Sampling Plans and Sample Size for Design V&V Under ISO 13485

Sampling plans and sample size justification for design verification and validation under ISO 13485

1. What ISO 13485 actually requires (and does not)

Clauses 7.3.6 and 7.3.7 sit inside the design and development section of ISO 13485:2016, the medical-device quality management system standard covering the device life cycle, including design and development, production, and servicing ³⁵. Design and development controls may be excluded only where regulatory requirements permit, with the justification recorded ³⁵.

The important point for sampling: the ISO 13485 clause text does not itself specify sampling plans, acceptance quality limits, or statistical sample-size rationale for verification or validation ³⁵. The standard requires that verification and validation be planned, that acceptance criteria and methods be defined, and that results and records be retained, but it delegates the "how many units and why" question to the sponsor's chosen statistical methods. That is why manufacturers reach for the acceptance-sampling and statistical-interpretation standards below, and why FDA guidance repeatedly asks for a documented statistical rationale rather than a fixed number.

A parallel hook exists on the US side. Under the Quality System Regulation, statistical techniques and sampling plans must be written and based on a valid statistical rationale; FDA guidance reflects this by stating that where 100 percent inspection is not performed, the sampling plan should provide reasonable statistical assurance and the statistical choices should be documented and justified ¹⁹⁶. With FDA's Quality Management System Regulation aligning 21 CFR 820 to ISO 13485, the same expectation now bridges both frameworks.

2. Acceptance sampling by attributes (pass/fail characteristics)

Attributes sampling is the most common basis for design verification acceptance testing where each unit is judged conforming or nonconforming.

ISO 2859-1 — Sampling procedures for inspection by attributes, Part 1: sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection. Specifies single, double, and multiple sampling plans indexed by AQL; applicable to products, sub-assemblies, components, materials, and records ¹³⁷. The current metadata shows a 2026 edition ¹³⁷.
ISO 2859-2 — Part 2: sampling plans indexed by limiting quality (LQ) for isolated-lot inspection, for lots in isolation where the ISO 2859-1 switching rules do not apply ¹⁴⁴. This is the standard to reach for when you have a single design-verification lot rather than a continuing series.
ISO 2859-3 — Part 3: skip-lot sampling procedures, to reduce inspection on high-quality product from suppliers with effective quality controls ¹⁵³.
ISO 28590 / ISO 2859-10 — introductions to the ISO 2859 series that summarize the schemes in Parts 1 to 5 and give guidance on selecting the appropriate attributes inspection system ²⁸²⁹.

FDA recognition (attributes):

Standard	Title	FDA recognition no.	Extent
ANSI/ASQ Z1.4-2003 (R2018)	Sampling Procedures and Tables for Inspection by Attributes	5-62	Complete ⁸⁵⁸⁶
ISO 2859-1 (2nd ed. 1999-11-15, incl. Cor 1:2001, Amd 1:2011)	Sampling schemes indexed by AQL for lot-by-lot inspection	5-88	Complete ⁸³⁸⁸

Both Z1.4 and ISO 2859-1 are recognized in full, so a Declaration of Conformity to either is available to support a premarket submission ⁸⁵⁸³. Z1.4 and ISO 2859-1 are technically aligned attributes systems; Z1.4 is the more common US reference.

3. Acceptance sampling by variables (measured characteristics)

When the verification characteristic is a measurement on a continuous scale (force, dimension, volume), a variables plan achieves the same protection with a smaller sample than the equivalent attributes plan, at the cost of a normality assumption.

ISO 3951-1 — single sampling plans indexed by AQL for lot-by-lot inspection, one quality characteristic and one AQL; assumes a continuing series of lots, negligible measurement error, and an approximately normal distribution, with lot acceptability based on an estimate of percent nonconforming ¹¹⁵. Current edition is 2022 ¹⁰⁰.
ISO 3951-2 — general specification for single sampling plans indexed by AQL for independent quality characteristics ¹⁰⁶¹⁰⁷.
ISO 3951-3:2007 — double sampling schemes indexed by AQL ¹¹¹.
ISO 3951-4:2011 — procedures for assessment of declared quality levels (note: intended for assessing a declared quality level, not for lot acceptance) ¹⁰⁴.
ISO 3951-5 — sequential sampling plans indexed by AQL, known standard deviation ⁹⁴.
ISO 3951-6:2023 — single sampling plans for isolated-lot inspection indexed by limiting quality (LQ) ⁹⁹¹¹².

