1What the regulations and guidance ask for.
ICH Q10, EU GMP and FDA’s drug guidance all ask for the effectiveness of CAPAs to be evaluated. None of them sets a method or a duration.
| Source | Status | What it says |
|---|---|---|
| ICH Q10, sections 3.2.2 and 3.2.4 | ICH guideline, published by FDA as guidance and in EU GMP Part III | In commercial manufacturing, “CAPA should be used and the effectiveness of the actions should be evaluated”. Management review covers “the effectiveness of process and product changes including those arising from corrective action and preventive actions”.¹ |
| EU GMP Chapter 1, 1.4(xiv) and 1.10 | EU GMP guideline, in operation since 31 January 2013 | “The effectiveness of such actions should be monitored and assessed, in line with Quality Risk Management principles.” The product quality review covers significant deviations or non-conformances “and the effectiveness of resultant corrective and preventive actions taken”.² |
| EU GMP Chapter 8, 8.9 and 8.18 | EU GMP guideline, in operation since 1 March 2015 | Quality defect procedures should address the need for CAPAs “and for the assessment of the effectiveness of those CAPAs”. The effectiveness of CAPAs taken for a quality defect “should be monitored and assessed”.³ |
| FDA quality systems guidance, 2006 | FDA guidance for drug CGMP | Procedures should ensure “a selected action is taken within a defined timeframe, and the effectiveness of the action taken is evaluated”.⁴ |
| 21 CFR Part 820 | US regulation for devices, in its current form since 2 February 2026 | Incorporates ISO 13485:2016, whose Clauses 8.5.2 and 8.5.3 cover corrective and preventive action. FDA says the degree of action should be “appropriate to the magnitude of the problem and commensurate with the risks encountered”.⁵⁶ |
2Implementation, verification and effectiveness.
A CAPA raises three separate questions, and closing on the first is the usual failure.
- Implementation: was the action done? The revised procedure is in force, the people are trained, the part is replaced.
- Verification: does the action work as intended, without harming anything else? FDA’s CDRH course gives the example of spell-check added as a corrective action that then starts changing words in a presentation. The action has to be shown to be effective and not to adversely affect the product.⁷
- Effectiveness: did the problem stop, or fall, over a defined period? ISO 9000 defines effectiveness as the “extent to which planned activities are realized and planned results are achieved”.⁸
ICH Q10 makes the same split for changes. After implementation, a change is evaluated “to confirm the change objectives were achieved and that there was no deleterious impact on product quality”.¹ Q10 names CAPA among the things that drive change, so a corrective action made through change control gets both evaluations.¹
3Plan the check before the action takes effect.
Write the effectiveness criteria into the CAPA when the action is approved, not when the check falls due. Q10 asks change management to set prospective evaluation criteria for each proposed change.¹ For the data that feed a CAPA system, FDA’s CDRH course asks for measures that are quantitative when possible and, when qualitative, keep subjectivity to a minimum. It expects a company to be able to explain its statistical methods and how they set its thresholds for opening a CAPA.⁷ Effectiveness criteria need the same discipline.
A plan answers six questions. The answers below are an illustration for one CAPA, not a standard.
| Question | Example answer |
|---|---|
| What outcome should change? | Seal leaks on filling line 2 |
| How is one event counted? | One confirmed leak per run, whatever records report it |
| Over what population? | Runs on line 2 that used the revised procedure |
| Measured against what exposure? | Leaks per 100 runs, not a raw count |
| For how long? | Until the old rate would have produced at least three leaks |
| What result passes, and who decides? | A rate at or below the target, assessed by the CAPA owner and approved by QA |
Match the measure to the action. A training action is judged by errors on the trained task, not by training records. A supplier action is judged by the rejection rate of that supplier’s lots, an equipment fix by failures per operating hour. A count without its denominator can mislead: 12 failures in 150 runs and 5 in 60 are nearly the same rate.
4How long the check should run.
ICH Q10, EU GMP Chapters 1 and 8, FDA’s 2006 guidance and 21 CFR Part 211 set no duration. They ask that effort follow risk: Q10 ties investigation effort to “the level of risk, in line with ICH Q9”, and Chapter 1 ties effectiveness monitoring to quality risk management principles.¹²
Set the length by exposure, not by the calendar. The check has to run long enough that, had the action done nothing, the problem would probably have shown itself. Work it out from the baseline:
- Find the old rate. Say 8 failures in the 100 runs before the action.
- Choose a window in which the old rate would have produced at least three events. Here that is about 40 runs.
- Read the result. If nothing had changed, the chance of seeing no events in a window where three were expected is about 5% (e^−3, under a Poisson model). No failures in 40 runs is reasonable evidence; no failures in 10 runs is not.
For a rare problem the same arithmetic gives a long window. At two events a year, three expected events take about 18 months. A quiet 90 days would be expected about six times in ten even if nothing had changed (e^−0.5, about 0.61). For those, verify the action directly, for example by challenging a new control with test cases, and keep watching through trending and the annual product review. In the EU that review covers the effectiveness of CAPAs, and in the US the annual review covers the investigations made under 21 CFR 211.192.²⁹
Fix the window before the results arrive. Extending it until the numbers look good, or cutting it short after a quiet month, defeats the check.
