.

The gate before a recall exists

A device recall, in FDA’s enforcement data, starts life as a correction or removal a firm reports under 21 CFR 806.10. The rule requires that report within ten working days, but only when the action was initiated to reduce a risk to healththe device poses, or to remedy a violation of the Act that may itself present a risk to health. Nothing else in Part 806 requires a report. A firm that pulls a device to relabel it, tighten a spec, or fix a defect that never rose to a health risk is not required to tell FDA at all — 21 CFR 806.1(b) exempts exactly that category, by name: actions taken to improve performance or quality that do not reduce a risk to health or remedy a violation presenting one.

Where the report goes instead

806.10 also steps aside when the same information already reached FDA another way. A correction already reported under Medical Device Reporting(21 CFR 803) doesn’t generate a second, duplicate report under Part 806, and a removal FDA itself ordered under its recall authority (21 CFR Part 810) isn’t reported back to the office that ordered it. Neither of those is the exemption in 806.1(b) — both are still health-risk corrections, still on FDA’s radar through a different door. The 806.1(b) exemption is the only one of the three that means FDA doesn’t receive the information under this part at all.

Exempt is not the same as undocumented

The exemption doesn’t mean the correction disappears from the world, only from FDA’s public record of it. 21 CFR 806.20still requires the firm to keep a record of every correction or removal it decides not to report — the device identifier, a description of the event, the corrective action taken, and the firm’s own justification for not reporting it. That record has to survive two years beyond the device’s expected life. It just never leaves the firm’s files. FDA never sees it, openFDA never indexes it, and no monitoring product, ours included, can watch a document that was never submitted.

A recall feed with nothing in it for a given firm is answering a narrower question than the one most people ask it. It says no reportable, health-risk correction has been classified. It says nothing about corrections that stayed under that bar by rule.

The practice

Reading a quiet recall history as a clean quality history is the mistake worth retiring. The honest reading is narrower and still useful: no correction has crossed the threshold Part 806 sets for reporting, as of the last classified record FDA has published. Whether a competitor’s silence reflects genuine quality discipline, a run of corrections that stayed on the exempt side of that line, or a report FDA hasn’t classified yet is a judgment call about a specific firm and a specific device — regulatory intelligence and method, never regulatory advice, and one for a regulatory professional weighing context this post doesn’t have.