The sponsor proposes risk. The IRB decides.

An investigational device study classified nonsignificant risk (NSR) never reaches FDA as a filing at all. Under 21 CFR 812.2(b), meeting the abbreviated requirements and winning the reviewing institutional review board’s approval is treated as having an approved IDE, full stop — no submission, no FDA decision letter, no FDA reviewer ever opens the file. The sponsor who proposes that classification is not the party who gets to make it stick.

A proposal, not a filing

Under the abbreviated requirements, a sponsor brings the IRB a nonsignificant-risk determination and the reasoning behind it — typically the risk analysis the sponsor’s regulatory team already holds — and asks the IRB to concur. That request is a proposal. The IRB is required to make its own, independent assessment of whether the device actually meets the significant-risk criteria the regulation sets out, not simply adopt the sponsor’s label for it. FDA’s own description of the abbreviated pathway is direct about who decides: unless FDA has already made the determination for a given study, the IRB reviews the sponsor’s classification and can change it, on its own reading of the device and the protocol.

Agreement closes the filing. Disagreement opens a report.

When the IRB concurs, the abbreviated pathway closes the loop entirely — 812.2(b) treats that IRB approval as the approved IDE, and the study proceeds on it with no parallel FDA submission anywhere in the record. When the IRB does not concur — when it reads the same device and reaches significant risk where the sponsor proposed nonsignificant risk — the sponsor’s obligation is not to appeal to the IRB or re-argue the risk analysis. Under 812.150(b)(9), the sponsor has to report the IRB’s determination to FDA within five working days of first learning of it. The clock measures when the sponsor found out, not when the IRB voted, and it runs whether or not the sponsor still disagrees with the finding.

A nonsignificant-risk study has no FDA file to check for a disagreement, because a disagreement is what creates the file.

What that means for the record a sponsor keeps

A sponsor’s internal memo concluding NSR is not the determination 812.2(b) actually relies on; the IRB’s own written concurrence is. A study file that holds the sponsor’s risk analysis but not the IRB’s documented determination is missing the fact the abbreviated pathway turns on. And because an IRB that calls a device significant risk reopens the filing question entirely — the study cannot proceed as NSR until an approved IDE exists for it — the practical habit is to track each reviewing IRB’s determination as a live status, not a box checked once at study start. A multi-site study can produce more than one answer, because each reviewing IRB assesses the file in front of it independently; a split result across sites is itself something the sponsor has to track, and, at any site where the IRB disagrees with the sponsor’s own classification, report.

Why the mix-up happens

The confusion is almost always a vocabulary import from the drug side, where a sponsor’s own safety assessment carries more of the practical weight before an IND ever exists. Carrying that assumption into a device program reads the abbreviated requirements backwards: the IRB is not a courtesy reviewer ratifying a sponsor’s risk call so the study can start a week sooner — it is the body the regulation actually assigns the determination to, with the sponsor’s reporting duty under 812.150(b)(9) existing precisely because the sponsor does not get the final word.

This describes the mechanism in the regulation — who makes the significant-risk determination, and what a disagreement requires — not a risk determination for any specific study or device. That reading is regulatory intelligence and method, never regulatory advice; whether a given device meets the significant-risk criteria is a question for the sponsor’s regulatory team and the reviewing IRB, on the actual file.