What is continued process verification, and how do I set one up and react to what it shows?

This is United States federal law, read for three classes: drugs including over-the-counter drugs, human food under the preventive controls rule, and dietary supplements. Which one you are in decides whether you owe anyone an ongoing verification program and what it has to contain, so that comes first. A monitoring plan built to the wrong rule is work you cannot use and cannot defend.

If you make a dietary supplement, start here. The federal supplement rule does not use the word validation anywhere in it, and subparts C and G of the food rule, which are where its validation and verification clauses sit, do not apply to supplement manufacturing that is in compliance with part 111 and section 761 of the Federal Food, Drug, and Cosmetic Act. So there is no federal requirement addressed to you that is called continued process verification, or anything close to it. You still have real duties, and they are named below. If a customer or an auditor has asked you for a CPV plan, the ask is coming from a contract or a certification rather than from a regulation, and reading which one it is changes what you have to build.

On this page: Which rule reaches you · What it actually is · Setting one up · Reacting to what it shows · What quietly ends it · What the inspection record shows · Three things to check

Which rule reaches you

The only text read here that uses the words is a certification standard, not a regulation: validated processes are subject to documented continued verification confirming that the process remains in a state of control (NSF/ANSI 455-4, 4.5.1.4). Everything else that reaches you says the same thing in its own vocabulary, and the vocabulary matters, because it decides which clause an investigator writes you up under.

Drugs, including over-the-counter drugs. The ongoing obligation sits inside the in-process control clause. Written procedures shall be established and followed that describe the in-process controls and tests or examinations to be conducted on appropriate samples of in-process materials of each batch, and such control procedures shall be established to monitor the output and to validate the performance of those manufacturing processes that may be responsible for causing variability in the characteristics of in-process material and the drug product (21 CFR 211.110(a)). Read the two verbs. Monitor the output is a continuing act, not a study you close out. Two more clauses carry the loop: records shall be maintained so that data therein can be used for evaluating, at least annually, the quality standards of each drug product to determine the need for changes in drug product specifications or manufacturing or control procedures (21 CFR 211.180(e)), and any unexplained discrepancy or the failure of a batch or any of its components to meet any of its specifications shall be thoroughly investigated (21 CFR 211.192).

Human food under preventive controls. There is no stage numbering here and no process validation in the drug sense. What there is instead is a verification duty. You must verify that the preventive controls are consistently implemented and are effectively and significantly minimizing or preventing the hazards, and to do so you must conduct activities that include the following, as appropriate to the facility, the food, and the nature of the preventive control and its role in the facility’s food safety system: calibration of process monitoring instruments and verification instruments, or checking them for accuracy; product testing for a pathogen, an appropriate indicator organism, or another hazard; environmental monitoring, if contamination of a ready-to-eat food with an environmental pathogen is a hazard requiring a preventive control; and review of records (21 CFR 117.165(a)). That records review has a clock on it: monitoring and corrective action records must be reviewed within 7 working days after they are created, or within a reasonable timeframe, provided that a preventive controls qualified individual prepares, or oversees the preparation of, a written justification for a timeframe that exceeds 7 working days (21 CFR 117.165(a)(4)(i)). On top of that, the food safety plan as a whole is reanalyzed at least once every 3 years (21 CFR 117.170(a)).

Dietary supplements. Two clauses decide this, and both are exclusions. Subparts C and G of the food rule do not apply to any facility with regard to the manufacturing, processing, packaging, or holding of a dietary supplement that is in compliance with the requirements of part 111 and section 761 of the Federal Food, Drug, and Cosmetic Act (21 CFR 117.5(e)). Subpart C is where validation and verification live, so those clauses above are not addressed to you. And in the 590 clauses of part 111, the word validation does not appear once. What part 111 asks for instead is a production and process control system covering all stages of manufacturing, packaging, labeling and holding (21 CFR 111.55); in-process specifications at any point, step or stage in the master manufacturing record where control is necessary (21 CFR 111.70(c)(1)), with adequate documentation of your basis for why meeting them, in combination with meeting component specifications, will help ensure the finished specifications are met (21 CFR 111.70(c)(2)); a responsibility to determine whether the specifications you established are met (21 CFR 111.73); and quality control review of all monitoring required under subpart E, and determining whether all in-process specifications established in accordance with 111.70(c) are met (21 CFR 111.123(a)(3), 111.123(a)(6)). That is an ongoing loop. It is just not called one, and nothing in it asks you to trend anything.

