The Repeat Nonconformity: Retraining That Did Not Work
Eleven months ago, a major nonconformity (CAPA-118) was raised because the leak test was not performed on three lots of catheter balloons. The recorded root cause was "operator error"; the corrective action was retraining six operators and revising the work instruction. You are now conducting the follow-up audit. You find: the leak test was skipped on one lot of 240 units, which was shipped after a deviation was closed by the production supervisor with the disposition "no impact," without QA approval; the work instruction cited in CAPA-118 has since been superseded twice; the training records show 4 of the 6 original operators were retrained, and 2 operators hired since the CAPA received no leak-test training; the deviation procedure requires QA disposition of all deviations affecting product acceptance. Separately, a caliper in the R&D lab has no calibration sticker; it is not used for production or acceptance testing.
1. Evaluate CAPA-118's effectiveness. Should it be reopened, and what does the evidence support?
Read the worked answer
The CAPA was not effective - the same nonconformity recurred, so effectiveness verification failed.
Evidence chain:
- Recurrence: the identical failure mode (leak test not performed) on a lot that was then shipped. Recurrence of the original nonconformity is the clearest possible effectiveness signal.
- Root cause was superficial: "operator error" is a description, not a cause. It stops the analysis before the questions that matter - was the leak test easy to skip (no poka-yoke, no interlock, no record field that forces a result), was the work instruction usable, was training effectiveness measured, was supervision adequate? The two subsequent revisions of the same work instruction suggest the document was not fit for purpose, and the two untrained new hires show training was treated as an event, not a controlled requirement.
- Action was not durable: retraining 4 of 6 operators, with no mechanism to train new hires, means the corrective action decays with normal turnover. That is a design flaw in the CAPA, not bad luck.
- Reopen or raise new? Either is defensible; best practice is to keep the link explicit - reopen CAPA-118 or raise a new CAPA that references it, states that effectiveness was not achieved, and requires re-analysis of root cause. Silently closing CAPA-118 as "corrected" while raising a fresh CAPA destroys the recurrence history that management review and auditors rely on.
- Escalate separately: 240 units shipped with a skipped test and a supervisor-only disposition needs a health hazard/risk evaluation and a reportability decision. The auditor does not make that decision - but the auditor must ensure it reaches someone with the authority to make it, in writing, during the audit.
2. How many findings should be reported, at what grades, and how should they be written?
Read the worked answer
Four separate findings - do not bundle, because they have different owners, different causes, and different corrective actions.
- Major - CAPA ineffective / recurrence. Criterion: corrective action requirements (ISO 13485:2016 8.5.2; 21 CFR 820.100) - action must be verified for effectiveness and proportionate to the risk. Evidence: CAPA-118 record; the current lot's missing leak test; superseded work instructions; training matrix showing 2 untrained operators. Include the prior audit reference explicitly so the repeat status is visible.
- Major - Nonconforming product dispositioned without required authority. Criterion: the firm's deviation procedure requiring QA disposition; nonconforming product controls (ISO 13485:2016 8.3; 21 CFR 820.90). Evidence: the deviation record, the supervisor's signature, absence of QA approval, shipping record for 240 units.
- Major or minor - Training/competence for personnel performing a critical operation. Criterion: competence and training requirements (ISO 13485:2016 6.2; 21 CFR 820.25). Evidence: 2 untrained operators performing leak testing; 2 of 6 retrained. Grade major if the untrained operators actually performed the test (patient-harm consequence); minor if they had not yet been assigned.
- Minor - Calibration identification of an R&D caliper. Criterion: monitoring and measuring equipment control, scoped to equipment used to demonstrate product conformity. Evidence: no calibration sticker. Keep it separate and minor, with a note that it is outside production use; bundling it with the leak-test findings inflates it, and inflating minor items damages the credibility of the major ones.
Writing each finding: requirement (clause) → condition (what was found, with lot numbers and record identifiers) → objective evidence (specific documents and observations) → significance (why it matters, including patient risk). No opinions, no "poor culture," no speculation about intent.
3. Design the follow-up that would actually verify effectiveness next time, and say what would count as proof.
Read the worked answer
Define the proof before the action, or the action will define itself as successful.
Effectiveness criteria (agreed at CAPA approval, written into the CAPA record):
- 100% record review of leak-test records for all lots produced during the verification window (recommend 6 months, minimum 20 lots), showing a completed leak-test result for every unit-lot, reviewed by QA - not by production.
- Zero deviations closed without QA disposition in the verification window, verified by listing all deviations and checking signatures.
- Training matrix closed: every operator currently assigned to leak testing is trained and the training is effectiveness-checked (observed performance, not attendance). Include a trigger that new hires cannot be assigned before training is complete.
- Error-proofing evidence, not just retraining: the corrected root cause should produce a change that makes skipping the test difficult or impossible (record field that blocks lot release, interlock, automated test capture). Retraining alone is not a corrective action for a recurrence - that is the lesson of CAPA-118.
- Trend: leak-test-related nonconformities and complaints at or below the pre-failure baseline for the window.
Who verifies: someone with authority independent of production, not the production supervisor, and not the same person who owned the original CAPA if a fresh evaluation is needed.
Timing: verification date set at CAPA approval; interim checkpoints (e.g., 60 days) so failure is caught early. Effectiveness is verified at the system level - did the change hold under turnover and instruction revisions - not by confirming that the training session happened. This follow-up is exactly what the previous CAPA lacked.