FDA recognition (variables):

Standard	Title	FDA recognition no.	Extent
ANSI/ASQ Z1.9-2003 (R2018)	Sampling Procedures and Tables for Inspection by Variables for Percent Nonconforming	5-48	Complete ⁸⁴⁸⁷

Z1.9 is recognized in full and establishes sampling plans and procedures for inspection by variables ⁸⁴. Note that FDA's recognized list here surfaces Z1.9 for variables but does not show a complete-recognition entry for the ISO 3951 series in the same query, so if you intend to declare conformity to a variables plan for a US submission, Z1.9 is the safer recognized reference ⁸⁴⁸⁷.

4. Sequential and reduced-inspection plans

For programs that want to minimize the number of tested units when quality is clearly good or clearly bad, sequential plans evaluate results as they accumulate:

ISO 28591:2017 — sequential sampling plans for inspection by attributes; suited to samples from processes and usable for lots and isolated lots ³⁴.
ISO 8422 (attributes, older edition shown as withdrawn) — sequential plans for attributes, suitable for acceptance sampling and verification of simple statistical hypotheses for proportions ³³.
ISO 8423:2008 / ISO 39511:2018 — sequential sampling plans for inspection by variables for percent nonconforming, known standard deviation, for a continuing series of lots with a normal (or near-normal) distribution ¹¹⁷¹²⁰⁹²⁹³.
ISO 2859-3 — skip-lot procedures (see Section 2) ¹⁵³.

5. Choosing among sampling systems

Selection guidance helps justify why a particular scheme, AQL/LQ, and inspection level were chosen, which is exactly the rationale ISO 13485 verification/validation planning expects to see:

ISO/TR 8550-1:2007 — guidance on the selection and usage of acceptance sampling systems for inspection of discrete items in lots, Part 1; general guidance on selecting a system, scheme, or plan for discrete items sampled with equal chance ²⁷.
ISO 28590:2017 / ISO 2859-10:2006 — introductions that guide selection within the attributes series ²⁸²⁹.
ISO/DTR 24962 — sampling-based conformity assessment; guidance for deriving a prior distribution and selecting an approach to design an acceptance sampling plan, covering isolated-lot and serial-lot inspection ³⁰.
ISO 24153:2009 — random sampling and randomization procedures. It supports the mechanics of drawing a random sample but explicitly does not advise on choosing the procedure, strategy, or sample size, pointing instead to ISO 2859, ISO 3951, ISO 8422/8423, and related standards for that ³¹.

6. Statistical basis for sample size: tolerance intervals, confidence, and reliability

Many design-verification acceptance criteria are framed as reliability/confidence statements (for example, "95% reliability at 95% confidence") rather than an AQL. The relevant statistical-interpretation standards are:

ISO 16269-6 — Statistical interpretation of data, Part 6: determination of statistical tolerance intervals, covering one-sided and two-sided intervals that contain at least a specified proportion of the population at a specified confidence level, with parametric and distribution-free methods ¹³⁵¹³⁴. The 2014 edition is under review; the 2005 edition is withdrawn ¹³⁵¹³⁴. Tolerance intervals are the standard tool for translating a reliability/confidence acceptance criterion into a required sample size.
ISO 3207:1975 — an earlier tolerance-interval standard, now withdrawn ¹²⁹.
ISO/AWI 11453 — Statistical interpretation of data: tests and confidence intervals relating to proportions; methods for two-sided confidence limits at a desired confidence level ²¹⁰.
ISO/NP 11524 — Estimation of nonconforming proportion and selection of sample size for inspection by attributes; directly relevant to attribute-based sample sizing (early-stage project) ²¹¹.

A practical caveat from the source review: the ISO tolerance-interval and proportion standards give the mathematics for confidence and tolerance intervals, but they are not written specifically as "design verification sample size" standards, and the catalog scope text does not tie them to sample-size justification in so many words ¹³⁵²¹⁰. In practice, ISO 16269-6 tolerance intervals and confidence-interval-for-proportion methods are what teams use to justify reliability-based sample sizes; the standards supply the method, and the manufacturer supplies the risk-based reliability and confidence targets.