One more exclusion is worth reading against your own operation, because it removes the same subparts for a different reason. Except as provided by subpart E of the food rule, subparts C and G do not apply to a qualified facility, which is on modified requirements instead (21 CFR 117.5(a)). A facility that qualifies is outside the verification clauses above and submits an attestation instead, and there are two it may choose between. Either that it has identified the potential hazards associated with the food being produced, is implementing preventive controls to address the hazards, and is monitoring the performance of those controls to ensure they are effective (21 CFR 117.201(a)(2)(i)), or that it is in compliance with State, local, county, tribal or other applicable non-Federal food safety law (21 CFR 117.201(a)(2)(ii)). A qualified facility that does not submit the first of those must give consumers the name and complete business address of the facility where the food was manufactured or processed (21 CFR 117.201(e)).

The certification you hold, which for many readers is the real answer. A scheme binds you because you claimed it. The over-the-counter drug standard carries the requirement in full, and its supplement counterpart does not. The supplement standard names establishing and maintaining a state of control as one of three objectives in its purpose clause (NSF/ANSI 455-2, 1.1), and then carries no clause asking for an ongoing process performance monitoring program at all. The over-the-counter drug standard names the same objective (NSF/ANSI 455-4, 1.1) and then requires the program: there is a defined process for monitoring process performance and product quality (NSF/ANSI 455-4, 4.6.1), and an ongoing program to collect and analyze product and process data that relate to product quality must be established (NSF/ANSI 455-4, 4.6.1.1). The difference between the two standards is the difference between an objective and a requirement, and it is the difference a customer asking a supplement plant for a CPV plan has usually not noticed.

The wider question of which regime reaches which product for process validation as a whole, and what each of them does and does not name, is answered on its own page: how many PPQ batches are actually required. This page assumes that question is settled and takes up what happens afterwards.

What it actually is

Strip the staging language and the thing has a definition, and the definition is about a condition rather than about a document. A state of control is a condition in which the set of controls consistently provides assurance of continued process performance and product quality (NSF/IPEC/ANSI 363, definitions). Mark where that comes from before you use it: a private certification standard for pharmaceutical excipient manufacture, which reaches you only where something else brings it. We give it because it is the cleanest statement of the condition among the sources read here, not because it governs your product. Continued process verification is whatever you do, week after week, to know whether you are still in that condition.

The fullest statement of it among the sources read here sits in a pharmaceutical quality system guideline. Its stated objective is to develop and use effective monitoring and control systems for process performance and product quality, thereby providing assurance of continued suitability and capability of processes (ICH Q10, 1.5.2). The system itself is spelled out: plan and execute a system for the monitoring of process performance and product quality to ensure a state of control is maintained; use quality risk management to establish the control strategy, which can include parameters and attributes related to materials and components, facility and equipment operating conditions, in-process controls and finished product specifications, and the associated methods and frequency of monitoring and control, with the control strategy facilitating timely feedback and feedforward and appropriate corrective and preventive action; provide the tools for measurement and analysis, including data management and statistical tools; analyze those parameters and attributes to verify continued operation within a state of control; identify sources of variation affecting process performance and product quality; and include feedback on product quality from internal and external sources, including complaints, product rejections, non-conformances, recalls, deviations, audits and regulatory inspections and findings; and provide knowledge to enhance process understanding, enrich the design space where one is established, and enable innovative approaches to process validation (ICH Q10, 3.2.1).

