Pathway | New Product Development

How Many Samples Are Needed for Medical Device Design Verification?

Representative samples, precision inspection, and controlled manufacturing support medical device design verification.

TLDR

TL;DR: TL;DR: There is no universal sample size for medical device design verification. The number of devices required depends on the design input being verified, the risk associated with failure, the type of data collected, the statistical confidence needed, and applicable device-specific standards. For certain pass/fail tests using a zero-failure binomial approach, 29, 59, or 299 independent samples may support different reliability claims at 95% confidence. However, these numbers are not universal FDA requirements. A defensible verification strategy must justify sample selection, represent the intended device configuration, and establish appropriate acceptance criteria before testing begins.

Understanding Medical Device Design Verification Sample Sizes

One of the most important questions medical device companies face when preparing for design verification is how many devices they need to manufacture and test. The answer can significantly influence development budgets, manufacturing schedules, material requirements, and the timing of regulatory submissions. For companies producing complex or expensive medical devices, manufacturing additional verification units may represent a substantial investment. At the same time, testing too few devices without an appropriate scientific justification can leave important questions about product performance unanswered.

The challenge is that medical device design verification does not follow a single sampling requirement applicable to every product. A catheter connection undergoing tensile testing, a mechanical component being evaluated for dimensional accuracy, and an electromechanical device undergoing functional testing may require different sample quantities and statistical approaches. Even within the same device, individual design requirements may need different verification methods. Selecting a sample size therefore requires understanding what is being verified, how the results will be evaluated, and what level of evidence is appropriate for the associated risk.

A successful verification strategy begins before the first test sample is manufactured. Engineering, quality, regulatory, and manufacturing teams must establish which design inputs require verification, identify suitable test methods, define acceptance criteria, and determine how the selected samples will represent the intended device. These decisions influence not only the statistical validity of the results but also the manufacturing activities needed to produce the verification units.

Part 1: What Does the FDA Require for Design Verification Sample Sizes?

Is There a Minimum FDA Sample Size?

The FDA does not establish one universal minimum sample size for all medical device design verification tests. Instead, sample-size expectations depend on the device, applicable regulatory requirements, test objectives, and any relevant FDA guidance or recognized consensus standards. A sample quantity appropriate for one verification activity may be insufficient or unnecessary for another.

The FDA’s guidance, Recommended Content and Format of Non-Clinical Bench Performance Testing Information in Premarket Submissions, recommends providing a scientific rationale for the number of samples tested. It explains that sample size should be based on the study objective and supported by risk analysis and a sampling plan. When testing follows a recognized consensus standard or device-specific FDA guidance, applicable recommendations for sample selection should also be considered.

This means that choosing a familiar number such as 30 or 50 samples does not automatically establish that a verification study is adequate. The manufacturer must be able to explain why the selected quantity, test configuration, and analysis method provide appropriate evidence that the relevant design requirement has been met.

For example, verifying the presence of a specified design feature may require a different approach from demonstrating that a mechanical connection meets a defined strength requirement across a population of manufactured devices. Both activities may contribute to design verification, but their objectives and evidentiary requirements are not identical.

How ISO 13485 and ISO 14971 Influence Verification Planning

ISO 13485:2016 establishes quality management system requirements relevant to medical device design and development, including design verification activities. Under the FDA’s Quality Management System Regulation (QMSR), which became effective February 2, 2026, ISO 13485:2016 is incorporated by reference into 21 CFR Part 820. Manufacturers must address applicable design and development requirements within this regulatory framework.

Risk management is another important consideration. ISO 14971:2019 provides a framework for identifying hazards, estimating and evaluating risks, and establishing appropriate risk controls throughout the device lifecycle. Although ISO 14971 does not prescribe a universal verification sample size, risk analysis can help determine the strength of evidence needed for individual device requirements.

A failure that could affect a critical safety function may justify a more stringent verification strategy than a failure associated with a noncritical characteristic. However, the relationship between risk and sample size should be established through a documented rationale rather than assuming that every high-risk requirement automatically requires the same number of samples.

Explore Our DV Manufacturing Support

Part 2: What Determines the Number of Verification Samples?

Risk Classification and the Consequences of Failure

The risk associated with a design requirement is an important consideration when establishing verification confidence. A device characteristic associated with a potentially serious patient or user harm may require stronger evidence of performance than a characteristic with limited safety implications. Risk analysis helps identify these differences and provides a foundation for selecting appropriate verification methods and acceptance criteria.