7. What FDA guidance expects for V&V sampling

FDA does not mandate a single sampling standard, but its guidance is consistent about wanting a documented, risk-based statistical rationale:

Design Control Guidance for Medical Device Manufacturers. Design verification confirms that design outputs meet design input requirements, based on tests, inspections, and analyses, documented in the Design History File ³⁴. Design validation is broader, assuring the design conforms to user needs and intended uses; it encompasses verification, is performed on initial production units, lots, or batches (or their equivalents) under actual or simulated use conditions, and may use a representative sample of users ¹²⁵. The guidance does not set a general sampling rule for validation ¹.
Quality Management System Information for Certain Premarket Submission Reviews (draft). Validation plans should include methods, acceptance criteria, and, as appropriate, statistical techniques with sample size, descriptions, and supporting rationale; validation should be on representative product, including initial production units/batches or equivalents, with a rationale for the product used ¹².
Essential Drug Delivery Outputs for Devices Intended to Deliver Drugs and Biological Products (draft). Design verification sampling plans should be risk-based, accounting for indication, patient population, the drug delivered, context of use, and design/manufacturing complexity; the protocol should include a statistical sampling plan with the number of lots to be tested and acceptance criteria, using representative commercial-process lots ¹⁹⁵.
Marketing Clearance of Diagnostic Ultrasound Systems and Transducers. Where 100 percent sampling is not performed, the sampling plan should provide reasonable statistical assurance, the statistical choices should be documented and justified, and the sample size should not be less than three ¹⁹⁶.
Process Validation: General Principles and Practices. Although this addresses process rather than design validation, its logic is frequently applied to V&V sampling: the number of samples should be set in the protocol and be adequate to provide sufficient statistical confidence of quality within and between batches, with the confidence level chosen from a risk analysis of the attribute; PPQ-stage sampling should exceed routine production sampling ²¹². cGMP sampling plans must result in statistical confidence, and samples must represent the batch ²¹³.

8. Practical takeaways

ISO 13485 clauses 7.3.6 and 7.3.7 require planned, recorded verification and validation with defined methods and acceptance criteria, but delegate sample size to the sponsor's statistical methods ³⁵.
For attributes acceptance testing, cite ISO 2859-1 or ANSI/ASQ Z1.4 (both FDA-recognized, complete) ⁸⁵⁸³; use ISO 2859-2 for isolated lots ¹⁴⁴.
For measured characteristics, cite ANSI/ASQ Z1.9 (FDA-recognized, complete) or the ISO 3951 series ⁸⁴¹⁰⁰.
For reliability/confidence acceptance criteria, use tolerance-interval methods (ISO 16269-6) and confidence-interval-for-proportion methods (ISO/AWI 11453), supplying your own risk-based reliability and confidence targets ¹³⁵²¹⁰.
Use ISO/TR 8550-1 and the ISO 28590 introduction to document why a given scheme, AQL/LQ, and inspection level were selected ²⁷²⁸.
Whatever plan you choose, FDA guidance expects the sampling plan and sample-size rationale to be written, risk-based, tied to acceptance criteria, and executed on representative product ¹²¹⁹⁵¹⁹⁶.

Limitations

Scope summaries and lifecycle status were available for the ISO standards, but not the full normative text, so exact sampling tables, formulae, and clause wording must be read in the standards themselves. Several statistically relevant ISO deliverables (ISO/AWI 11453, ISO/NP 11524, ISO/DTR 24962) are early-stage projects and not yet published ²¹⁰²¹¹³⁰. The FDA-recognized-standards query surfaced complete recognition for Z1.4, Z1.9, and ISO 2859-1, but did not return a complete-recognition entry for the ISO 3951 variables series in the same search, so confirm current recognition status in the CDRH database before declaring conformity ⁸⁵⁸⁴⁸⁸. AAMI technical information reports and the c=0 zero-acceptance-number plans commonly used in industry did not surface in these datasets and would be worth a targeted follow-up.