Mark the standing of that before you build to it. It is a harmonised guideline, not a United States regulation and not a standard you are audited against. It reaches you where something else brings it: a marketing application, a customer contract, or a certification that adopts it. The over-the-counter drug standard is one that does, and gives that guideline section as the basis for its own clause (NSF/ANSI 455-4, 4.6.1).

The phrase itself reaches most readers through agency guidance and through the software and consultants that follow it. That guidance is not read here, and nothing on this page describes what it says. Nothing on this page depends on it.

Setting one up

What to watch

The text that comes closest to a specification for the data set is in the certification standard, and it is written as a should rather than a must: the data collected should include relevant process trends and quality of incoming materials or components, in-process material, and finished products; the data should be statistically trended and by personnel trained in statistical process control and reviewed by the quality unit (NSF/ANSI 455-4, 4.6.1.2). Four sources, not one. Mark the standing before you build to it. In that standard the program itself is a must and what goes into it is a should (NSF/ANSI 455-4, 4.6.1.1), and both reach you only if you hold the certification. We give it here because it is the most concrete statement of the data set among the sources read here, not because it binds every reader.

The common plan watches only the last of the four. Finished-batch release results are easy to get, already in a table, and already reviewed. A plan built on them alone tells you that every batch passed, which you knew, and tells you nothing about whether the process moved. The federal clause is pointed the other way: it asks for control procedures on the manufacturing processes that may be responsible for causing variability in the characteristics of in-process material and the drug product (21 CFR 211.110(a)). That is a question about the process, and a release assay is not an answer to it.

The same mistake has a second form that is harder to see. A monthly average of a parameter is a number about a month, not about the process. Where a plan trends the average and not the spread, a process that is drifting in one direction and a process that is oscillating produce the same chart. If you keep one habit from this page, keep both: what the middle is doing, and how wide the scatter is.

Where the limits come from

This is the part that goes wrong most often, and it goes wrong quietly. Limits get computed from the process’s own recent output, so the chart is measuring the process against itself. It will alarm when the process changes and it will never alarm when the process was never good enough, because it has no idea what good enough is.

The drug rule asks for two things at once, and conditions the second. Valid in-process specifications for such characteristics shall be consistent with drug product final specifications and shall be derived from previous acceptable process average and process variability estimates where possible and determined by the application of suitable statistical procedures where appropriate (21 CFR 211.110(b)). Consistent with the final specification, and, where possible, derived from what the process actually does. Drop the first half and the limit says only that the process is doing what it was doing. Note the word acceptable in that clause too: a baseline drawn from a period your own quality unit would not sign off on is not a baseline.

For a supplement plant the equivalent is not a control chart, it is a written rationale. The clause asks for adequate documentation of your basis for why meeting the in-process specifications, in combination with meeting component specifications, will help ensure the finished specifications are met, and for quality control personnel to review and approve that documentation (21 CFR 111.70(c)(2), 111.70(c)(3)). Where that basis is written down, you have the thing a monitoring plan is built on. Where it is not, there is nothing to trend against, and the trend you build will be a picture of your own habits.

How often, and for how long

The honest answer to frequency is not a number, and the clearest statement of why is in the certification standard: monitoring and sampling of process parameters and quality attributes should be continued at the level established during the process qualification stage until sufficient data is available to generate significant variability estimates (NSF/ANSI 455-4, 4.6.1.4). Two things follow. You start at qualification-stage intensity rather than at a reduced routine level, and you step down when the data justifies it rather than when the calendar reaches a date. That clause uses should, and it is a scheme rather than a rule, so it binds you only if you hold the certification.

The cadences that are law are these, and they are slower than people expect. Records are maintained so the data can be used for evaluating, at least annually, the quality standards of each drug product, with written procedures for those evaluations that include a review of a representative number of batches, whether approved or rejected, and, where applicable, the records associated with the batch, and a review of complaints, recalls, returned or salvaged drug products and the investigations conducted under 211.192 for each drug product (21 CFR 211.180(e), 211.180(e)(1), 211.180(e)(2)). For food, monitoring and corrective action records within 7 working days (21 CFR 117.165(a)(4)(i)), and the plan as a whole at least once every 3 years (21 CFR 117.170(a)).

