Why Your Primary Endpoint Isn’t Actually Primary

Written by HATEM RABEH, MD, MSc Ing

Your Clinical Evaluation Expert And Partner

in
S

The distinction between primary and secondary endpoints is not a formatting preference. It is a fundamental design choice that shapes your entire clinical investigation. It determines your sample size, your statistical plan, your interpretation strategy, and ultimately, whether your clinical evidence will stand up under MDR scrutiny.

Yet in practice, I see this distinction mishandled more often than it should be. Endpoints are poorly defined. Hierarchies are unclear. Statistical adjustments are missing. And when the investigation concludes, the clinical evaluation team inherits data they cannot interpret with confidence.

What MDR Actually Requires

The MDR does not explicitly define primary and secondary endpoints in technical detail, but Article 62 and Annex XV make clear that clinical investigations must be designed to provide valid scientific evidence. MDCG 2020-6 on sufficient clinical evidence reinforces that investigations must be methodologically sound, with clear objectives, appropriate statistical power, and predefined success criteria.

This is where endpoint selection becomes regulatory, not just scientific. Your primary endpoint is the foundation of that validity. It must answer the core clinical question your device is intended to address. It must be clinically meaningful, measurable, and defensible as the criterion for claiming performance or safety.

If your primary endpoint does not directly support your intended purpose, your investigation may generate data, but it will not generate the evidence you need for conformity assessment.

Primary Endpoint: The Single Question That Matters

A primary endpoint is the main outcome measure used to determine whether the investigation meets its objective. It should be singular. Not multiple. Not a collection of interesting observations. One clearly defined measure that represents clinical success or failure.

This is the endpoint that drives your sample size calculation. It determines the statistical threshold you must meet. It defines what success looks like before you enroll the first patient.

In practice, I see sponsors struggle with this singularity. They want to measure everything. Device performance, safety, usability, patient satisfaction, clinical improvement, all at once. But calling everything primary dilutes the focus and introduces statistical complexity that most investigations cannot handle.

Your primary endpoint must also be clinically relevant. Surrogate markers can be acceptable, but only if they have established correlation with meaningful clinical outcomes. A biochemical marker with no validated link to patient benefit will not convince a Notified Body that your device is safe and performs as intended.

Finally, the primary endpoint must be measurable with precision and reproducibility. Vague terms like “clinical improvement” or “satisfactory performance” do not work. You need a defined scale, a clear timepoint, and a method that trained investigators can apply consistently across sites.

Secondary Endpoints: Supporting Evidence, Not Backup Plans

Secondary endpoints explore additional questions. They provide supporting evidence. They capture data that enriches understanding of the device but is not the primary basis for claiming success.

The distinction is important because secondary endpoints do not drive the core hypothesis. They do not determine sample size. They are analyzed after the primary endpoint, and their interpretation is contextual.

But secondary endpoints are not throwaway data. They matter for building a complete clinical picture. They support claims in your IFU. They feed into your clinical evaluation report and your risk-benefit analysis. They inform post-market surveillance planning.

What makes a good secondary endpoint? It should address a clinically relevant question that extends beyond the primary objective. Examples include durability of effect, quality of life measures, subgroup analyses, or safety outcomes that complement the primary safety endpoint.

The mistake I often see is treating secondary endpoints as backups. If the primary endpoint fails, sponsors look to secondary endpoints to salvage the investigation. This is methodologically unsound. Secondary endpoints cannot retroactively become primary. If your investigation fails on its primary endpoint, it fails.

Statistical Implications You Cannot Ignore

The primary endpoint determines your statistical plan. Sample size calculations are based on the expected effect size, variability, and desired power for that single endpoint. If you dilute this by trying to power the study for multiple primary endpoints, you either need a much larger sample, or you need to apply multiplicity adjustments.

Multiplicity adjustments are not trivial. If you test multiple primary endpoints without adjustment, you inflate your Type I error rate. In plain terms, you increase the chance of declaring success when none exists. Regulators know this. Notified Bodies will scrutinize your statistical section for multiplicity handling.

Secondary endpoints also introduce multiplicity, but the consequences are different. Since they are not the basis for claiming success, the statistical standards are less stringent. However, if you plan to use secondary endpoint data to support labeling claims or clinical performance, you still need a predefined analysis plan and appropriate adjustment methods.

In practice, this means your CIP must state clearly: which endpoint is primary, how it will be analyzed, what threshold defines success, and how secondary endpoints will be handled statistically. Vagueness here translates directly into deficiencies during ethical and regulatory review.

