This Technical Blog was written with input from Endpoint’s Regulatory and Medical Device expert Dr. Sam Gilbert, VMD.

Content has been informed by the presentation and discussion following the recent webinar “Regulatory Strategy for Medical Device Preclinical Programs”.

When I review a proposed preclinical study, one of the first things I consider is how the resulting data will eventually be interpreted by regulators.

I am not trying to predict exactly what an FDA reviewer will decide. I am trying to put myself in the reviewer’s position and ask a more useful question: If this protocol and its supporting documentation landed on my desk, what would I look for?

The reviewer is not only asking whether the study was well executed. They need to understand what the sponsor is trying to establish, whether the study was designed appropriately for that purpose, and whether the evidence supports the conclusion they are being asked to accept.

A technically successful study is not necessarily a successful regulatory study.

Before reviewing a protocol, I want to understand the role the study plays within the broader evidence-generation strategy. A feasibility study, a pilot study, and a definitive GLP study should not be designed – or judged – the same way.

I often encourage teams to start with the end in mind: the regulatory pathway, the milestones leading to approval, the evidence ultimately required, and the specific role each study will play in generating it. Only then can the protocol be evaluated against the conclusion it is supposed to support.

This is where broad objectives can become a problem. A protocol may state that the study will assess safety, performance, usability, biocompatibility, or some combination of all four. But what will the sponsor actually be able to conclude when the study is complete?

Collecting an endpoint does not automatically mean the study can support a conclusion about it.

A study may include blood work, organ collection, histology, functional assessments, and observations related to device use. That does not necessarily mean the sample size, model, study duration, or overall design is sufficient to satisfy every regulatory requirement associated with those data.

Sponsors understandably want to do more with less. A major part of preclinical strategy is finding ways to reduce the number of studies, animals, time points, and unnecessary costs. But trying to answer every question in a single model may ultimately create more work if the study does not adequately support several of its intended conclusions.

Sometimes the more efficient strategy is to separate questions and answer each in the model best suited to it.

I rarely ask whether a model is simply “clinically relevant.” I ask whether it is appropriate for the evidence the study must generate.

What outputs are required? What endpoints and success criteria will matter? Can the finished device be used in the model? Can the procedure reasonably simulate the intended clinical technique? Does the anatomy support the necessary access, deployment, or interaction with the target tissue? Can the model demonstrate the biological response the sponsor will later need to describe?

In some cases, one model can address several important questions. In others, the best model for safety is not the best model for effectiveness or usability. The decision should follow the intended conclusion rather than the desire to place every endpoint into one convenient study.

The justification also matters. A regulatory submission needs to explain clearly why the model was chosen, what it represents, what it does not represent, and why those limitations do not prevent the study from supporting its intended purpose.

Saying that a model is commonly used is rarely enough. The connection between the model and the claim must be understandable to someone who was not involved in the development process.

The protocol sections I spend the most time refining are the endpoints, success criteria, and interpretation plan, and rarely the operational ones. Most protocols describe what will be measured. Far fewer define, in advance, what constitutes success.

That becomes especially important after the data are available. If the protocol does not explain prospectively what an acceptable result looks like, the sponsor may later appear to be interpreting the findings post hoc around the outcome they received.

This does not mean every endpoint needs a simple pass-or-fail threshold. Many medical device studies, particularly large animal safety studies, involve sample sizes that will never support conventional statistical significance. Some findings require a scientific and weight-of-evidence interpretation rather than a numerical cutoff.

The protocol should still explain how the data will be assessed.

What is each endpoint intended to contribute? What findings would be expected? What would raise concern? How will variability be considered? How would an unexpected observation affect the study’s overall conclusion?

These questions are not easier to answer after the study. By then, the team is reacting to real findings, often under time and financial pressure.

Defining the interpretive framework prospectively makes the eventual narrative more credible and helps the study withstand questions about post hoc analysis. FDA has increasingly focused on success criteria for preclinical endpoints, making this an important part of protocol development rather than something to address only when the report is written.

A scientifically elegant design can still become a poor regulatory study if the procedure cannot be performed consistently, the site lacks the necessary proficiency, or the resulting documentation is incomplete or difficult to interpret.

GLP compliance does not compensate for an immature device, an unrefined procedure, or a model the team has not adequately rehearsed.

