Why neurological endpoints fail regulatory review

Written by HATEM RABEH, MD, MSc Ing

Your Clinical Evaluation Expert And Partner

in
S

This happens more often than it should. Teams spend months gathering data, only to discover that what they measured does not answer the clinical question regulators need answered.

For neurological devices, endpoint selection is not just a statistical exercise. It determines whether your clinical evidence can demonstrate safety and performance under MDR Article 61. And in neurology, that choice is harder than it looks.

The regulatory expectation for neurological devices

MDR Article 61 requires sufficient clinical evidence to demonstrate conformity with safety and performance requirements. For neurological devices, this means showing that the device produces a clinically relevant effect on neurological function, symptoms, or outcomes.

But neurology is not cardiology. You cannot measure a neurological benefit the way you measure ejection fraction or blood pressure reduction. Neurological disorders are complex, multifactorial, and often subjective in presentation.

MDCG 2020-6 on sufficient clinical evidence reinforces that clinical data must address the intended purpose and the claims made. If your device claims to improve motor function in Parkinson’s disease, your endpoints must reflect motor function in a way that is clinically interpretable and recognized in the neurological community.

This is where many files break down. The team selects an endpoint because it is measurable, not because it is meaningful.

What makes a neurological endpoint valid?

A valid neurological endpoint must meet three criteria. It must be clinically relevant, methodologically sound, and interpretable within the context of the neurological condition.

Clinically relevant means it reflects a change that matters to the patient or the treating physician. A two-point shift on a 100-point scale may be statistically significant but clinically irrelevant if it does not translate to functional improvement.

Methodologically sound means the measurement tool is validated, reliable, and appropriate for the population. Using a cognitive assessment designed for Alzheimer’s patients to evaluate traumatic brain injury outcomes will raise questions, even if the tool is validated.

Interpretable means the endpoint can be explained in clinical terms. Reviewers need to understand what a change in the endpoint represents in the real clinical setting. If you cannot explain why a 15% improvement in your chosen metric matters, the endpoint is not interpretable.

I have seen teams create custom scoring systems because existing scales did not fit their device perfectly. The result is always the same: the Notified Body asks for validation studies, the timeline extends, and the file stalls.

The difference between surrogate and clinical endpoints

Surrogate endpoints are appealing because they are easier to measure and faster to demonstrate. But in neurology, surrogates are rarely sufficient on their own.

A surrogate endpoint is a marker that is expected to predict clinical benefit but is not itself a measure of clinical benefit. Examples include imaging findings, biomarker levels, or electrophysiological signals.

A clinical endpoint is a direct measure of how the patient feels, functions, or survives. Examples include symptom scales, functional assessments, and quality of life measures.

MDR does not prohibit the use of surrogate endpoints. But it requires that the relationship between the surrogate and the clinical outcome be well established. In neurology, that relationship is often weak or unproven.

Consider a device that modulates brain activity to reduce epileptic seizures. Measuring changes in EEG patterns is a surrogate. Measuring seizure frequency is a clinical endpoint. The EEG change may correlate with seizure reduction, but if that correlation is not validated in the literature, the EEG data alone will not satisfy the clinical evidence requirement.

So what do you do if you have strong surrogate data but limited clinical outcome data?

You justify the surrogate with literature. You show that other studies have validated the relationship. You explain why the surrogate is accepted in clinical practice. And you supplement it with whatever clinical outcome data you can gather, even if limited.

Endpoint selection for different neurological conditions

The choice of endpoint depends heavily on the neurological condition being addressed. What works for stroke rehabilitation does not work for chronic pain. What works for movement disorders does not work for cognitive impairment.

For movement disorders like Parkinson’s disease, the Unified Parkinson’s Disease Rating Scale (UPDRS) is the standard. It is validated, widely used, and interpretable. If your device targets motor symptoms, you use the motor subsection. If you target overall disease burden, you use the total score.

For stroke rehabilitation, functional independence measures like the Barthel Index or the modified Rankin Scale are expected. These reflect the patient’s ability to perform activities of daily living, which is the primary concern in stroke recovery.

For pain management, especially neuropathic pain, visual analog scales (VAS) or numerical rating scales (NRS) are common. But pain is subjective and influenced by many factors. Reviewers will expect you to account for confounders and show consistency over time.

For cognitive impairment, tools like the Montreal Cognitive Assessment (MoCA) or the Mini-Mental State Examination (MMSE) are recognized. But cognitive endpoints are challenging because small changes may not be clinically meaningful, and ceiling effects can obscure real improvements in high-functioning patients.

What happens if no validated scale exists for your specific indication?

You have three options. You can adapt an existing scale and justify the adaptation. You can develop a new scale and validate it. Or you can use a combination of measures to capture the clinical effect from multiple angles.

The third option is often the most practical. No single scale captures every dimension of a neurological condition. Combining a functional measure with a symptom scale and a quality of life assessment gives a more complete picture.

The timing and frequency of endpoint measurement

When you measure matters as much as what you measure. Neurological conditions evolve over time. Acute effects are different from chronic effects. Immediate responses are different from sustained improvements.

If your device is used for acute intervention, such as clot retrieval in stroke, endpoints must be measured at discharge and at defined follow-up intervals. Immediate technical success does not guarantee functional recovery.

If your device is used for chronic management, such as deep brain stimulation for Parkinson’s disease, endpoints must be measured over months or years. Short-term improvements may not persist. Long-term data is required to demonstrate durability of effect.

PMCF planning under MDR Article 61(11) and Annex XIV Part B requires ongoing collection of clinical data. For neurological devices, this means continuing to measure endpoints in real-world use, not just in controlled trials.