The Hierarchy Matters During Interpretation

When your investigation concludes, the results are interpreted in a hierarchy. Primary endpoint first. If it is not met, the investigation did not achieve its objective. This does not mean the data is useless, but it does mean you cannot claim the device meets its intended purpose based on that investigation alone.

Secondary endpoints are then interpreted in context. Positive secondary outcomes can provide supportive evidence, but they do not override a negative primary result. Conversely, if the primary endpoint is positive but secondary endpoints raise safety concerns, the risk-benefit assessment becomes more complex.

This hierarchy must be understood before the investigation starts, not after. I have reviewed clinical evaluation reports where sponsors tried to reframe secondary outcomes as primary because the true primary endpoint was inconclusive. This does not work. The protocol is a binding document. You cannot change the rules after seeing the data.

Common Pitfalls in Endpoint Selection

One common mistake is selecting a primary endpoint that is easy to measure but clinically irrelevant. Device deployment time, for example, might be measurable and reproducible, but if it does not correlate with clinical benefit, it will not support your conformity claim.

Another pitfall is choosing a composite primary endpoint without justification. Composite endpoints combine multiple outcomes into a single measure. This can be appropriate when the components are clinically related and of similar importance, but it often obscures interpretation. If your composite includes both major adverse events and minor inconveniences, what does success actually mean?

A third issue is failing to predefine the analysis population. Will you analyze intent-to-treat, per-protocol, or both? For the primary endpoint, this must be specified in advance. Post-hoc decisions about which population to analyze invite bias and weaken the integrity of the results.

Finally, I see protocols where the timepoint for primary endpoint assessment is poorly justified. Why six months and not three? Why hospital discharge and not thirty-day follow-up? If the timepoint is arbitrary, reviewers will question whether it was chosen to optimize appearance rather than clinical validity.

How This Feeds Into Clinical Evaluation

Your clinical evaluation report must appraise the clinical investigation data. That appraisal depends on understanding what the investigation was designed to show. If the primary endpoint was met, you have direct evidence supporting a specific claim. If it was not met, you must explain the implications and whether other data compensates.

Secondary endpoints can enrich your clinical evaluation, but they do not carry the same evidentiary weight. A positive secondary outcome on quality of life, for example, supports a broader benefit profile, but it does not substitute for a failed primary endpoint on device performance.

This distinction becomes critical during Notified Body review. Auditors will cross-reference your CIP, your clinical investigation report, and your CER to check consistency. If your CER claims clinical benefit based on secondary endpoints while the primary endpoint was inconclusive, expect challenges.

Practical Guidance for Writing a CIP

When drafting the objectives and endpoints section of your CIP, start by stating the primary objective in one sentence. Then state the primary endpoint in one sentence. If you cannot do this clearly, your design is not ready.

For secondary endpoints, list them separately and provide a brief rationale for each. Why is this outcome important? How will it be measured? What will the data contribute to the overall assessment?

In your statistical section, specify the sample size calculation based solely on the primary endpoint. Describe how secondary endpoints will be analyzed, and if you plan to use them for labeling or claims, explain the adjustment for multiplicity.

In your analysis plan, define the hierarchy explicitly. State that the primary endpoint will be analyzed first, and that secondary endpoints will be interpreted in the context of the primary result. This protects you from later accusations of data dredging.

Finally, ensure that your success criteria are predefined and unambiguous. “Clinical improvement” is not a criterion. “Reduction of at least 20% in symptom score at six months compared to baseline, with p<0.05" is.

What Happens When You Get It Wrong

If your endpoint selection is weak, the consequences are not immediate but they are inevitable. Ethics Committees will request clarifications or reject the protocol. Notified Bodies will issue deficiencies during design review. Clinical evaluators will struggle to interpret your data. And in the worst case, your investigation will generate data that cannot support conformity.

I have seen investigations run to completion, only to discover that the primary endpoint was too insensitive to detect a meaningful effect, or that the timepoint was too early to capture the relevant outcome. The data exists, but it does not answer the question that regulators need answered.

This is expensive, both in time and resources. More importantly, it delays patient access to potentially beneficial devices and erodes trust with investigators, Ethics Committees, and Notified Bodies.

Final Thought

The distinction between primary and secondary endpoints is not bureaucratic. It is the backbone of scientific rigor in clinical investigation. It forces you to define what success looks like before you start. It ensures your investigation is adequately powered. It protects the integrity of your conclusions.

When I review a CIP with a single, well-justified primary endpoint and a logical set of secondary endpoints, I know the sponsor understands clinical investigation design. When I see a list of undifferentiated outcomes with no hierarchy, I know there will be problems downstream.

Choose your primary endpoint carefully. Define it precisely. Design your investigation around it. Everything else is secondary, and that is exactly as it should be.

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