FDA 483 Citations for Statistical Techniques and Sampling Plans Under 21 CFR 820.250

22 hours ago

5 min read

FDA investigators cited 21 CFR 820.250 three ways on system-generated Form FDA 483s in fiscal year 2025:

"Sampling plans are not [written] [based on valid statistical rationale]," 820.250(b), 12 times
"Procedures for identifying valid statistical techniques required for establishing, controlling, and verifying the acceptability of process capability and product characteristics have not been [adequately] established," 820.250(a), 11 times
"Procedures to ensure sampling methods are adequate for their intended use have not been [adequately] established," 820.250(b), 9 times

Sampling and statistics sit underneath many other requirements: receiving inspection, lot acceptance, process validation, design verification and complaint trending. When an investigator asks "why did you test 10 units?" and the answer is "we always have," a statistics citation follows, and the conclusions built on that sample are in doubt.

What the citations mean

Examples written after "Specifically" usually describe:

Sample sizes chosen by habit, with no rationale
Acceptance sampling with no defined AQL, LTPD or confidence and reliability level
Validation studies using three units with no justification
Sampling plans not linked to the risk of the characteristic being checked
No procedure describing which statistical methods are used and when
Trend analysis of complaints or nonconformances with no defined method
Sampling plans not reviewed after process or product changes

What 21 CFR 820.250 required

Under the Quality System Regulation, where appropriate, each manufacturer had to establish and maintain procedures for identifying valid statistical techniques required for establishing, controlling and verifying the acceptability of process capability and product characteristics (820.250(a)).

Sampling plans, when used, had to be written and based on a valid statistical rationale. Each manufacturer had to establish procedures to ensure that sampling methods are adequate for their intended use and that, when changes occur, the sampling plans are reviewed. These activities had to be documented (820.250(b)).

What changed under QMSR

Under QMSR, in effect since February 2, 2026, statistical requirements come from several ISO 13485:2016 clauses:

Clause 8.1 requires the organization to plan and implement monitoring, measurement, analysis and improvement processes, including determination of appropriate methods, including statistical techniques, and the extent of their use.
Clauses 7.3.6 and 7.3.7 require design verification and validation plans to include methods, acceptance criteria and, as appropriate, statistical techniques with rationale for sample size.
Clause 8.4 requires analysis of data, including feedback, conformity to requirements, process and product characteristics and trends, suppliers, audits and service reports, using appropriate methods including statistical techniques.

The expectation is the same as before, but the link between sample size and risk is clearer. A sampling plan should reflect how serious a failure of that characteristic would be.

How to respond to the 483

FDA generally considers responses received within 15 business days of the close of the inspection before deciding on further action.

Justify or replace the cited sampling plans

For each plan cited, document the statistical basis: the standard used, the AQL or confidence and reliability level, and why it fits the risk of the characteristic. Where the old plan was inadequate, assess what it may have missed.

Assess past decisions

For validations or acceptance decisions based on unjustified samples, decide whether the conclusions still hold. Retest or extend studies where needed.

Write a statistical techniques procedure

Approved sampling standards, such as ISO 2859-1, ANSI/ASQ Z1.4 or ISO 3951
A table linking risk levels to AQLs or confidence and reliability levels
Methods for attribute and variable data
Methods for process capability, trend analysis and data analysis
Review of sampling plans after changes
Training or qualification for people who apply the methods

Apply it everywhere

Review receiving inspection, in-process and final acceptance, validation, design verification and trending against the procedure.

What a compliant statistical techniques procedure contains

Scope and responsibilities
Approved methods and standards
Risk-based selection of AQL or confidence and reliability levels
Sample size rationale for validation and verification
Process capability and control charting
Trend analysis methods for complaints, nonconformances and service data
Review of plans after changes
Records

Common mistakes in statistics responses

Citing a standard without stating the AQL, inspection level or rationale
Using the same sample size for every characteristic regardless of risk
Fixing acceptance sampling but leaving validation sample sizes unjustified
Writing a procedure that no one in production understands
Forgetting to review plans after a process change

Example: how one response section might read

This illustrates structure only; your response must describe your own facts.

Observation 16: Sampling plans are not based on valid statistical rationale. Specifically, seal strength for each lot is tested on five pouches with no documented rationale.

Response:

Correction: Based on the risk analysis, seal strength was classified as a major characteristic. A sampling plan providing 95 percent confidence that at least 95 percent of seals conform was adopted on a stated date, requiring 59 samples with zero failures.
Assessment: Historical seal strength data from 80 lots showed a process capability index above 1.67, supporting the conclusion that past lots conformed.
Corrective action: A statistical techniques procedure was issued with a risk-based table of confidence and reliability levels for attribute and variable sampling.
Extended review: All 42 sampling plans in receiving, in-process and final inspection were reviewed against the procedure by a stated date.

Frequently asked questions

Is ANSI/ASQ Z1.4 acceptable for medical devices?

Yes, when the AQL and inspection level are chosen and justified for the risk of each characteristic.

What confidence and reliability levels should we use?

There is no single required level. Many firms use higher levels, such as 95 percent confidence and 99 percent reliability, for critical characteristics and lower levels for minor ones. Document the basis.

Do we need statistics for complaint trending?

ISO 13485 clause 8.4 requires appropriate methods, including statistical techniques, for data analysis. Define how trends are identified and what triggers action.

How A.P. Lyon documents help

A.P. Lyon's Statistical Techniques Procedure for ISO 13485 and FDA QMSR covers sampling standards, risk-based selection, process capability and trend analysis. It supports the Lot Inspection Procedure, the Validation Protocol and Report Procedure and the Complaint Trending Procedure.

Documents are editable Word and Excel files and must be adapted to your products and processes. Sample sizes must be justified for your own risks.

Next steps

Justify or replace each cited sampling plan
Reassess conclusions built on unjustified samples
Issue a statistical techniques procedure and apply it across the system
Respond within 15 business days with dates for each action

To build a complete set of procedures at a volume discount, use Build Your QMS. For help with a 483 response, see A.P. Lyon consulting services.

Not sure where your quality system stands? Take the free QMS gap check: about 20 questions, an instant report of your gaps by clause, and the procedures that close each one.