When Analytical Validation Alone Is Not Enough for Your IVD
You submitted an impressive analytical validation package. The Notified Body sent it back. Not because the data was weak, but because the clinical question was never answered. This happens more than you think.
In This Article
Most IVD manufacturers build their entire regulatory strategy around analytical performance. Sensitivity. Specificity. Precision. Linearity. They invest months in these studies, and rightfully so. But then comes the question that stops everything: Does this device perform as intended in the target population?
Analytical validation tells you the device works under controlled conditions. Clinical performance tells you it works in real use. The gap between these two is where many submissions collapse.
Under IVDR Article 58, clinical evidence is required unless you can justify its absence. That justification is narrower than most manufacturers assume. And when you cannot avoid clinical data, the question is not whether you need a performance study. The question is how to design one that actually answers what regulators and Notified Bodies will ask.
When Analytical Data Is Not Sufficient
The default assumption in many IVD projects is that analytical validation is enough. After all, you demonstrated that your assay detects the target analyte with high accuracy. You showed repeatability and reproducibility. You validated against reference materials.
But here is the issue. Analytical performance is measured in ideal conditions. Controlled sample types. Known concentrations. Trained operators. A clinical setting is none of that.
Your device will be used on samples from patients with comorbidities, interfering substances, edge-case presentations. The operator may not be a laboratory specialist. The workflow may introduce variables you never tested. This is why IVDR Article 58 requires you to demonstrate performance in conditions that reflect actual use.
Manufacturers submit analytical validation reports and assume they have covered clinical performance. The Notified Body rejects the file because no data addresses how the device performs with real patient samples in the intended use setting.
If your device is used to diagnose a condition, analytical validation alone does not prove diagnostic accuracy. If it monitors disease progression, you need data showing it reflects clinical status over time. If it guides treatment decisions, you must show it provides actionable information that aligns with clinical outcomes.
The gap is not optional to fill. It is a regulatory requirement.
What Triggers the Need for Clinical Data
There is no universal checklist, but some patterns emerge from Notified Body reviews and MDCG guidance. Certain device characteristics, intended uses, and claims make clinical data unavoidable.
High-risk classifications are the most obvious trigger. Class C and D IVDs almost always require clinical performance data. The stakes are too high to rely on analytical studies alone. These devices influence major medical decisions, and regulators expect direct evidence that the device performs in the target population.
Novel analytes or novel technologies also trigger this requirement. If your device measures something that has not been widely used in clinical practice, analytical validation does not address whether the measurement is clinically meaningful. You may detect a biomarker accurately, but does it correlate with disease presence, severity, or response to treatment? That question requires clinical data.
Similarly, if you introduce a new measurement principle or platform, you cannot assume equivalence to established methods. Even if the analytical performance is comparable, the clinical interpretation of results may differ. Reviewers will ask for evidence that clinicians can use your device the same way they used the predicate.
The moment your device introduces a clinical interpretation challenge, analytical data becomes insufficient. The regulatory focus shifts from “does it measure accurately” to “does it inform clinical decisions correctly.”
Another common trigger is a claim that implies clinical utility. If you state that your device improves patient outcomes, speeds diagnosis, or reduces unnecessary interventions, that is a clinical claim. You must support it with clinical data. Analytical validation does not demonstrate utility. It only demonstrates measurement capability.
Finally, if your device targets a population with high variability, you may need clinical data even for straightforward analytes. Pediatric populations, patients with rare diseases, or settings with limited infrastructure introduce variables that analytical studies do not capture. The question becomes whether the device remains reliable when those variables are present.
Designing a Performance Study That Answers the Right Question
Once you accept that clinical data is unavoidable, the next step is designing a study that actually satisfies regulatory expectations. This is where many manufacturers stumble. They design studies that generate data but do not answer the clinical performance question.
The first principle is to define the clinical question clearly. What does the device need to demonstrate? Diagnostic accuracy? Prognostic value? Ability to monitor treatment response? Each question requires a different study design, different endpoints, and different comparators.
If you are claiming diagnostic accuracy, your study must compare device results to a clinical reference standard. That reference is not always another IVD. It may be histopathology, clinical follow-up, or an independent diagnostic algorithm. Whatever it is, it must be accepted as the truth against which your device is judged.
The sample size must be adequate to demonstrate the claimed performance with statistical confidence. Too often, manufacturers propose studies with 50 or 100 samples and expect to claim 95% sensitivity. The confidence intervals are too wide. The Notified Body will reject it or request additional data.
Sample size calculations are missing, or they are based on assumptions that do not align with the claimed performance. Reviewers see this immediately and flag it as a critical gap.
The study population must reflect the intended use population. If your device is intended for use in symptomatic patients, enrolling only confirmed positive cases does not demonstrate real-world performance. You need a spectrum of patients that includes the diagnostic gray zone, borderline results, and conditions that might interfere or confuse the diagnosis.
Blinding and independence are critical. If the device result influences the reference standard determination, or if the same operator performs both tests, the study is biased. Regulatory reviewers are trained to spot this. They will question the validity of your conclusions.
Endpoint definition matters. If you are measuring agreement with a reference method, you need to predefine what level of agreement is acceptable and clinically meaningful. If you are assessing clinical utility, you need to show that device results influence clinical decisions in the expected direction.
Prospective Versus Retrospective Data
Another common question is whether retrospective data is acceptable or whether a prospective study is required. The answer depends on what you are trying to demonstrate and what data is available.