For example, a mechanical connection that must remain intact during clinical use may require testing that evaluates its ability to withstand specified loading conditions. A cosmetic surface characteristic that does not affect safety or essential performance may justify a different verification approach. The sample size should reflect the purpose of the test, the nature of the requirement, and the uncertainty that the verification activity is intended to address.

Risk should not be considered in isolation. The selected confidence level, reliability objective, test method, expected variability, and relevant standards also influence the final sample quantity. A risk-based sampling plan is most useful when it establishes a consistent method for making these decisions before testing begins.

Attribute Data vs. Variable Data

One of the most significant factors affecting verification sample size is whether the test generates attribute data or variable data.

Attribute data generally describes a categorical outcome, such as pass or fail, acceptable or unacceptable, or the presence or absence of a defined characteristic. For example, a verification test may determine whether a device successfully withstands a specified load without failure. If the study is intended to support a population reliability claim based on zero observed failures, a binomial statistical approach may be appropriate.

Variable data consists of numerical measurements, such as tensile strength, pressure, dimensional accuracy, displacement, or force. Rather than recording only whether a device passes, variable testing captures how the measured characteristic behaves relative to its specification. These measurements can provide information about the distribution and variability of performance.

Variable data may support a more efficient sampling strategy when an appropriate statistical model and sufficiently capable measurement method are available. However, a smaller sample size is not automatically justified simply because numerical measurements are collected. The appropriate calculation depends on the intended statistical inference, variability, distribution assumptions, specification limits, and confidence requirements.

For example, measuring the pull strength of catheter connectors can provide information about average strength and variation between assemblies. Whether that information is sufficient to demonstrate compliance depends on the acceptance criteria and the statistical method selected. If the objective is to establish that a specified proportion of the population exceeds a minimum strength requirement, a tolerance-limit approach may be relevant. If the objective is to estimate the average strength within a defined margin, a confidence-interval calculation may be more appropriate.

Device Configurations, Worst-Case Selection, and Manufacturing Variation

The number of samples is only one part of verification planning. Manufacturers must also determine which device configurations should be tested and whether the selected units adequately represent the intended product family.

Medical devices may be offered in different lengths, diameters, materials, or configurations. Testing every possible combination may be impractical, but selecting a single configuration without justification may fail to address important performance differences. In some circumstances, a worst-case or bracketed testing approach can help identify representative configurations. The appropriate selection depends on which characteristics influence the requirement being verified.

ASTM F3172, a recognized guide addressing device size and sample-size selection for endovascular device design verification, provides a relevant example of this principle. The guide addresses both device configuration selection and statistical approaches to determining sample quantities for individual design requirements.

Manufacturing variation must also be considered. Verification samples manufactured from one carefully selected component lot may not represent the variation expected across different suppliers, material lots, or production conditions. Where relevant, sampling plans should account for these sources of variability and justify how the selected units represent the device configuration being evaluated.

Discuss Your Verification Build Requirements

Part 3: How to Calculate Medical Device Verification Sample Sizes

Understanding 95% Confidence and 95% Reliability

One commonly discussed approach to medical device verification sampling uses a specified confidence level and reliability objective. These terms are related but describe different statistical concepts.

Reliability, in this context, refers to the proportion of a defined population expected to meet a particular requirement. Confidence describes the statistical strength of the inference made from the sample results. A 95% confidence and 95% reliability demonstration using an appropriate one-sided binomial method aims to establish a 95% lower confidence bound of at least 95% population reliability.

For a zero-failure attribute test, the required sample size can be calculated using the following relationship:

n ≥ ln(1 − C) / ln(R)

Where:

  • n is the required number of independent samples.
  • C is the desired confidence level.
  • R is the required population reliability.

For example, using 95% confidence and 95% reliability:

n ≥ ln(0.05) / ln(0.95) = 58.4

Rounding upward gives a minimum of 59 independent samples, assuming every sample passes and the binomial model’s assumptions are satisfied.

Why Are 29, 59, and 299 Samples Commonly Used?

The following table illustrates the minimum sample quantities for selected reliability objectives at 95% one-sided confidence using an exact zero-failure binomial approach.

Confidence Reliability objective Minimum samples Allowed failures
95% 90% 29 0
95% 95% 59 0
95% 99% 299 0

These values are mathematical results under specific assumptions, not universal regulatory sample-size requirements. They assume independent, representative Bernoulli trials with a consistent underlying probability of success, a fixed sample size, and zero observed failures. They also apply to a defined attribute and population, not automatically to every requirement or configuration within a medical device program.