Read the annual one for what it is. It asks you to evaluate quality standards once a year to determine the need for changes in specifications or in manufacturing or control procedures. It does not ask you to run a chart. The running part of the obligation is the in-process control clause, and an annual review is not a substitute for it.

Reacting to what it shows

There are two ways to get this wrong and they pull in opposite directions. One clause names both: procedures should describe how trending and calculations are to be performed and should guard against overreaction to individual events as well as against failure to detect unintended process variability (NSF/ANSI 455-4, 4.6.1.3). Write the reaction rule before you have a signal in front of you, because after you have one it is hard to tell which of the two you are doing.

For a drug maker the reaction duty is the same clause that governs batch release, and it is broader than most plans treat it. All drug product production and control records, including those for packaging and labeling, shall be reviewed and approved by the quality control unit to determine compliance with all established, approved written procedures before a batch is released or distributed. Any unexplained discrepancy, including a percentage of theoretical yield outside the established maximum or minimum, or the failure of a batch or any of its components to meet any of its specifications, shall be thoroughly investigated, whether or not the batch has already been distributed. The investigation shall extend to other batches of the same drug product and other drug products that may have been associated with the specific failure or discrepancy. A written record of the investigation shall be made and shall include the conclusions and followup (21 CFR 211.192). Three duties inside that get dropped: distribution does not close it, it reaches other batches and other products, and the conclusions and the follow-up have to be in the record and not in somebody’s memory.

For food the reaction is written as procedure, and it is owed as appropriate to the nature of the hazard and the nature of the preventive control, except where the corrections paragraph applies. Corrective action procedures must describe the steps to be taken to ensure that appropriate action is taken to identify and correct a problem that has occurred with implementation of a preventive control, that appropriate action is taken where necessary to reduce the likelihood the problem will recur, that all affected food is evaluated for safety, and that all affected food is prevented from entering commerce where you cannot ensure it is not adulterated or misbranded (21 CFR 117.150(a)(2)). That machinery is not owed for everything. You do not need to comply with the corrective action paragraphs where you take action, in a timely manner, to identify and correct conditions and practices that are not consistent with the named food allergen or sanitation controls, or to identify and correct a minor and isolated problem that does not directly impact product safety (21 CFR 117.150(c)). Deciding which of those a signal is, is itself the judgment this section is about. And the review itself can trigger the escalation: where a review of records finds that the records are not complete, that the activities conducted did not occur in accordance with the food safety plan, or that appropriate decisions were not made about corrective actions, you are in the unanticipated problem route (21 CFR 117.150(b)(1)(iii)), which can require reanalysis of the plan (21 CFR 117.150(b)(2)(ii)).

For a supplement plant the reaction clause is a rejection clause with a narrow door in it. For specifications established under 111.70(a), (b)(2), (b)(3), (c), (d), (e) and (g) that you do not meet, quality control personnel must reject the component, dietary supplement, package or label unless they approve a treatment, an in-process adjustment or reprocessing that will ensure the quality of the finished dietary supplement and that it is packaged and labeled as specified in the master manufacturing record (21 CFR 111.77(a)). Read the whole section, because two categories have no door at all. An unmet component identity specification means the component is rejected and must not be used in manufacturing the supplement (21 CFR 111.77(b)), and an unmet specification on a product you received from a supplier for packaging or labeling means the product is rejected and may not be packaged or labeled for distribution as a dietary supplement (21 CFR 111.77(c)). Where the door is open, the approval requires a material review and a disposition decision based on a scientifically valid reason (21 CFR 111.90(b)(1)). Hold on to that phrase. A scientifically valid reason is a higher bar than a plausible one, and it is the bar the record has to show you cleared.