Prospective studies are stronger because they eliminate many sources of bias. The protocol is predefined. Samples are collected specifically for the study. The reference standard is applied consistently. Reviewers trust prospective data more because it is harder to manipulate.
But prospective studies are expensive and time-consuming. If your device is in a competitive market, waiting 12 months for a prospective study may not be feasible. This is where retrospective data becomes attractive.
Retrospective data can be acceptable if it is well-documented, representative, and includes appropriate comparators. The challenge is that retrospective studies often lack the rigor of a predefined protocol. Sample selection may be biased. Reference standard application may be inconsistent. These gaps are visible to reviewers.
A hybrid approach is sometimes used. A retrospective study provides initial evidence, and a prospective study is planned as part of post-market clinical follow-up. This allows you to submit the file earlier while committing to generate stronger evidence post-market.
If you propose retrospective data, expect to justify why it is representative and unbiased. If you cannot make that case convincingly, plan for a prospective study from the start.
What Notified Bodies Actually Look For
Understanding what Notified Bodies look for in a performance study saves you months of back-and-forth. Their review is not arbitrary. They follow MDCG 2022-2 and assess whether your study answers the clinical performance question.
They check whether the study objective aligns with your intended use and claims. If you claim diagnostic use but your study only shows analytical agreement, they will reject it. The study must demonstrate what you claim.
They review the population inclusion and exclusion criteria. Are you testing the device in the population where it will be used? If your device is for suspected infections but you only tested confirmed cases, that is a gap. The study does not reflect real use.
They assess the reference standard. Is it appropriate? Is it applied independently? If the reference is weak or circular, the entire study loses credibility.
They examine the statistical analysis plan. Was it predefined? Are the endpoints clear? Is the sample size justified? If these elements are missing or added post hoc, reviewers assume the study was not rigorously planned.
They look for protocol deviations and how they were handled. Deviations happen in every study, but how you document and justify them matters. If deviations are frequent or unexplained, it signals poor study conduct.
They review the results critically. Do the confidence intervals support the claims? Are subgroup analyses justified? Are negative or inconclusive results acknowledged? If the report only highlights favorable results and ignores limitations, reviewers become skeptical.
Study reports present results without discussing limitations or alternative interpretations. This does not reassure reviewers. It makes them question whether the manufacturer understands the data.
Integrating Performance Data Into Your Clinical Evidence
A performance study is not a standalone document. It must integrate into your overall clinical evidence framework. This means connecting it to your analytical validation, your literature review, and your post-market surveillance plan.
In your Clinical Evidence Report, the performance study should address gaps that analytical data cannot fill. It should demonstrate that the device performs in real conditions, with real patients, in a way that supports the intended use.
If the study reveals limitations, acknowledge them and explain how they will be addressed post-market. For example, if your study included a limited number of rare disease cases, commit to collecting additional data through PMCF. This shows regulatory maturity.
The performance study also informs your risk management. If the study identifies failure modes, edge cases, or interfering conditions, these must be reflected in your risk analysis and your labeling. Clinical data and risk management must align.
Finally, the study should feed into your PMCF plan. What questions remain unanswered? What populations were underrepresented? What performance metrics need long-term monitoring? Your PMCF plan should be the logical extension of your premarket clinical evidence.
Final Reflection
When you cannot avoid clinical data, the question is not whether to do a performance study. The question is whether you understand what needs to be demonstrated and how to design a study that convinces regulatory reviewers.
Analytical validation is necessary but not sufficient. Clinical performance is what ties your device to real use. The gap between these two is not filled with assumptions. It is filled with evidence.
Next, we will address what happens when your performance study reveals results that do not fully align with your claims. How do you handle inconclusive data, unexpected findings, and performance gaps without derailing your submission?
Peace,
Hatem
Clinical Evaluation Expert for Medical Devices
Follow me for more insights and practical advice.
Frequently Asked Questions
What is a Clinical Evaluation Report (CER)?
A CER is a mandatory document under MDR 2017/745 that demonstrates the safety and performance of a medical device through systematic analysis of clinical data. It must be updated throughout the device lifecycle based on PMCF findings.
How often should the CER be updated?
The CER should be updated whenever significant new clinical data becomes available, after PMCF activities, when there are changes to the device or intended purpose, and at minimum during annual reviews as part of post-market surveillance.
What causes CER rejection by Notified Bodies?
Common reasons include inadequate equivalence demonstration, insufficient clinical data for claims, poorly structured SOTA analysis, missing gap analysis, and lack of clear benefit-risk determination. Structure and logical flow are as important as the data itself.
Which MDCG guidance documents are most relevant for clinical evaluation?
Key documents include MDCG 2020-5 (Equivalence), MDCG 2020-6 (Sufficient Clinical Evidence), MDCG 2020-13 (CEAR Template), MDCG 2020-7 (PMCF Plan), and MDCG 2020-8 (PMCF Evaluation Report). IVDR Article 58, MDCG 2022-2
Need Expert Help with Your Clinical Evaluation?
Get personalized guidance on MDR compliance, CER writing, and Notified Body preparation.
✌
Peace, Hatem
Your Clinical Evaluation Partner
Follow me for more insights and practical advice.
– IVDR 2017/746 Article 58
– MDCG 2022-2: Clinical Evidence for In Vitro Diagnostic Medical Devices
Deepen Your Knowledge
Read Complete Guide to Clinical Evaluation under EU MDR for a comprehensive overview of clinical evaluation under EU MDR 2017/745.
