Most IDE studies skip the IDE application

The fork lives in 812.2(b), not 812.20

812.20 is the section most people mean when they say ‘IDE application’ — the filing to FDA that opens a formal review. But 812.2(b) reaches the study first, and only for one branch of devices. For an investigation of a device other than a significant risk device, the regulation's abbreviated requirements apply instead: label the device under 812.5, obtain and maintain IRB approval after presenting the IRB with a brief explanation of why the device isn't significant risk, secure informed consent from each subject, meet the monitoring obligations, and keep the required records and reports. Nothing in that list involves FDA. A nonsignificant-risk study's version of an IDE application is the explanation handed to an IRB, not a filing FDA ever sees unless something goes wrong.

The sponsor makes the call first

812.3(m) defines a significant risk device along four tests: an implant that presents a potential for serious risk; a device purported or represented for use in supporting or sustaining human life that presents that potential; a device of substantial importance in diagnosing, curing, mitigating, or treating disease, or otherwise preventing impairment of human health, that presents that potential; or a catchall — a device that otherwise presents a potential for serious risk to a subject's health, safety, or welfare. None of the four tests gets applied by FDA before a study starts. 812.2(b) puts the first determination on the sponsor, who has to give the reviewing IRB a brief explanation of why the device clears the nonsignificant-risk bar. If the IRB disagrees — determines the study actually involves a significant risk device — 812.66 requires it to notify the investigator and sponsor, and the sponsor can't proceed except as 812.30(a) allows. A risk determination a sponsor gets wrong doesn't quietly stay wrong; it routes the study back through the track it should have used from the start.

What the filing actually has to contain

For a significant risk device, 812.20 requires a complete report of prior investigations of the device and either an accurate summary of the investigational plan's purpose, protocol, risk analysis, device description, and monitoring procedures — the elements 812.25(a) through (e) name individually — or the complete plan itself. The sponsor owes FDA the full plan and the full report, not the summary, whenever no IRB has reviewed them, FDA has found an IRB's review inadequate, or FDA asks for them directly. A protocol document built to satisfy an IRB's own review checklist isn't automatically the document 812.20 requires if the file ever needs to go to FDA instead.

A clock that runs on silence

812.30 gives FDA thirty days to act on a complete application, and the default direction of that clock cuts the other way from a 510(k)'s. An investigation may not begin until thirty days after FDA receives the application, but if FDA hasn't notified the sponsor by then that the study is approved, approved with conditions, or disapproved, the sponsor may proceed — approval by the agency's silence, not its letter. A submission tracker built on the assumption that every FDA pathway ends in an explicit clearance is watching for an event 812.30 doesn't actually require.

An IDE risk-determination record built around this fork — the 812.2(b) explanation, the 812.3(m) test behind it, and the full-plan trigger in 812.20 — is previewed in the launch catalog. A team whose IRB and sponsor read the risk test differently is exactly the case a well-built Q-Submission request exists to resolve before a study starts rather than after an inspector asks about it; if your program documents that resolution differently, the shelf takes that correction directly.