IVDR performance evaluation is the IVD world's equivalent of MDR clinical evaluation. Instead of live-patient clinical data, an IVD manufacturer proves safety and performance through three linked pillars: scientific validity, analytical performance, and clinical performance. The evidence usually comes from blood, saliva, tissue, or other specimens, not from studies in which the device is applied directly to a patient's body.

By Tibor Zechmeister and Felix Lenhard.

TL;DR

  • Under IVDR, clinical evaluation is called performance evaluation, and it is structured around three pillars: scientific validity, analytical performance, and clinical performance.
  • Performance evaluation rarely requires studies on living patients. Most evidence comes from specimens: blood, saliva, tissue, urine, swabs.
  • The three pillars build on each other. Scientific validity justifies why an analyte matters, analytical performance proves the device can measure it, clinical performance proves the measurement delivers clinical benefit.
  • The deliverable is a Performance Evaluation Plan followed by a Performance Evaluation Report, analogous to the MDR's CEP and CER .
  • Roughly 80% of IVDR's framework overlaps with MDR. Classification and performance evaluation are the two major differences founders need to learn from scratch.

Why this matters

In Tibor's conversations with IVD startups, the single biggest source of confusion is the word "clinical". Founders coming from the MDR world, or from medical device textbooks written before 2022, expect to run a clinical investigation on patients. They budget for sites, ethics committees, recruitment, and months of informed consent logistics. Then they discover their product, a multiplex assay, a lateral flow test, a companion diagnostic, does not require anything of the kind.

IVDR uses a different word for a reason. An in vitro diagnostic device, by definition, operates outside the living body. The evidence standard is adapted to that reality. Tibor's framing: "In IVDR, clinical evaluation becomes performance evaluation. No live human subjects in most cases. Instead, blood, saliva, tissue samples are used to prove diagnostic accuracy or identification rate."

That shift changes the budget, the timeline, the sample sourcing strategy, the statistical plan, and the team you need to hire. It does not reduce the overall evidence burden. Many startups assume performance evaluation is "clinical evaluation lite" and underinvest. Notified bodies then push back, and the certification slips by six to twelve months while the performance evaluation is reconstructed.

What IVDR actually says

The IVDR is Regulation (EU) 2017/746. It became fully applicable on 26 May 2022, with staggered transitional provisions for legacy devices . Performance evaluation is governed by IVDR Article 56 and detailed in Annex XIII .

The regulation sets out that performance evaluation is "a continuous process by which data are assessed and analysed to demonstrate the scientific validity, analytical performance and clinical performance of a device for its intended purpose as stated by the manufacturer" .

Three pillars. All three required. All three documented.

Scientific validity is the association between the analyte or marker measured and the clinical condition or physiological state. Example: is there a recognised, published, biologically plausible link between elevated troponin and myocardial infarction? Scientific validity is largely a literature and state-of-the-art exercise. The manufacturer does not invent the association. The manufacturer demonstrates that the association is established.

Analytical performance is the ability of the device to correctly detect or measure the analyte. This is the domain of precision, accuracy, sensitivity, specificity, linearity, limit of detection, limit of quantification, measurement range, and interfering substances. Analytical performance lives in the lab. It does not require patient-facing studies.

Clinical performance is the ability of the device, when used as intended, to yield results that correlate with a clinical condition or a physiological or pathological process in accordance with the target population and intended user. Clinical performance is where diagnostic sensitivity, diagnostic specificity, positive and negative predictive values, and likelihood ratios live. Evidence can come from literature, routine diagnostic data, or a dedicated clinical performance study, depending on device class and novelty.

The output of this work is a Performance Evaluation Report that the notified body reviews. Higher-class devices (Class C and Class D under IVDR) face much deeper NB scrutiny than Class A or most Class B devices .

The regulation also requires a Performance Evaluation Plan written before the evidence generation starts. The plan defines which evidence will come from literature, which from in-house lab studies, which from retrospective specimens, and which, if any, from prospective clinical performance studies.

A worked example

Consider a small startup developing a fluorescence-based assay for a specific cancer biomarker. The device is intended to help clinicians monitor treatment response in patients already diagnosed and undergoing therapy.

Scientific validity. The team assembles a structured literature review covering every peer-reviewed publication linking the biomarker to the specific cancer and to treatment response. They document guideline references, pathway biology, and known confounders. They flag where the evidence is strong and where it is weak. This work does not produce new data. It establishes that the biomarker is a legitimate target. If the literature cannot support the link, the product cannot proceed. Tibor has seen founders skip this step, assuming the biology is "obvious", then struggle during NB review to produce a coherent scientific validity narrative.

Analytical performance. The team runs in-house studies. Precision across multiple operators and days. Accuracy against a reference method. Linearity across the claimed measurement range. Limit of detection and limit of quantification. Interference testing using haemolysed, lipaemic, and icteric samples. Sample stability under refrigeration, freezing, and room temperature. Each study has a protocol, a raw data set, a statistical analysis, and a report. The entire package feeds the Performance Evaluation Report.

