An IVDR performance evaluation is three tests, not one

A CE-marked medical device closes its clinical case with one report built to answer one question: does the evidence support safety, effectiveness, and an acceptable benefit-risk ratio, under MDR Article 61. An in vitro diagnostic device closes its case differently, and a team that pattern-matches an IVDR performance evaluation onto the clinical evaluation report structure it already knows will build the wrong document. IVDR Article 56 splits the evidence into three legs that don't collapse into each other: scientific validity, analytical performance, and clinical performance. Each has its own test, its own kind of study, and its own way of going wrong — and a performance evaluation report that blends them into one narrative has usually only actually demonstrated one of the three.

Three legs, one report

Article 56(1) requires conformity with the general safety and performance requirements to rest on scientific validity, analytical performance, and clinical performance data, sufficient to establish clinical evidence for the device's intended purpose. Scientific validity is the association between an analyte and a clinical condition or physiological state — the reason measuring this substance says anything about this disease at all. Analytical performance is the device's ability to correctly detect or measure that analyte: trueness, precision, sensitivity, specificity, limit of detection. Clinical performance is the device's ability to yield results that correlate with the clinical condition or physiological state in the intended population. All three feed one document — the performance evaluation report referenced in Section 1.3.2 of Part A of Annex XIII — but the regulation tests each leg on its own terms before that document gets to synthesize them.

Scientific validity is the leg a familiar analyte tempts you to skip

For a well-established analyte, it's tempting to treat the disease association as self-evident and skip straight to performance testing — everyone already knows what this marker means. IVDR doesn't accept that as a substitute for the documented association Article 56 actually asks for. The scientific-validity case still has to be built through a systematic review of the literature already establishing the analyte-condition link, assembled — in practice, for most manufacturers — as its own standalone report feeding the overall performance evaluation. Familiarity with the analyte isn't evidence; the literature review is.

Analytical performance has to be your own data

As a general rule, analytical performance has to be demonstrated through the manufacturer's own analytical performance studies — run on the actual device, against the actual analyte, not inherited from a comparable product's published data. This is the one leg of the three where borrowed evidence has essentially no role: a competitor's trueness and precision figures say nothing about whether your own assay, reagents, and manufacturing process hit the same marks.

Clinical performance's easier path narrows with device class

Article 56(4) allows clinical performance to draw on three sources: the manufacturer's own clinical performance studies, peer-reviewed scientific literature, or published experience gained from routine diagnostic testing — but studies are required unless the manufacturer can duly justify relying on one of the other two instead, and a new or novel analyte rarely clears that bar. The justification gets harder to sustain as device class rises: Class C and D devices additionally face Notified Body review of the performance evaluation report itself, and under Article 29 they require a Summary of Safety and Performance — validated by the Notified Body and posted publicly to Eudamed — a disclosure obligation with no equivalent at the lower classes.

Where this crosses the rest of the file

The three-legs structure runs on a different logic than the MDR clinical evaluation's own equivalence route: a CER can lean on an already-marketed comparator device's literature under Article 61(3), while IVDR's version of borrowed evidence is narrower from the outset and closes further as class rises, rather than opening a single alternate pathway the way MDR's equivalence provision does. It's the same lesson in a different shape as a biological evaluation plan's own contact-and-duration grid: a framework that proposes a starting structure for evidence still leaves the manufacturer to justify, on the record, why a given leg was satisfied the way it was — not to treat the framework's existence as the justification.

A performance evaluation record built around this structure — the three legs demonstrated and documented separately, clinical-performance sourcing justified rather than defaulted to literature, and the Class C/D reporting obligations tracked on their own line — is previewed in the launch catalog. If your program's evidence sourcing works differently, the shelf takes that correction directly.