IVDR Performance Evaluation: Scientific Validity Is Not the CER Habit It Looks Like

IVDR Article 56 requires manufacturers to demonstrate three things, not one: scientific validity, analytical performance, and clinical performance. Teams that built their regulatory instincts on MDR Clinical Evaluation Reports tend to arrive at performance evaluation already knowing the move — find the literature, summarize it, cite it. That move is built for clinical performance. Applied to scientific validity, it produces a report that reads like due diligence and functions like a gap.

Three pillars, one habit

Article 56(3) requires the manufacturer to specify and justify the level of clinical evidence necessary to demonstrate conformity, and Annex XIII Part A ties that evidence to scientific validity data, analytical performance data, and clinical performance data — documented, per Section 1.3.2, in a performance evaluation report that covers all three. For manufacturers who also run MDR Clinical Evaluation Reports, the instinct is to treat this as a familiar exercise with a new name. It is not. A CER's literature-and-clinical-data synthesis is built around one question — does the device perform as intended for its clinical purpose. IVDR performance evaluation asks that question and a prior one: does the thing the device measures actually mean what the manufacturer claims it means.

Where the literature-review reflex breaks

Scientific validity is defined around the association between an analyte or marker and a clinical condition or physiological state — a question that, for an established marker, may already have decades of published evidence behind it. That is exactly why the literature-review reflex feels safe here: the papers exist, the search returns results, the citation list looks substantial. But MDCG 2022-2 does not ask for a citation list. It asks for an appraisal — a documented assessment of each source's relevance to the manufacturer's specific claim and the methodological quality behind it. A marker can be well established in general and still leave the manufacturer's particular assay format, patient population, or intended use outside what the literature actually covered.

• Scientific validity is a marker claim, not a device claim. It establishes that the analyte means what the manufacturer says it means — separate from whether this specific device measures it accurately, which is analytical performance's job.
• A literature review has to be appraised, not just assembled. MDCG 2022-2's data-retrieval and appraisal approach evaluates relevance and quality; a reference list without that appraisal does not satisfy Annex XIII on its own.
• Legacy assays are not automatically covered. A marker used clinically for years can still have a published record that does not map cleanly onto the manufacturer's assay format, cut-off, or intended population — the gap a literature search alone will not surface.
• Novel markers rarely have the option. Where no substantial published record exists, scientific validity has to be built from the manufacturer's own data, not assumed from an absence of literature to the contrary.

A literature review can carry scientific validity when the literature genuinely closes the gap. Most manufacturers never check whether it does before writing the section as if it had. Why appraisal is the missing step

What this costs, and when

The gap rarely surfaces at the manufacturer's own review — it surfaces at the Notified Body's, when a reviewer asks for the appraisal behind a citation list that was never appraised, or asks what the literature actually says about the manufacturer's specific patient population rather than the marker in general. By that point the fix is a data-generation project on a Notified Body clock, not a writing exercise. The manufacturers who avoid it are the ones who treat scientific validity as a distinct IVD regulatory workstream from the start: state the marker-condition claim explicitly, appraise the evidence against it, and name the gaps before a reviewer does. That same discipline carries forward into Class D EURL batch verification and the broader IVDR legacy-device transition — each stage rewards teams that treated performance evaluation as three separate evidence bases rather than one narrative wearing three headings.

None of this asks for more evidence than IVDR already requires — it asks for the evidence to be organized around the claim it is actually supporting. A performance evaluation report that treats scientific validity as its own appraised, gap-checked workstream holds up the same way whether the marker is decades old or new to the market. One that inherits an MDR literature-review habit tends to work, right up until a Notified Body asks the question the review was never built to answer.