"Refer to the IFU" is not a clear instruction.

I have read enough IVD protocols in the last fifteen years to know which sections are written carefully and which are filled in. The eligibility criteria are usually written carefully — sponsors and authors think hard about who counts as a subject. The endpoints are written carefully. The statistical analysis plan is written carefully. And then there is a category of fields that recurs in protocol after protocol with a phrase that has, somewhere along the way, become accepted shorthand for "we are not going to specify this." The phrase is "refer to the IFU."

It appears in sample handling sections. It appears in storage and shipping. It appears in test procedure descriptions. It appears, on occasion, in fields where it does not even apply, suggesting it is now a reflexive insertion rather than a deliberate one. The intent is reasonable: the assay's Instructions for Use document is the authoritative specification, and a protocol that contradicts the IFU creates a regulatory hazard. If the protocol points to the IFU, the protocol cannot conflict with it. The argument is sound on its face.

The argument fails for one simple reason. The investigator does not have the IFU at the moment they need it. The site coordinator processing a sample at 4:30 p.m. on a Friday does not stop to retrieve the IFU. The PI signing a consent does not consult the IFU between the patient's questions. The data manager building the CRF does not pull from the IFU when defining field constraints. The IFU lives, in practice, in a regulatory binder somewhere in the site's office or on a sponsor's portal that requires a separate login. The site staff who are doing the actual study work are doing it from the protocol — and what the protocol says, in the fields where it could have been specific, is "refer to the IFU."

The result is that the most consequential operational specifications on a clinical trial are deferred to a document that nobody actually opens during the trial. Sample handling deviations, storage protocol breaches, shipping condition mismatches — the things that compromise samples and disqualify subjects — happen at the moments where the protocol failed to specify and the IFU was not consulted. The IFU did not fail. The protocol failed.

The fix is not complicated. It is to write the relevant content of the IFU into the protocol, in the protocol's voice, in the section where the site staff will actually look. Yes, this means the IFU's text is reproduced in the protocol. Yes, this means the protocol is longer. Both are acceptable costs. The alternative is a protocol that is shorter on paper and longer in execution, because every "refer to the IFU" is an unanswered question that becomes a deviation, a query, or a missed sample at some point in the study.

The deeper concern is what the phrase signals about how the protocol was written. Protocols that say "refer to the IFU" in many places are usually protocols that were assembled rather than authored. A medical writer assembled the document from sponsor-provided sections, the IFU was not in the writer's pile, and the placeholder hardened into the final draft. Protocols that were authored — by someone who took responsibility for the document as a whole and asked, at every section, what the site staff would need to know to do their work — do not contain the phrase, because the author would have felt the gap and filled it.

What I tell sponsors who are designing IVD protocols is this. Treat the protocol as a document the PI and site coordinator will read at every visit, with no other resources at hand. Anything the protocol defers to another document — the IFU, the lab manual, the central lab specification — is a question the site is going to have to interrupt their work to answer, and the interruptions accumulate into protocol deviations whether or not anyone notices. The protocol is the source of truth, or it is not. Halfway is the worst position.

The IFU is a regulatory artifact. It exists for a different audience and a different purpose. Treating it as a substitute for protocol specificity confuses the two functions. The protocol's job is to tell the people running the study what to do. If it does not tell them, no one will.