Why your clinical endpoints may not support your claims

Written by HATEM RABEH, MD, MSc Ing

Your Clinical Evaluation Expert And Partner

in
S

This happens more often than you think. Manufacturers gather impressive-looking data. Strong p-values. Large sample sizes. Well-conducted studies. Then they face rejection because the evidence does not actually address what the device is supposed to do.

The disconnect sits between what was measured and what needs to be demonstrated. This is the surrogate endpoint problem.

What MDR Actually Requires

MDR Article 61 and Annex XIV make clear that clinical evaluation must demonstrate safety and performance for the intended purpose. Not for some related outcome. Not for a proxy measure. For the actual claim.

MDCG 2020-6 on sufficient clinical evidence reinforces this. The evidence must address the benefit-risk profile relevant to the device’s intended use and target population. If your device claims to reduce complications, you need data on complications. If it claims to improve quality of life, you need quality of life measures.

This sounds obvious when stated directly. Yet I routinely see clinical evaluations built entirely on surrogate endpoints that never connect back to the actual clinical benefit claimed.

The Surrogate Endpoint Trap

Consider a cardiac monitoring device that claims to reduce hospital readmissions for heart failure patients. The manufacturer provides studies showing the device accurately detects fluid retention changes. Sensitivity 95%. Specificity 92%. Technically impressive.

But detection accuracy is a surrogate endpoint. It measures device function, not patient outcome. The relevant question for the claim is different: Does using this device actually reduce readmissions?

That requires evidence showing that patients monitored with the device had fewer hospital readmissions compared to patients managed with standard care. You need intervention studies, not just technical performance data.

The gap between these two is where many clinical evaluation reports fail.

Why Manufacturers Default to Surrogates

I understand the temptation. Surrogate endpoints are easier to measure. They require smaller sample sizes. They show results faster. They cost less to study.

Patient outcome studies are expensive and time-consuming. A study measuring actual reduction in myocardial infarction requires hundreds of patients and years of follow-up. A study measuring improvement in lipid profiles needs fifty patients and three months.

The regulatory pressure to get devices to market creates strong incentive to use the faster, cheaper option.

But if the surrogate does not reliably predict the patient outcome, the evidence is not sufficient for MDR purposes.

When Surrogates Can Work

Not all surrogate endpoints are inadequate. Some have strong established relationships with patient outcomes. But this relationship must be demonstrated, not assumed.

Take blood pressure measurement devices. Reduction in blood pressure is technically a surrogate endpoint. The real patient outcomes are stroke, myocardial infarction, heart failure, death. But decades of evidence have established that blood pressure control reduces these outcomes. The link is validated.

So for a blood pressure monitor, demonstrating measurement accuracy is clinically relevant. The surrogate is accepted because the connection to patient benefit is proven by extensive literature.

Here is the key question: Is the link between your endpoint and the patient outcome established in the medical literature?

If yes, you can reference that literature to bridge from your surrogate data to the claimed benefit. If no, you need direct patient outcome data.

The Bridging Requirement

When you use surrogate endpoints, your clinical evaluation report must explicitly make this connection. You must show that changes in the surrogate endpoint translate to changes in patient outcomes for your specific application.

This requires systematic literature review. Not just citation of a few papers. A structured search and appraisal demonstrating that the endpoint you measured is a validated predictor of the outcome you claim.

I have seen reports that measure biomarkers with no evidence that changing those biomarkers improves patient health. The assumption is made but never supported. That does not satisfy MDR requirements.

Patient-Relevant Outcomes

What makes an endpoint patient-relevant? It directly reflects how the patient feels, functions, or survives.

Mortality is patient-relevant. Symptom relief is patient-relevant. Return to work is patient-relevant. Quality of life scores are patient-relevant if properly validated.

Laboratory values are usually not patient-relevant unless the link to clinical benefit is established. Imaging findings are usually not patient-relevant unless they predict outcomes. Technical specifications are never patient-relevant on their own.

Your intended purpose statement should guide you. If the device claims to improve mobility, you need mobility measures. If it claims to reduce pain, you need pain measures. If it claims to extend survival, you need survival data.

The claim and the evidence must align directly.

The Regulatory Reviewer Perspective

When I review a clinical evaluation report, I map every claim in the intended purpose to the evidence provided. I ask: Where is the data that supports this specific statement?

If the device claims to reduce complications but only provides data on procedure time, there is a gap. Faster procedures might reduce complications. They might not. Without evidence showing the actual complication rate, the claim is not supported.

Notified Bodies approach this the same way. They will identify every claim element and look for corresponding evidence. If the evidence does not directly address the claim, they will issue a non-conformity.

