Your Test Doesn’t Treat Anyone. It Decides Who Gets Treated.
Regulatory strategy for IVDs — analytical validity, clinical validity, and the IVDR clock for assays and the instruments that run them.
A Diagnostic Never Touches the Patient. It Can Still Kill Them.
The harm from an IVD is entirely indirect, which is what makes the category counterintuitive. A false negative on an infectious disease assay sends an untreated patient home. A false positive on a cancer marker starts a therapy nobody needed. The device sat in a laboratory the whole time — the damage was done by a clinician acting on a number.
So regulation follows the consequence, not the contact. Classification tracks the decision your result drives; evidence expectations track the population you claim; and the specimen you never see — collected by someone else, transported by someone else — is part of your system. The whole file is an argument that a number produced in your box means what your label says it means.
Collection, transport, storage. Half your error budget is spent before the sample arrives.
Two Questions. Most Sponsors Only Answer the First.
Analytical validity and clinical validity are different claims with different evidence. Confusing them is the most common structural error in IVD development.
Question one · Analytical validity
Does it measure what you say, accurately?
- Precision: repeatability and reproducibility, across sites, days, operators, and lots
- Accuracy: trueness against a reference method or material
- Sensitivity: limit of blank, detection, and quantitation
- Specificity: interference and cross-reactivity — the things that look like your analyte
- Range, stability, and lot-to-lot consistency
Question two · Clinical validity
Does the number mean what you claim?
- Clinical sensitivity and specificity against a defined clinical truth
- Predictive values — and the prevalence that makes them move
- The intended-use population, not a convenience sample
- A comparator: gold standard, composite reference, or clinical outcome
- Cut-off justification — where you drew the line and why
IVDR moved most tests into notified body review — and the notified bodies are the bottleneck.
Under the Old Directive Most Tests Self-Certified. Under IVDR Most Do Not.
The IVDR inverted Europe’s risk classification, pulling the large majority of assays into notified body assessment for the first time — against a notified body population far too small for the volume. The transition deadlines are staged and moving through them now: Class D devices had their application and agreement dates in 2025 with market access to end-2027; Class C applications are due by 26 May 2026, agreements by September 2026, and transition to end-2028; Class B and sterile Class A follow in 2027 with transition to end-2029.
And legacy devices must be registered in EUDAMED by 27 November 2026 even while running on transition provisions. The practical reality behind those dates is queue: review timelines commonly run over a year, and applications filed today complete well into 2027. The deadline that matters is not the transition date — it is the date your notified body will accept the file.
What an IVD Program Plans Around.
Three numbers that decide the program.
Analytical and clinical. Different claims, different evidence — and only one of them is a bench exercise.
The variable that turns a 99%-specific test into a false-positive machine. It belongs in the claim, not the appendix.
The IVDR Class C application deadline — and the notified body queue behind it is the real constraint.
Six Failure Modes We Are Brought In to Prevent.
Most are a strong assay meeting a weak claim.
Analytical validity offered as clinical
Excellent precision data presented for a claim that needed clinical performance in the intended population.
Convenience samples
Banked specimens that do not represent the intended-use population — and performance that will not survive contact with it.
Prevalence arithmetic ignored
A screening claim in a low-prevalence population, with predictive values nobody computed before the reviewer did.
Pre-analytical variables unstudied
Collection tube, transport time, and freeze-thaw — the error budget outside your box that lands inside your claim.
IVDR met at the deadline
An application filed as the transition date approaches, queued behind everyone else with the same plan.
Software treated as instrument firmware
Algorithms and middleware carrying the interpretation, with no software file behind them.
IVD Leadership That Reads the Prevalence Before the Protocol.
Our IVD leads have built analytical and clinical validation packages, justified cut-offs, and moved portfolios through the IVDR transition.
“A 99% specific test sounds excellent until you screen a population where the disease is rare. Then it is a machine for generating anxiety and biopsies.”
The discipline we bring to assays, analyzers, and diagnostic systems.
Building an IVD? Test the Claim Against the Population, Not the Bench.
Bring senior IVD leadership in before the clinical study locks — and before the IVDR queue closes on you.
Senior-led. Embedded in your team. No junior hand-offs.