Why your diagnostic’s clinical benefit claim keeps failing review

Written by HATEM RABEH, MD, MSc Ing

Your Clinical Evaluation Expert And Partner

in
S

This confusion appears in nearly every diagnostic device submission I review. Manufacturers confuse performance characteristics with clinical benefit. They present sensitivity, specificity, and accuracy as if these metrics answer the clinical benefit question. They don’t.

The regulatory framework is clear, but the path from regulation to acceptable evidence is not obvious. Most teams know they need to demonstrate clinical benefit under MDR Article 61. Fewer understand what that demonstration actually requires for a diagnostic device.

What Clinical Benefit Actually Means for Diagnostics

Clinical benefit for a diagnostic device is not about how well it measures something. It’s about what happens to the patient because the measurement exists.

MDR Annex XIV Part A defines clinical benefit as the positive impact of a device on the health of an individual, expressed in terms of meaningful, measurable, patient-relevant outcome. For therapeutics, this connection is direct. For diagnostics, it operates through a chain.

The diagnostic provides information. Clinicians use that information to make decisions. Those decisions lead to actions. Those actions affect patient outcomes.

Most deficiency letters I see point to this missing connection. The manufacturer provides excellent technical performance data. Then they leap directly to claiming clinical benefit without establishing the intermediate steps.

The Evidence Chain Reviewers Expect

When I review a clinical evaluation for a diagnostic device, I look for three linked evidence sets. Each must be present. Each must connect to the next.

First: Analytical performance. This is the technical accuracy, precision, sensitivity, and specificity of the measurement itself. This data answers: does the device measure what it claims to measure, and how reliably?

This is necessary but not sufficient. It establishes that your device works as a measurement tool. It does not establish clinical benefit.

Second: Clinical performance. This data establishes the connection between the diagnostic result and clinical decision-making. It answers: do clinicians use this information? How does it change their diagnostic or therapeutic approach?

MDCG 2022-2 provides the framework here. Clinical performance means the ability to yield diagnostically relevant information. You must show that the information your device provides is actually used in clinical practice and influences medical decisions.

For some established diagnostic parameters, this connection is well-documented in literature. For novel biomarkers or new testing approaches, you need to generate this evidence yourself.

Third: Patient outcomes. This data connects the clinical action to actual patient benefit. It answers: do patients have better outcomes when clinicians have access to this diagnostic information?

This is where most submissions become thin. The manufacturer assumes that if the test is accurate and if doctors use it, benefit automatically follows. Reviewers do not accept this assumption.

When Literature Can Support the Chain

For well-established diagnostic parameters measured through standard methods, published literature often provides the complete evidence chain. If you’re developing another troponin assay for cardiac injury, extensive outcome literature exists connecting troponin measurement to therapeutic decisions and patient outcomes.

But three conditions must be met before you can rely on this literature:

Your device must measure the same parameter. Not a similar parameter. The same one. A novel cardiac biomarker cannot borrow troponin outcome data.

Your device must have equivalent analytical performance to the methods used in the outcome studies. If the outcome literature is based on laboratory methods and your device is a point-of-care test with different performance characteristics, the connection weakens.

The clinical setting must align. Outcome data from emergency department chest pain protocols may not transfer cleanly to home monitoring scenarios.

When these conditions hold, your clinical evaluation can reference established outcome literature to complete the evidence chain. When they don’t hold, you need to generate your own outcome evidence or clearly acknowledge the gap and address it through post-market activities.

The Novel Biomarker Problem

The evidence challenge becomes acute when your diagnostic measures something new or uses a new methodology for an established parameter.

I reviewed a submission for a device measuring a proprietary inflammatory marker panel for early sepsis detection. The analytical validation was excellent. The manufacturer cited literature showing that early sepsis detection improves outcomes. They claimed clinical benefit.

The Notified Body rejected this reasoning. The outcome literature was about early detection through established criteria and markers. This device used different markers. The critical question remained unanswered: do patients have better outcomes when physicians make decisions based on this specific panel versus existing approaches?

This question requires comparative outcome data. Most manufacturers do not want to hear this because generating outcome data is expensive and time-consuming. But the alternative is a clinical benefit claim that cannot be defended.

For truly novel diagnostics, the regulatory pathway often involves conditional approval with mandatory PMCF studies designed to generate the missing outcome evidence. This approach is acceptable if the submission acknowledges the evidence gap clearly and if the PMCF plan is designed to close it within a reasonable timeframe.

Companion Diagnostics and Linked Benefit

Companion diagnostics present a special case. These devices guide therapeutic decisions for specific drugs or treatments. The clinical benefit question becomes: do patients selected using this diagnostic have better therapeutic outcomes than unselected patients or patients selected by other methods?

The evidence standard here is typically higher because the diagnostic directly gates access to therapy. If your companion diagnostic incorrectly excludes patients who would benefit or includes patients who won’t respond, the consequences are immediate and significant.

Regulatory reviewers expect to see evidence that the diagnostic-therapeutic combination produces better outcomes than alternative approaches. This usually means data from the clinical trials that established the therapeutic efficacy, analyzed by diagnostic result.