You cannot reason your way out of this. You cannot argue that the surrogate logically implies the outcome. You must demonstrate it with data.

Composite Endpoints and Multiple Measures

Some devices affect multiple patient outcomes. A diabetes management system might reduce hypoglycemic events, improve HbA1c, and enhance quality of life. You have several relevant endpoints.

This is appropriate if all these outcomes are part of your intended purpose. But be careful with composite endpoints that mix patient-relevant outcomes with surrogate measures.

I have seen studies report a composite endpoint combining mortality, hospitalization, and biomarker changes. The first two are patient outcomes. The third is a surrogate. If the composite shows benefit but the patient outcomes alone do not, the evidence is weaker than it appears.

Always report patient-relevant outcomes separately. Do not bury them in composites where their individual contribution is unclear.

Prioritizing Outcomes

Some outcomes matter more than others. Mortality and major morbidity are primary concerns. Symptom improvement and quality of life are important but secondary. Convenience and efficiency are tertiary.

Your clinical evidence should address the primary outcomes first. If your device affects serious health outcomes, that is where your evidence must be strongest. You cannot compensate for weak safety or efficacy data with strong usability or satisfaction data.

The hierarchy of evidence importance must match the hierarchy of clinical importance.

PMCF and Endpoint Selection

Post-market clinical follow-up must continue measuring the endpoints that matter. If your pre-market evidence relied on surrogate endpoints with literature bridging, your PMCF should aim to gather direct patient outcome data.

This is where you confirm that the predicted relationship actually holds in real-world use. That the surrogate-outcome link validated in controlled studies applies to your device in routine clinical practice.

PMCF is not just about collecting more of the same data. It is about addressing the gaps and uncertainties that remained after pre-market evaluation. If you relied on surrogates, confirming patient outcomes is a key gap to address.

The PMCF plan should explicitly identify which outcomes are being measured and why. What questions are you answering? What uncertainties are you reducing? What claims are you confirming?

Practical Guidance for Choosing Endpoints

Start with your intended purpose statement. Write down every claim. For each claim, identify the patient outcome that directly reflects that claim.

Ask: What would a patient or clinician consider meaningful improvement or benefit? That is your target endpoint.

Then evaluate whether you have data on that endpoint. If not, assess whether a surrogate is acceptable. This requires literature review to validate the surrogate-outcome relationship.

If the literature does not support the surrogate for your specific context, you need to plan studies that measure the patient outcome directly. This might delay market entry, but it prevents later regulatory problems.

I know this is not always commercially convenient. But the MDR does not make exceptions for inconvenience.

Documenting the Rationale

Your clinical evaluation report must explain why the chosen endpoints are appropriate. This is not optional. Reviewers will question endpoint selection if it is not clearly justified.

If you use surrogate endpoints, document the validation evidence. Cite specific studies showing the relationship between the surrogate and patient outcomes. Explain why this relationship applies to your device and population.

If you use patient-relevant outcomes, explain how they reflect the claimed benefits. Show that the measures are validated and appropriate for your target population.

This documentation creates transparency. It shows you made deliberate, evidence-based choices rather than simply using whatever data was convenient.

When Evidence Gaps Exist

Sometimes you genuinely cannot obtain direct patient outcome data pre-market. The studies would take too long, cost too much, or require patient numbers you cannot access. This happens.

In these cases, you must acknowledge the gap. Explain why direct outcome data is not available. Justify the use of surrogate endpoints with whatever validation evidence exists. And commit to gathering patient outcome data post-market through PMCF.

This is not ideal, but it is honest. Reviewers can accept this if you are transparent about the limitation and have a clear plan to address it.

What they will not accept is pretending the surrogate data fully demonstrates patient benefit when it does not. That is where manufacturers get into serious trouble.

Final Considerations

The choice between surrogate endpoints and patient outcomes is not purely scientific. It is regulatory, clinical, and commercial. You must balance feasibility against evidence quality against regulatory acceptability.

But the MDR sets a clear standard. The evidence must demonstrate the claimed benefits for the intended purpose. If your endpoints do not directly reflect those benefits, you need to bridge that gap with validated relationships or acknowledge the limitation.

There is no shortcut around this. Technical performance is necessary but not sufficient. Clinical benefit must be demonstrated, not inferred.

When you build your clinical evaluation strategy, make endpoint selection an early priority. Map claims to outcomes. Identify which data you have and which gaps exist. Make deliberate choices about surrogates based on validation evidence, not just convenience.

This approach will strengthen your clinical evaluation and reduce the risk of late-stage regulatory issues. It aligns your evidence with what regulators actually need to see.

And it ensures that your clinical claims are grounded in evidence that matters to patients, not just evidence that is easy to collect.

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