Clinical evidence planning starts with the intended use

Before a cardiovascular device team chooses an endpoint or budgets a clinical study, it should agree on the claim the evidence is meant to support. A study can be well executed and still leave the product’s central claim unanswered. The planning question is whether the proposed observations would justify what the team intends to say about the device, for the people and conditions in which it will be used.

Start with the intended use and proposed indications, then make the evidentiary consequences explicit. This is especially important when a development program combines hardware performance, software output and a clinical workflow: each can work as designed without establishing the full clinical claim.

Define the claim before selecting the study

FDA’s final guidance on pivotal investigations recommends that study objectives support the intended use and desired labeling claims, and that enrolled subjects reflect the target population. Its principal focus is PMA-supporting studies, although relevant principles can inform clinical studies for 510(k) and De Novo submissions. [1, sections 1, 6.3 and 6.4] These are guidance recommendations; the planning method below is Zaliance’s professional interpretation.

Write a concise proposed-use statement that identifies the clinical purpose, target condition, patient population, user and setting. Add the boundaries that could change the evidence question: anatomical location, severity, duration of use, required accompanying care and the role of the device output in a treatment decision.

For internal planning, ask each function to translate that statement into its own decision. Engineering should identify the performance the design must deliver. Clinical should identify the patient-level question. Regulatory should identify the proposed submission claim. If these translations describe different products, resolve the difference before committing to a protocol.

Distinguish performing a task from delivering the claimed benefit

Consider a hypothetical peripheral vascular treatment device. The team has bench evidence that its delivery system can reach and operate at a specified lesion. It now wants to describe a durable treatment result in a defined patient population.

Successful delivery and immediate technical performance are relevant, but they do not answer whether the treated vessel remains patent over the claimed period or whether patients need another intervention. Those questions concern different observations. Make the gap visible before treating one favorable result as evidence for all three claims.

Proposed claim element	Question to resolve	Planning consequence
Delivery and operation at the target lesion	Can the device perform the specified procedure under relevant conditions?	Identify which bench, simulated-use and clinical observations address that task.
A sustained vessel result	What precisely counts as patency, at what time, and who assesses it?	Define the assessment method and follow-up schedule before selecting the endpoint.
Fewer repeat interventions	Compared with what, and under which rules for deciding to intervene?	Specify the comparison and how treatment decisions enter the outcome assessment.

This is a claim-mapping example, not a recommended study or a classification determination. The team might decide to narrow a proposed claim, commission additional evidence or change the development sequence. Those are substantive program choices; a protocol cannot make them disappear.

Build the population around the claim’s boundaries

In the hypothetical vascular program, “adults with vascular disease” would not adequately describe the population needed for planning. The team would need to decide which vessel territories, lesion characteristics, prior treatments and clinical presentations belong within the proposed use. It should explain how enrollment criteria correspond to those decisions.

Excluding a difficult group may be reasonable for a particular investigation. Record what that exclusion means for the eventual evidence claim. If the commercial plan still includes that group, identify how the gap will be resolved rather than assuming that a broad population label supplies the missing support.

Operational choices also deserve scrutiny. A study run only by highly experienced operators can help characterize performance under that condition. Before extending the conclusion to a wider user base, ask whether the intended training, workflow and use environment have been represented sufficiently for the claim.

Specify what the endpoint means clinically

FDA recommends clinically meaningful endpoints tied to study objectives and intended use, with measurement and analysis specified in the protocol. [1, section 7.1] Apply that principle by making the endpoint understandable without relying on its short name.

For the hypothetical sustained vessel result, specify the observation time, the assessment method and the treatment of events that prevent the planned assessment. A reintervention before the follow-up image is not simply an inconvenient missing image; decide prospectively how it relates to the claim being tested. Address deaths, withdrawals and unsuccessful procedures in the analysis plan rather than leaving them for interpretation after results are known.

For a repeat-intervention claim, examine how the decision to intervene will be made. If clinicians in one group are assessed more often or use a different intervention threshold, the comparison may incorporate differences in management as well as device performance. The study team should determine which controls and adjudication procedures are appropriate.

Safety deserves its own evidence questions. Name the plausible harms, their observation windows and the methods for identifying them. A favorable technical result should not be allowed to stand in for an assessment of procedural or later device-related harm.

For diagnostic claims, define the comparison correctly

A separate hypothetical diagnostic example illustrates a different problem. Suppose software flags a suspected cardiovascular condition for clinician review. Agreement with another software system would answer a comparison question; it would not, by itself, establish which system correctly identifies the condition.

FDA’s diagnostic reporting guidance distinguishes reference-standard accuracy estimates from agreement with a non-reference comparator. It recommends reporting the comparison and its limitations appropriately. [2, sections 3–4 and 6] For this example, define how the target condition will be established independently of the software under evaluation.

Also keep the claimed role of the output visible. Flagging a case for review, supplying a diagnosis and improving a patient outcome are different propositions. Choose evidence that addresses the proposition actually being made. Do not turn a classification-performance result into an outcome-benefit claim through a change in marketing language.

Decide what evidence is needed, then size the work

Clinical evidence planning does not automatically mean commissioning a new trial. FDA’s 510(k) guidance describes clinical-data requests for questions that nonclinical methods cannot adequately address because of the indications or technology. [3, section IV.F] Assess the particular pathway, existing evidence and device-specific expectations before deciding what new work is necessary.

For the hypothetical program, prepare a short evidence map for each claim: available evidence, its relevance to the proposed configuration and population, the unresolved question, and the work proposed to answer it. An existing dataset might resolve one uncertainty while leaving a different population or follow-up interval uncovered.

Only then develop the sample-size rationale. Specify the endpoint, comparison, effect or precision sought, expected event frequency where relevant, and allowance for incomplete follow-up. There is no defensible universal patient count for the varied claims in this example. A statistician can size a defined question; a larger study cannot rescue an undefined one.

Keep the evidence plan aligned as the product changes

Treat the proposed-use statement and evidence map as controlled planning inputs. When a design, user group or claim changes, identify which assumptions remain valid and which need reconsideration. Document the clinical and regulatory rationale before deciding that existing evidence transfers to the revised product.

The useful output is a program decision: proceed with a defined evidence plan, narrow the claim, resolve a specific uncertainty with FDA, or generate additional preliminary information. By the time a protocol is authorized, the team should be able to explain exactly which claim its primary result will support—and which claims will still remain unproven.

References

1. FDA. Design Considerations for Pivotal Clinical Investigations for Medical Devices. Final guidance, November 7, 2013; sections 1, 6.3, 6.4 and 7.1.
2. FDA. Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests. Final guidance, March 2007; sections 3–4 and 6. Its scope concerns qualitative diagnostic tests; it is not a complete study-design manual.
3. FDA. The 510(k) Program: Evaluating Substantial Equivalence in Premarket Notifications [510(k)]. Final guidance, July 28, 2014; section IV.F.