When such data exists, the benefit claim is straightforward. When it doesn’t exist, or when your diagnostic is a follow-on to an established companion diagnostic using different methodology, the evidence requirements mirror the novel biomarker challenge.

What Acceptable Evidence Looks Like in Practice

Let me describe two contrasting submissions I reviewed recently.

Submission A: A point-of-care coagulation monitor for anticoagulation therapy management. The clinical evaluation presented analytical performance data showing good agreement with laboratory methods. It then presented multiple outcome studies demonstrating that patient self-testing with POC coagulation monitors, when combined with self-management or physician management, leads to better time-in-therapeutic-range and reduced thromboembolic events compared to standard laboratory monitoring.

The evidence chain was complete. The device measures INR accurately. Patients and physicians use INR results to adjust anticoagulation therapy. Studies show that POC testing in this context improves outcomes. Approved.

Submission B: A novel imaging device for skin lesion analysis using multispectral imaging and AI interpretation. The clinical evaluation presented diagnostic accuracy data compared to histopathology. Sensitivity and specificity were good. The evaluation then claimed clinical benefit based on earlier melanoma detection leading to improved survival.

The evidence gap was obvious. No data connected this device to earlier detection in practice. No data showed that physicians changed their biopsy decisions based on this device’s input. No outcome data demonstrated actual survival benefit in a population screened with this device versus standard approaches.

The Notified Body issued a major deficiency. The manufacturer needed to either provide comparative outcome data or restructure the clinical benefit claim to reflect the actual evidence level with appropriate PMCF commitments.

PMCF Strategy for Benefit Evidence

When pre-market evidence cannot complete the benefit chain, post-market clinical follow-up must be designed to close the gap. This is not optional under MDR Article 61(11).

A PMCF plan that only monitors safety and confirms analytical performance is insufficient if clinical benefit evidence was weak at submission. The PMCF plan must specify how outcome data will be collected, what endpoints matter, and what timeline is realistic.

For a diagnostic device with a novel approach, an acceptable PMCF plan might include:

A registry study tracking patients tested with the device, documenting what clinical actions resulted from test results, and comparing outcomes to matched controls or historical data.

A comparative effectiveness study in a subset of sites where clinical outcomes are systematically tracked.

Prospective data collection at sites that adopt the device, with pre-specified outcome measures and analysis timepoints.

The key is specificity. “We will collect outcome data” is not a plan. “We will establish a registry at ten sites, enroll 500 patients over two years, track hospital admissions and treatment modifications, and compare outcomes to propensity-matched controls from the same sites in the pre-implementation period” is a plan.

Reviewers evaluate whether your PMCF plan will actually generate evidence that could validate or refute your benefit claim. If it won’t, they will require a more robust plan before approval.

The Software and AI Complication

When your diagnostic includes AI-based interpretation or clinical decision support, the benefit evidence requirements intensify.

Analytical performance data for AI diagnostics shows how accurately the algorithm identifies features or classifies images. Clinical performance data shows how physicians respond to the AI output. But the benefit question becomes more complex.

Do outcomes improve because the AI catches cases physicians would miss? Do outcomes improve because physicians work faster? Do outcomes worsen because physicians over-rely on the AI and miss cases where it fails?

The literature on diagnostic AI is growing but often lacks long-term outcome data. Most published studies focus on diagnostic accuracy metrics. Fewer track what happened to the patients after the AI provided its interpretation.

For AI diagnostics, the PMCF plan becomes critical. You need prospective data on clinical integration: how the AI is actually used, what actions physicians take based on its output, and what outcomes result. This data often reveals surprises about how clinical practice adapts to the technology.

What This Means for Your Next Submission

Before you write the clinical benefit section of your next diagnostic device clinical evaluation, map the evidence chain on paper.

Start with your device’s output. What measurement or information does it provide?

Next, document the clinical decision. What do physicians do with this information? Is this documented in guidelines, standard practice, or your own data?

Then, trace to outcomes. What happens to patients when physicians act on this information? Where is this documented?

Identify the gaps in this chain. Each gap is a deficiency waiting to happen unless you address it explicitly and commit to closing it through PMCF.

Your clinical benefit claim should match your evidence level. If you have complete outcome data, claim direct benefit. If you have decision-making data but limited outcome data, claim that the device provides information that supports clinical decision-making, and commit to outcome validation through PMCF.

Precision in claims prevents deficiencies and builds reviewer confidence.

Final Thought

The difference between analytical performance and clinical benefit is not semantic. It reflects the core purpose of medical device regulation: ensuring that devices improve patient outcomes, not just that they work as measurement tools.

When you understand this distinction and structure your evidence to address it, your clinical evaluations become defensible. When you ignore it, you face deficiency letters that require substantial additional work and delay your market access.

The evidence requirements are demanding, particularly for novel diagnostics. But they exist for a reason. Diagnostics that don’t improve outcomes aren’t beneficial, no matter how accurately they measure.

Your next submission will show whether you understand this difference.

Peace,
Hatem
Clinical Evaluation Expert for Medical Devices
Follow me for more insights and practical advice.