FDA Pivotal Clinical Study Design Guidance

01 What the document says

This guidance describes the principles FDA expects sponsors to apply when designing the pivotal study, the definitive investigation whose results are meant to support a determination of safety and effectiveness or substantial equivalence. It distinguishes the pivotal stage from earlier exploratory work (feasibility and pilot studies) and from post-market studies, and it is aimed at sponsors, investigators, IRBs, and FDA reviewers alike. It is not a statistics textbook; it sets out what a well-designed study must address.

For therapeutic devices, the document covers randomization, controls, blinding, choice of endpoints, and the sources of bias that undermine each. For diagnostic devices it has a separate discussion that is directly relevant to IVDs: the need to define the intended-use population and clinical context, the choice of a reference standard or comparator, how to avoid verification bias and spectrum bias, how to handle specimens of unknown status, and the role of sensitivity, specificity, predictive values, and agreement measures as endpoints.

Throughout, the guidance stresses pre-specification. The study hypothesis, success criteria, primary analysis, handling of missing or indeterminate results, and sample size justification should all be fixed in the protocol and statistical analysis plan before data are collected, with any later changes documented and explained. It also encourages sponsors to seek FDA feedback on the design through the Pre-Submission program.

02 What it means when you plan a study

• Define the intended-use population and the clinical question first; the inclusion criteria, site types, and comparator all follow from that definition and cannot be retrofitted.
• Avoid verification bias by applying the reference method to every enrolled specimen, not only to those the candidate test calls positive, which affects specimen volume and shipping logistics at every site.
• Pre-specify success criteria and the primary analysis; the sample size must be justified against those criteria, and a study that is merely 'large' without a stated hypothesis is harder to defend.
• Exploratory and feasibility studies are valuable for estimating prevalence and failure rates, but their data generally do not count toward the pivotal result, so budget them separately.
• Multi-site designs reduce single-site bias but add variability; the protocol should state how site effects will be examined.

03 Pathways it applies to

• 510(k) Clinical Studies
• CLIA Waiver Studies
• Dual Submission Studies

Source document: Design Considerations for Pivotal Clinical Investigations for Medical Devices (fda.gov).