Before reaching a definitive study, I want the basic feasibility questions addressed. The device should be sufficiently stable. The procedure should be reproducible. The team should understand the anatomy and access. The CRO should be selected not only because it claims the necessary capabilities, but because it is a good fit for the product, procedure, sponsor, and level of oversight required.

A definitive study should not be the first time the team discovers that the access is more difficult than expected or that a key procedural step does not work reliably in the chosen model.

Cadaveric work, pilot studies, and non-GLP rehearsals can be extremely valuable here. They allow teams to refine the procedure, understand how the device interacts with tissue, train personnel, and resolve operational questions before entering the study they intend to rely on. The closer a program gets to definitive evidence generation, the fewer fundamental questions should remain.

Ultimately, the regulatory reviewer will see the protocol, the data, the deviations, the study report, and the rationale provided in the submission.

That record must stand on its own.

Very few preclinical studies produce flawless results. There may be complications, deviations, outliers, adverse observations, unexpected pathology findings, or results that require more interpretation than the team anticipated. Those findings do not automatically invalidate the study.

The first interpretation is often emotional. The second is usually scientific. I encourage clients not to confuse the two.

When unexpected data emerge, the first step is to slow down. What does the finding actually show? Is it isolated or consistent? Is it related to the device, the procedure, the model, animal management, or another factor? Does it change the conclusion? Does it require further analysis, a root-cause investigation, additional testing, or simply a clear scientific explanation?

A strong report does not hide an inconvenient finding or selectively present only the favourable story. It acknowledges the complete data set, distinguishes observation from interpretation, and explains the finding in a way that is scientifically supportable.

Consistency is equally important. The interpretation in the study report should align with the regulatory submission, internal engineering reports, future publications, and the broader product-development narrative. Contradictory explanations across documents create avoidable questions and can undermine confidence in otherwise useful data.

The preclinical section of a regulatory document is not an isolated collection of reports. It must fit coherently with the device description, intended use, design history, clinical plan, and other evidence being presented.

A good reviewer pass does not end with the question, “Would I accept this study?”

I also want to know what I would ask next.

Where is the design likely to attract scrutiny? Is the model justification vulnerable? Is the sample size adequate for every conclusion being claimed? Are the endpoints and success criteria clearly defined? Does the procedure reflect intended use? Are there limitations that should be acknowledged proactively? Is the sponsor relying on a scientific rationale that has not yet been documented?

It is not necessary to eliminate every potential question. In many programs, the least burdensome approach intentionally pushes beyond the most conservative testing plan. That can be entirely appropriate, provided the sponsor understands where risk has been introduced and has a defensible rationale for the decisions made.

My typical approach is to first understand what the most comprehensive testing program might look like. From there, the team can deliberately pare it back by leveraging endpoints, reducing sample sizes where justified, minimizing time points, or sequencing studies around regulatory milestones.

The important part is knowing what was removed, why it was removed, and what the plan will be if the streamlined approach proves insufficient. A structured preclinical risk review can help teams examine those assumptions before committing to the study plan.

That same reviewer mindset is valuable when responding to regulatory feedback. A long deficiency letter can feel like a demand to start over, but the wording matters. FDA may ask for new testing, but it may also allow an issue to be addressed through existing data, an amended report, or a scientific rationale. Each concern should be read carefully and answered directly rather than assuming every question requires another study.

I recommend conducting a deliberate reviewer pass at two points: before the protocol is finalized and before the study report is approved.

Before approving the protocol, I ask:

• Does this study support the conclusion we intend to claim?
• Is the model appropriate for that conclusion?
• Have we defined success before collecting the data?
• Where will the design attract scrutiny?
• What will FDA ask first?
• If the outcome is mixed, do we already know how we will interpret it?

Before the report is finalized, I ask whether the data have been presented completely, interpreted consistently, and integrated into the broader product story. Have unexpected findings been addressed? Are the conclusions limited to what the study can genuinely support? What questions is a reviewer most likely to ask next?

This exercise cannot guarantee that FDA will agree with every decision. That is not the goal.

The goal is to stop evaluating the study exclusively from the sponsor’s point of view.

A protocol should not merely describe an experiment that can be performed. It should describe a study capable of producing evidence that can be understood, evaluated, and used in a regulatory decision.

That is the difference between asking whether a study is well designed and asking whether it is designed to carry the conclusion placed on it.