Now the failure none of those clauses catches. A signal appears, someone dispositions it, and the disposition is reasonable. It happens again the next month, and the same disposition is reasonable again. Nothing in any of the clauses above requires anyone to notice that it is the twelfth time, because each event was closed on its own merits. That is why the trending duty and the reaction duty are written as one thing rather than two: the data should be statistically trended and reviewed by the quality unit (NSF/ANSI 455-4, 4.6.1.2), and trends from process performance and product quality monitoring belong in the corrective and preventive action system alongside complaints, product rejections, non-conformances, recalls, deviations and inspection findings (ICH Q10, 3.2.2). The pattern is the finding. The individual events never were.

What quietly ends it

A validated state ends when the process changes, and the changes that end it are usually the ones nobody logged as changes. Equipment gets logged. A supplier’s grade change, a new operator on the shift, a batch size taken up because an order came in large, a slightly different mixing time that made Fridays work better: those are alterations to the arrangement in exactly the way a new machine is, and they are the ones that arrive without a change record.

The certification standard states the duty in one line: changes from the validated state are managed through a change control process (NSF/ANSI 455-4, 4.5.1.3). The food rule ties the same idea to reanalysis, which is required whenever a significant change in the activities conducted at your facility creates a reasonable potential for a new hazard or a significant increase in a previously identified hazard, and whenever you find that a preventive control, a combination of them, or the plan as a whole is ineffective (21 CFR 117.170(b)(1), 117.170(b)(4)). The drug rule reaches it through the procedures themselves: written production and process control procedures, including any changes, shall be drafted, reviewed and approved by the appropriate organizational units and reviewed and approved by the quality control unit (21 CFR 211.100(a)), and any deviation from the written procedures shall be recorded and justified (21 CFR 211.100(b)).

Monitoring and change control are the same system read from two ends. Change control asks what you did on purpose. Continued verification asks what happened, including the things nobody did on purpose. A plant that runs one without the other finds out late. The equipment half of this argument, and how a qualified state expires without anyone deciding it should, is its own question.

What the inspection record shows

FDA publishes its record of inspectional observations. Three clauses carry the ongoing loop for a drug maker, and they are cited at very different rates.

Across the same window the Drugs program carries 23,578 observation rows across 5,007 distinct inspections, so all three clauses together are a small share of what gets written. The gap between the first row and the other two is the part worth sitting with. The clause about acting on what the records show is cited more than three times as often as the clause about having the monitoring in place at all.

What these figures do not evidence. The first clause is not a continued verification clause. It governs production record review and discrepancy investigation generally, and most of what sits under it has nothing to do with trending; it is simply the clause a signal you failed to investigate lands on. A count of citations against a clause is evidence that the clause gets cited, and nothing more. It is not a rate, because these are counts of observation rows and of the inspections they sit in, not of firms, and not of the inspections where nothing was written. It says nothing about the quality of the monitoring programs that do exist: a clause cited for control procedures being absent is not evidence that the ones in place are sound. The recorded observation text is largely standardized, so it describes what was written down rather than everything that was asked across the table. Read it as a rough map of where this loop breaks, not as a probability that it will break for you.

Three things to check

Take dated copies of your current plan and its limits before you change anything. A monitoring plan and its acceptance criteria are controlled records, and revising one is your own quality unit’s decision, not ours.

1. Find the sentence that sets each limit, and read what sits after it. If the limit was computed from the last twelve months of your own output and nothing ties it to the finished specification, you have a chart that cannot fail. The rule asks for both, consistent with final specifications and, where possible, derived from previous acceptable process average and process variability estimates (21 CFR 211.110(b)).
2. Count how many of the four data sources the certification standard names you actually collect. Incoming materials, in-process material, finished product, and process trends (NSF/ANSI 455-4, 4.6.1.2). Most plans we read carry one and describe it as a program.
3. Pull the last twelve dispositions of your most common signal and read them together. If they say the same thing twelve times, the pattern is the finding and nobody has made it yet. That is the situation the investigation clause is written for, and it reaches other batches and other products, not just the one in front of you (21 CFR 211.192).