I have reviewed PMCF plans where the team proposed measuring endpoints at one year post-implant and then never again. That does not work for a chronic neurological device. The plan must include periodic reassessment over the device’s lifetime in the patient.

Frequency of measurement also matters. Measuring too frequently can burden patients and introduce variability. Measuring too infrequently can miss important changes. The schedule must align with the natural progression of the condition and the expected timing of the device effect.

Handling subjectivity and variability in neurological endpoints

Neurology is inherently subjective. Many neurological symptoms cannot be measured objectively. Pain, fatigue, mood, and cognitive function rely on patient-reported outcomes or clinician assessments.

Subjectivity does not invalidate an endpoint. But it requires careful handling. You must account for inter-rater variability, placebo effects, and confounding factors.

Inter-rater variability means different assessors may score the same patient differently. This is especially true for clinical scales that involve judgment, such as the UPDRS or cognitive assessments. Training assessors and using standardized protocols reduces variability, but it does not eliminate it.

Placebo effects are strong in neurology. Patients with chronic pain, movement disorders, or mood symptoms often improve with any intervention, including sham procedures. If your clinical data does not account for placebo, reviewers will question whether the observed effect is real.

Confounding factors include medication changes, disease progression, rehabilitation, and psychosocial support. Neurological conditions do not occur in isolation. Showing that your device contributed to the observed improvement requires isolating its effect from other influences.

Aligning endpoints with the intended purpose and claims

Every endpoint must map directly to a claim in your labeling. If your IFU states that the device improves motor function, your endpoints must measure motor function. If it states that the device reduces seizure frequency, your endpoints must measure seizure frequency.

This seems obvious, but misalignment happens often. The clinical team selects endpoints based on what is feasible to measure, and the regulatory team writes claims based on what sounds marketable. The result is a gap that becomes visible during review.

Notified Bodies will compare your claims to your clinical data line by line. If a claim is not supported by an endpoint, it is not supported by evidence. If an endpoint was measured but not claimed, reviewers will ask why.

The intended purpose defines the scope of your clinical evaluation. For a neurological device, the intended purpose must specify the condition, the population, the expected clinical benefit, and the mechanism of action. Your endpoints must cover all of these elements.

For example, a device intended to improve gait in Parkinson’s patients must measure gait. A device intended to reduce chronic pain must measure pain intensity and interference with daily activities. A device intended to improve cognitive function must measure cognition using validated tools.

If your intended purpose is broad, your endpoint strategy must be broad. If it is narrow, your endpoints can be focused. The key is consistency.

What happens when the endpoint does not show a benefit

Not all clinical studies show positive results. Sometimes the endpoint does not demonstrate the expected benefit. This does not automatically mean the device is unsafe or ineffective, but it does require explanation.

If the endpoint did not change, you must ask why. Was the endpoint inappropriate? Was the study underpowered? Was the patient population too heterogeneous? Was the intervention period too short?

The clinical evaluation must address negative or neutral findings transparently. MDR Article 61 requires appraisal of all relevant clinical data, including unfavorable results. Hiding negative data is not an option.

If the endpoint failed because it was not sensitive enough, the evaluation must explain what a more appropriate endpoint would have been and whether such data exists from other sources.

If the endpoint failed because the study was flawed, the evaluation must critique the study limitations and explain why the result does not reflect the true safety and performance of the device.

If the endpoint failed because the device does not work as intended, the clinical evaluation must recommend design changes, additional studies, or a narrower intended purpose.

I have seen manufacturers try to reinterpret negative results as positive by focusing on subgroup analyses or secondary endpoints. This rarely works. Reviewers see through it. The better approach is to acknowledge the limitation, explain it, and provide a path forward.

Building the justification into the clinical evaluation report

The clinical evaluation report must include a clear rationale for every endpoint selected. This is not optional. It is part of the appraisal process required by MDR Annex XIV Part A.

The rationale should address:

• Why the endpoint is clinically relevant for the intended purpose
• Why the endpoint is validated and recognized in the neurological community
• How the endpoint aligns with the claims and labeling
• What literature supports the use of this endpoint in similar devices or conditions
• What limitations the endpoint has and how they were mitigated

If multiple endpoints were used, explain how they complement each other. If surrogate endpoints were used, explain their relationship to clinical outcomes. If novel endpoints were introduced, provide validation data.

This rationale must be written before the study starts, not after the data is collected. Justifying endpoints retrospectively is unconvincing. Reviewers can tell when the justification was constructed to fit the available data rather than to guide the study design.

The clinical evaluation must also discuss whether the endpoints were measured consistently across studies. If different studies used different endpoints, the evaluation must explain why and how the results can be integrated.

Final thoughts on endpoint selection in neurology

Endpoint selection is not a technical detail. It is the foundation of your clinical evidence strategy. For neurological devices, where the clinical effect is often subtle, subjective, and multifactorial, the choice of endpoint determines whether your clinical evaluation succeeds or stalls.

Choose endpoints that are clinically meaningful, methodologically sound, and aligned with your intended purpose. Justify your choices with literature and clinical reasoning. Measure consistently and interpret transparently.

And remember that endpoint selection is not a one-time decision. As your device evolves, as real-world data accumulates, and as the clinical understanding of the condition deepens, your endpoints may need to evolve as well.

PMCF is where this evolution happens. Use it to refine your endpoint strategy, validate your initial assumptions, and demonstrate that the clinical benefit observed in controlled settings persists in routine use.

Neurological devices face unique challenges, but they also offer unique opportunities to improve patient outcomes. The key is to measure what matters, not just what is easy.

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