IVDR Performance Studies: Why Article 58 Authorization Isn't Automatic

IVDR Article 58 splits performance studies into two tracks, and the dividing line is not the one most teams reach for. It is not "is this a companion diagnostic study" or "are we using banked samples." It is whether the study is interventional, adds an invasive procedure, or takes samples specifically for the study rather than as remnants of standard-of-care collection. Get the classification wrong in either direction and you either file a notification the regulation says needed full authorization, or you build an authorization dossier the study never required.

The fork Article 58 actually draws

Article 58(1) puts a performance study on the authorization track when any one of three things is true: the study is an interventional clinical performance study, surgically invasive sample-taking is done solely for the purposes of the study, or the study involves additional invasive procedures or other risks to subjects beyond what standard of care would otherwise require. Any one trigger is enough. A study that would otherwise read as low-risk — a validation run alongside routine testing, say — crosses into the authorization track the moment it adds a procedure, a visit, or a sample draw the subject would not have had anyway.

Why the leftover-sample carve-out gets misapplied

Article 58(2) treats companion diagnostic studies the same as any other Article 58(1) study, with one exception: a study using only leftover samples needs notification rather than authorization. Teams routinely over-read that exception. "Leftover" under ISO 20916 means remnants of specimens already collected for standard-of-care purposes — not samples banked for research generally, and not samples drawn prospectively on a study-specific schedule even if the draw itself is routine. The moment a protocol adds a sampling visit, a larger draw than clinical care required, or any procedure performed because the study asked for it rather than because the patient's care called for it, the study is back on the Article 58(1) authorization track regardless of how the sponsor labeled it.

• Genuine remnant, not a research bank. The sample has to be left over from a standard-of-care collection — not drawn or retained because the study exists.
• No added procedure for the study alone. Any invasive step performed specifically to support the performance study, however minor, forfeits the carve-out.
• Notification is still a filing, not a shortcut. The study must still be notified to the competent authority before it starts.
• Ethical review still applies. Article 58(3) requires review by an ethics committee under national law on both tracks, with at least one lay member taking part.

Article 58 is not asking whether a companion diagnostic is involved. It is asking whether any subject bore risk, procedure, or sampling that standard of care would not otherwise have required. Why the leftover-sample test fails under scrutiny

The ethics and modification clocks don't wait for EUDAMED

Competent-authority review and ethics-committee review run in parallel under Article 58(3), not in sequence — sponsors who treat ethics approval as a step that follows authorization, rather than one that runs alongside it, build in weeks of avoidable delay. Once a study is underway, Article 71 governs substantial modifications: the sponsor notifies the relevant member state, and the earliest a modification can be implemented is 38 days later, unless the authority raises a ground for refusal under Article 71(3)(a) or an ethics committee issues a negative opinion — a window the authority can extend by a further seven days if it needs to consult outside experts. The June 2025 MDCG 2025-5 Q&A guidance gives a non-exhaustive list of what the Medical Device Coordination Group treats as substantial, and it is worth checking a planned amendment against that list before it is drafted, not after.

None of this is exotic once it is mapped, but it rewards doing the mapping early. A sponsor that classifies a study correctly at protocol design avoids both failure modes: filing a notification for a study that needed authorization, and over-building an authorization dossier for a study that only needed one. For multi-country studies, the regulation lets a sponsor name a single coordinating member state and submit one application — but with the EUDAMED performance studies module still outside the modules that became mandatory to use in May 2026, that coordination runs through national systems in practice, not a single EU-wide portal. Build the IVD regulatory strategy around the Article 58 classification first; everything else in the study plan follows from which track you are actually on.