5When the check fails.
A failed or inconclusive check is a finding, not an embarrassment. Record it, keep the original assessment, and reopen the investigation. The action may not have been fully implemented, the population may have changed, or the cause may have been wrong.
EU GMP accepts that the true root cause cannot always be found and asks for consideration to be given to identifying and addressing “the most likely root cause(s)”. It asks that human error be justified, having made sure that process, procedural or system-based errors have not been overlooked.² FDA’s CDRH trainer warns that training used “over and over” as the corrective action, while the problem persists, should prompt a closer look at the investigation.⁷ Q10 routes the outcome upward: management review covers the effectiveness of changes arising from CAPA, and the review of the quality system assesses CAPA process performance.¹
6How it runs in Seal.
Neil, Seal’s AI agent, can compare a CAPA’s event rates. It reads the CAPA action, its implementation evidence and effectiveness plan, comparable events and operating exposure before and after the change, and the exclusions, related changes and open investigations, then drafts a recurrence comparison with its limitations. The CAPA owner assesses effectiveness and QA approves the conclusion. The worked example shows why the denominator matters: leakage events fall from 12 in 150 runs to 5 in 60, but the rate rises from 8.00 to 8.33 per 100 runs.
In the CAPA blueprint, the effectiveness review keeps its baseline, population, exposure and follow-up window, and closure can be configured to require the approved review. A failed or inconclusive check follows a reassessment path that preserves the earlier decision.
References
- 1ICH Q10, Pharmaceutical Quality System, ICH, Step 4 version of 4 June 2008, sections 3.2.2, 3.2.3, 3.2.4 and 4.1. https:/
/ database. ich. org/ sites/ default/ files/ Q10%20Guideline. pdf (read 6 November 2026) - 2EudraLex Volume 4, Part I, Chapter 1: Pharmaceutical Quality System, European Commission, in operation since 31 January 2013, sections 1.4(xiv) and 1.10. https:/
/ health. ec. europa. eu/ system/ files/ 2016- 11/ vol4- chap1_ 2013- 01_ en_ 0. pdf (read 6 November 2026) - 3EudraLex Volume 4, Part I, Chapter 8: Complaints, Quality Defects and Product Recalls, European Commission, 13 August 2014, in operation since 1 March 2015. https:/
/ health. ec. europa. eu/ document/ download/ b1eb2292- cb0d- 4e3f- aea9- e3fe79faf6e3_ en?filename=2014- 08_ gmp_ chap8. pdf (read 6 November 2026) - 4Guidance for Industry: Quality Systems Approach to Pharmaceutical CGMP Regulations, FDA, September 2006, section IV.D.4. https:/
/ www. fda. gov/ media/ 71023/ download (read 6 November 2026) - 521 CFR Part 820, Quality Management System Regulation, sections 820.7 and 820.10, eCFR. https:/
/ www. ecfr. gov/ current/ title- 21/ chapter- I/ subchapter- H/ part- 820 (read 6 November 2026) - 6Medical Devices; Quality System Regulation Amendments, final rule, FDA, 89 FR 7496, 2 February 2024, response to comment 20. https:/
/ www. govinfo. gov/ content/ pkg/ FR- 2024- 02- 02/ html/ 2024- 01709. htm (read 6 November 2026) - 7Corrective and Preventive Actions, transcript, CDRH Learn, FDA, content current as of 4 January 2018. https:/
/ www. fda. gov/ training- and- continuing- education/ cdrh- learn/ corrective- and- preventive- actions- transcript (read 6 November 2026) - 8ISO 9000:2015, Quality management systems: Fundamentals and vocabulary, Clause 3, entry 3.7.11, ISO Online Browsing Platform, edition marked withdrawn. https:/
/ www. iso. org/ obp/ ui/ en/ #iso:std:iso:9000:ed- 4:v1:en (read 6 November 2026) - 921 CFR 211.180(e), General requirements for records and reports, eCFR. https:/
/ www. ecfr. gov/ current/ title- 21/ section- 211. 180 (read 6 November 2026)
AQuestions and answers
What is the difference between CAPA verification and an effectiveness check?
Verification asks whether the action works as intended without side effects, for example that a new alarm triggers at the right limit. An effectiveness check asks whether the problem the CAPA targeted has stopped or fallen over a defined period. A CAPA can pass the first and fail the second.
Is a fall in the number of events enough?
No. Compare rates against exposure. Twelve failures in 150 runs and five in 60 are almost the same rate, so the lower count says nothing about the action. Use events per batch, run, operating hour or lot.
Can a CAPA be closed before the effectiveness check is finished?
The guidance does not set the order, but the evaluation should still be done and recorded. If your procedure closes the CAPA when the actions are implemented, keep the effectiveness check as a tracked record with its own criteria, due date and owner, so it cannot drop out of view.
What if the problem was a one-off with no history to compare?
A recurrence rate will not show much quickly. Verify the action directly, for example by challenging the new control with test cases, then watch the process through routine trending and the annual product review.
Does completed training prove a training CAPA worked?
No. Training records show the action was implemented. To show effect, look at performance on the task, such as errors or deviations on the steps the training covered, over a period long enough for the old rate to have shown.