Clinical performance. Because the intended population is patients already diagnosed with this cancer, and because retrospectively collected specimens are available from collaborating biobanks, the team negotiates access to several hundred characterised specimens with known clinical outcomes. They test the specimens blinded, compare results against the established reference standard, and calculate diagnostic sensitivity and specificity with confidence intervals. No patient is ever enrolled into a live study. The evidence is real, the sample size is justified statistically, and the ethical requirements for secondary specimen use are respected.

The total cost and timeline of this approach is a fraction of what an equivalent MDR clinical investigation would cost for a therapeutic device. That is the design intent of IVDR's performance evaluation pathway. It is not a shortcut, it is a regulation adapted to the nature of diagnostics.

The Subtract to Ship playbook

Felix's Subtract to Ship rule for IVDR performance evaluation: write the plan before you buy a single reagent. Every hour spent on a well-structured Performance Evaluation Plan saves a week of rework later.

Step 1. Classify first. IVDR has fewer classification rules than MDR but the rules are unintuitive. Classification drives NB involvement, drives the depth of performance evaluation, and drives the Performance Evaluation Report review. Get classification wrong and the entire plan is wrong. .

Step 2. Write the intended purpose carefully. As with MDR, intended purpose is the single document that determines everything else. For an IVD, it must specify the analyte, the specimen type, the target population, the clinical context, the intended user, and the claimed decision support. A vague intended purpose will force a broader, more expensive performance evaluation.

Step 3. Do the scientific validity work before the lab work. Founders instinctively want to build the assay first. Tibor's advice is the opposite. A literature-driven scientific validity review takes weeks, not months, and routinely surfaces issues that would have wasted a year of analytical development. Sometimes the literature reveals that the analyte is not the right target at all.

Step 4. Plan analytical performance around the published CLSI and ISO protocols. Analytical performance for IVDs has a rich standards landscape. Using recognised protocols protects the submission from methodological challenge by the NB.

Step 5. Decide the clinical performance route explicitly. For each performance characteristic, choose: literature, routine diagnostic data, retrospective specimens, or prospective study. Document the rationale. If a prospective clinical performance study is required, treat it with the same gravity as an MDR clinical investigation, because the regulatory process is equivalent in weight.

Step 6. Treat performance evaluation as continuous. IVDR makes it explicit that performance evaluation does not end at certification. Post-market performance follow-up feeds back into the Performance Evaluation Report over the device's lifetime. Build the infrastructure for continuous data review before you ship, not after.

Step 7. Budget for a harder notified body search. Tibor's observation: the pool of notified bodies designated for IVDR is significantly smaller than the pool for MDR. Capacity is tight. Start the NB conversation earlier than an MDR equivalent, not later.

Reality Check

  1. Do you know whether your device falls under IVDR or MDR, and can you justify the answer in writing?
  2. Have you written an intended purpose that specifies analyte, specimen, population, clinical context, user, and decision support?
  3. Can you name the three pillars of performance evaluation without looking them up?
  4. Do you have a Performance Evaluation Plan in draft before starting analytical studies?
  5. Is your scientific validity literature review structured, reproducible, and ready for NB review?
  6. Have you chosen recognised protocols for every analytical performance claim?
  7. Do you have a documented decision, with rationale, for how each clinical performance characteristic will be evidenced?
  8. Have you identified at least two notified bodies designated for IVDR in your device class and engaged them in preliminary conversation?

If three or more of these answers are no, the project is not yet at the stage where it can commit to a certification timeline.

Frequently Asked Questions

Does IVDR require clinical investigations on patients? In most cases, no. Performance evaluation relies on specimens and laboratory data. A prospective clinical performance study is only required for certain higher-class devices, novel analytes, or when literature and retrospective data are insufficient .

Is performance evaluation easier than clinical evaluation? Different, not easier. The effort shifts from patient recruitment to specimen sourcing, analytical rigour, and literature work. The documentation burden is comparable.

Can we use MDR clinical evaluation templates for IVDR? The structural logic transfers, but the content does not. The three pillars of IVDR performance evaluation do not map one-to-one onto MDR's clinical evaluation stages. A dedicated IVDR template is safer.

What is the Performance Evaluation Report? The consolidated document summarising all evidence under the three pillars, equivalent in function to the Clinical Evaluation Report under MDR. It is part of the technical documentation submitted to the notified body.

How often must performance evaluation be updated? Continuously, based on post-market performance follow-up. The exact update cadence depends on device class and risk profile .

Sources

  1. Regulation (EU) 2017/746 on in vitro diagnostic medical devices, Article 56 and Annex XIII [MDR VERIFY].
  2. Regulation (EU) 2017/745 on medical devices, Article 61 and Annex XIV, for comparison.
  3. EN ISO 13485:2016+A11:2021 and EN ISO 14971:2019+A11:2021.