The difference between 59 and 299 samples illustrates why reliability objectives have a substantial effect on verification planning. Increasing the required population reliability from 95% to 99%, while maintaining 95% confidence, increases the zero-failure sample requirement considerably. However, selecting the higher target without a documented technical or risk-based rationale can create unnecessary testing costs without necessarily improving the relevance of the evidence.

The calculation also assumes that the samples provide independent information. Repeatedly testing the same device does not automatically create additional independent samples. Similarly, testing many devices from a single highly controlled build may not adequately represent important manufacturing or material variation. These limitations must be considered when applying a statistical sampling method.

When Can Fewer Samples Be Justified?

A smaller sample quantity may be appropriate when the verification objective, test method, statistical analysis, and associated risks support that decision. For example, certain design requirements can be verified through dimensional inspection, engineering analysis, or a demonstration that directly evaluates the relevant characteristic. Other requirements may use continuous measurements and statistical methods that provide meaningful evidence without requiring the sample quantities associated with a zero-failure attribute demonstration.

A smaller sample size may also be supported by applicable consensus standards, prior technical evidence, justified worst-case testing, or a specific statistical design. However, these approaches should not be used merely to reduce testing costs. The manufacturer must establish why the evidence is adequate for the particular design requirement.

Importantly, a small number of successful tests does not automatically demonstrate a high level of population reliability. If five devices pass a test, the result establishes that those five devices met the acceptance criteria under the tested conditions. Any broader statistical claim requires an appropriate method and supporting assumptions.

Explore Device Verification Builds

Part 4: Common Mistakes in Verification Sampling Plans

Applying One Sample Size to Every Design Requirement

A frequent planning mistake is assigning the same number of samples to every verification test without evaluating the differences between requirements. Although a standardized sampling procedure can support consistency, a single fixed quantity may not be appropriate for all testing activities. Mechanical strength, dimensional accuracy, functional performance, and other characteristics may require different statistical methods and levels of evidence.

A more effective approach begins with the design input requirements and establishes an appropriate verification method for each. The sampling plan should identify what is being measured, how the results will be evaluated, and why the selected sample quantity is sufficient. This creates a clearer relationship between design requirements, risk management, test protocols, and the conclusions drawn from the results.

Using Samples That Do Not Represent the Finished Device

Verification testing may generate misleading conclusions when the tested devices differ materially from the configuration intended for production. Engineering prototypes can be useful for development testing, but they may use different materials, manufacturing processes, components, or assembly methods from the final device.

FDA bench-testing guidance generally recommends that tested devices represent the final, finished product and account for relevant manufacturing and processing conditions. Where samples differ from the finished configuration, manufacturers should explain why the differences do not undermine the applicability of the test results.

For devices that undergo sterilization, environmental conditioning, or other processing steps that may affect performance, these conditions may need to be represented in the verification strategy. The relevance of each condition depends on the device and the requirement being evaluated.

Failing to Account for Destructive Testing and Sample Allocation

Some verification activities permanently alter or destroy the tested device. Tensile testing, burst testing, fatigue testing, and certain mechanical evaluations may leave units unsuitable for subsequent use. If the same devices are allocated to multiple tests without considering their effects, the test sequence may compromise the validity of later results.

Verification planning should therefore consider sample allocation alongside statistical requirements. Engineering and manufacturing teams need to determine which tests can be performed sequentially on the same device, which require independent units, and whether prior testing changes the characteristics being evaluated.

This is particularly important for expensive or low-volume devices. An incomplete allocation plan can result in additional manufacturing builds, material shortages, or delays while replacement samples are produced. Establishing test quantities and sequencing before the verification build begins helps align manufacturing requirements with the approved protocols.

Changing Sample Sizes After Reviewing Results

Sample-size decisions should be established before formal testing whenever the statistical design requires a fixed sample quantity. Reducing the sample size after observing favorable results, or increasing it selectively following an unfavorable result, can compromise the intended statistical interpretation unless the approach is supported by a properly designed adaptive or sequential testing method.

Unexpected results should be investigated and documented. When a test fails, the team must determine whether the result reflects a device problem, test-method issue, manufacturing variation, or another relevant factor. Any changes to the protocol or additional testing should follow appropriate quality procedures and preserve the integrity of the verification evidence.

