Resin substitution, cracked housings and a release decision on a running line

Shipping as-is is not defensible, because the organisation cannot currently state what the units are or whether they conform.

Reasons the evidence fails:

• The change affected material and a special process (ultrasonic welding), and revalidation was limited to dimensional checks on five parts from one lot - no functional or leak testing, no weld-strength data (7.5.6).
• The weld parameters were altered by a technician without a documented change control, risk assessment or validation, so the validated state described in the process validation report is no longer the process in use.
• The complaint rate is fifteen times baseline with a plausible causal link to the change; clause 8.2.1 feedback and 8.2.2 complaint handling require this to be evaluated, not monitored passively.

Containment:

1. Quarantine the 4,000 units; do not release. Segregate the affected lot range by traceability (7.5.9).
2. Identify whether the units were moulded with the new grade and whether the weld parameters applied - batch records and machine logs decide this, not memory.
3. Notify affected customers and distributors of the potential issue and hold shipments; assess whether field action is needed under 8.2.3, with the regulatory-reporting decision documented even if the conclusion is 'no report required'.
4. Retrieve and inspect returned units to build failure data; run a targeted accelerated or functional test comparing old-grade and new-grade assemblies.
5. Release only after the acceptance criteria are defined, testing is complete against them, and the responsible person signs a documented release decision (8.2.6). If the data cannot be produced before the due date, the shipment slips - the customer relationship is not a justification for releasing unverified product.

Scope of revalidation. Two linked changes: material and process. They should be validated together but evaluated separately, so a failure can be attributed.

• Material: confirm the new grade's specification, certificate of analysis, and biocompatibility and shelf-life implications. If the material is patient-contacting or affects the sterile barrier, a risk-based decision on biocompatibility and ageing testing is required, not an assumption of equivalence.
• Moulding: establish the process window for the new grade (melt temperature, pressure, cooling), and run an operational qualification across the window plus a performance qualification on three consecutive production lots.
• Ultrasonic welding: re-establish amplitude, hold time and energy settings through designed experiments, then validate with weld-strength and leak testing at the specification limits, plus a challenge at the worst-case parameter combination.

Acceptance criteria must be written before the runs: dimensional within drawing tolerance, weld tensile or peel strength at or above the historical minimum, leak/pressure decay within specification, and visual criteria for flash and cracks. Criteria that are invented after seeing the data are not criteria.

Sample sizes should be statistically justified against the defect rate that matters - for a 3 per 10,000 failure mode, a five-piece check has almost no chance of detecting it. The validation plan should specify sample size, confidence/reliability rationale, and who approves it: quality, not production engineering alone. Records must include raw data, not just a 'pass' summary.

Change control: the resin substitution should have triggered an incoming change notification, impact assessment, risk re-evaluation and a documented decision on revalidation before use. Add a supplier agreement clause requiring advance notification of material or process changes (7.4.2), and add material grade to the incoming inspection plan.

The housing is the symptom; four system failures produced it.

1. Supplier change notification (7.4.1, 7.4.2). The supplier notified six weeks late and no consequence followed. Actions: written agreements require advance notification and prior approval of changes; supplier monitoring includes change-history as a performance indicator; the supplier's approval status is reviewed. The purchasing data must specify the material grade and prohibit substitution without written acceptance.

2. Management of change (7.3.9 where design-related, 7.5.6 for process). A material change and a weld-parameter adjustment both bypassed impact assessment. Actions: a single change-control procedure with defined triggers (material, process, software, supplier, site), a mandatory risk-assessment step, and a documented revalidation decision. The technician's parameter adjustment is the critical finding - it shows the validated state is not protected on the floor. Parameter access should be locked, with changes only through the change-control route.

3. Competence and training (6.2). Signing 'read and understood' on behalf of two operators is falsification of a record, not a training shortcut. Actions: training records must identify the person trained and the trainer; effectiveness is verified by observation or assessment before independent work; the supervisor's action is addressed through the disciplinary process and the affected records are corrected, with the correction itself documented rather than backdated.

4. Risk management (7.1). The FMEA was not revisited after a material change. Actions: define explicit triggers for risk-file review, link the change-control form to the risk file, and require risk review before production release of a changed product. Occurrence 3 for a failure now running at fifteen times baseline is itself evidence that the risk file is not maintained.

Finally, decide whether a CAPA is raised for each failure or one CAPA with four workstreams - provided each root cause and each effectiveness check is separately defined (8.5.2).