Review Our Manufacturing Capabilities

Part 5: Planning Design Verification Builds Around Sample Requirements

Establishing a Verification Sample Plan Before Manufacturing

Once verification requirements and sample quantities have been established, the manufacturing team must determine how the necessary units will be produced. This includes evaluating component availability, material specifications, device configurations, assembly procedures, traceability, and the documentation needed to connect each sample with its manufacturing history.

A verification program may require multiple device configurations, different conditioning groups, destructive and nondestructive testing, or samples from more than one manufacturing lot. These requirements can significantly affect the total number of units needed. The statistical sample size for an individual test should therefore not be confused with the total manufacturing quantity required to execute the entire verification program.

For example, a device may require 59 units for one zero-failure attribute test while other requirements use different sample quantities or configurations. Some units may be eligible for more than one test, while others must be manufactured separately because testing is destructive or the required conditioning differs. A complete build plan should account for these distinctions, along with reasonable allowances for setup, documented manufacturing losses, and approved contingency requirements.

Connecting Verification Planning With Controlled Manufacturing

Verification results are most useful when the manufacturing history of the tested devices is appropriately documented. Component traceability, controlled work instructions, assembly records, inspection results, and change management help establish what was tested and how those devices were produced. This information becomes particularly important when verification results are used to support subsequent regulatory submissions or manufacturing decisions.

We support medical device companies through production-intent design verification builds, pilot manufacturing, controlled assembly, and manufacturing documentation. Our integrated engineering, manufacturing, and quality capabilities allow us to coordinate the practical requirements of producing verification units while supporting traceability and repeatability.

Within our ISO 13485-certified quality management system, we help establish appropriate manufacturing instructions, inspection activities, material controls, and documentation for verification-stage programs. Our ISO Class 7 cleanroom capabilities further support precision assembly and controlled manufacturing operations where required.

By connecting verification planning with manufacturing execution, we help device companies prepare the units needed for testing while maintaining the flexibility necessary for development-stage programs. The appropriate sample-size calculations, acceptance criteria, and verification conclusions remain specific to each device and should be established through the responsible engineering, quality, and regulatory processes.

Discuss Your DV Build

Frequently Asked Questions About Medical Device Process Validation

Does the FDA require 30 samples for design verification?

No. There is no universal FDA requirement to test 30 samples for every medical device design verification activity. The appropriate sample size depends on the requirement, test method, associated risk, statistical approach, and applicable guidance or consensus standards. Manufacturers should establish and document a scientific rationale for the selected quantity.

Why are 59 samples commonly used in medical device testing?

A sample size of 59 is associated with a zero-failure binomial demonstration of at least 95% population reliability at 95% one-sided confidence. This result depends on assumptions about independence, representativeness, and the test outcome. It is not automatically appropriate for every design verification test.

What is the difference between attribute and variable testing?

Attribute testing records categorical outcomes, such as pass or fail, while variable testing produces numerical measurements, such as force, pressure, or dimensions. The appropriate sample-size calculation depends on the type of data, the statistical claim being made, and the characteristics of the test method.

Can the same medical devices be used for multiple verification tests?

Potentially, provided the test sequence does not invalidate subsequent measurements and the approach is consistent with the approved verification protocols. Destructive testing, environmental conditioning, or prior loading may affect device performance and require separate samples.

Do verification samples need to be manufactured under production conditions?

Samples should appropriately represent the device configuration and conditions relevant to the requirement being verified. FDA bench-testing guidance generally recommends finished-device representation and scientific justification when different samples are used. The necessary degree of production representativeness depends on the test and applicable requirements.

How many extra devices should be manufactured for design verification?

There is no universal contingency quantity. The total build should account for approved test sample sizes, device configurations, conditioning groups, destructive testing, manufacturing setup needs, and justified contingencies. Extra units do not automatically become part of the statistical sample unless their inclusion follows the approved testing plan.

Building a Defensible Design Verification Strategy

Determining how many samples are needed for medical device design verification requires more than selecting a familiar number. The appropriate quantity depends on the design requirement, risk, statistical method, device configuration, and the evidence needed to demonstrate acceptable performance. A scientifically justified sampling plan connects these considerations with the testing methods and acceptance criteria established for the device.

For companies approaching design verification, early coordination between engineering, quality, regulatory, and manufacturing teams can help prevent unnecessary builds and ensure that the required samples are produced under appropriate conditions. By integrating controlled manufacturing, documentation, traceability, and engineering support, we help medical device companies prepare for verification activities while establishing a stronger foundation for subsequent production.

Start Your Verification Conversation


References

Exit mobile version