How to run an OOS investigation (Phase I / Phase II)

The OOS procedure starts the moment the analyst sees the result. Before anything is discarded or changed:

1. Keep everything — sample solutions, standards, glassware where practical, instrument output and raw data files.
2. Do not re-inject, re-prepare or retest on your own initiative.
3. Inform the supervisor or the person defined in your SOP promptly, the same working day.
4. Record the result in the GMP record. An OOS result is never left out because it might be invalid.

An OOS result that is overwritten, deleted or tested again before an investigation is opened is a serious data-integrity finding. Controlled systems that lock submitted results and keep raw data unaltered make this much harder to do by accident.

Phase I asks one question: is there a clear, assignable laboratory error? The analyst and supervisor review the test together, using a checklist that typically covers:

• Correct test method, specification version and sample identity.
• Calculations and data transcription, including any spreadsheet used.
• Instrument performance — calibration status, system suitability, maintenance and audit trail.
• Standards and reagents — preparation, expiry, storage and certificates.
• Sample preparation — dilutions, extraction, glassware and filtration.
• Analyst training on the method, and any observations noted during the test.
• Chromatograms and raw data, including integration parameters and any reprocessing.

Some laboratories split this into Phase Ia, an obvious error such as a spillage or calculation slip found quickly, and Phase Ib, a deeper hypothesis-led laboratory investigation. MHRA guidance uses this split; the FDA guidance treats it as one laboratory phase.

If a laboratory error is confirmed

The original result can be invalidated only when the error is documented and scientifically justified. The test is then repeated, and the laboratory error itself should lead to corrective action — for example method clarification, retraining or instrument repair — so it does not recur.

If no laboratory error is found

The OOS result stands and the investigation moves to Phase II. “Analyst error” recorded without evidence is one of the most frequent OOS findings in inspections.

Phase II brings in QA and production. It looks for the root cause in manufacturing and sampling as well as in the laboratory, and assesses the impact on the batch and on other batches.

Manufacturing and sampling review

• Batch record review — process parameters, deviations, equipment and personnel.
• Materials — raw material lots, supplier history and any changes.
• Sampling — how, where and by whom the sample was taken, and whether it was representative.
• Trend — previous results for the product, similar products, and stability data.

Retesting

Retesting means testing a portion of the original sample again. It may be appropriate in Phase II when it helps test a hypothesis, but it must follow rules defined in the SOP before the investigation begins: how many retests, who performs them (preferably a different analyst), and how results will be evaluated. You cannot keep retesting until a passing result appears, and a passing retest does not on its own invalidate the original OOS result.

Resampling

Resampling — taking a new sample from the batch — should be justified by evidence that the original sample was not representative or was compromised, and follow a documented sampling plan.

Averaging and outliers

Averaging can hide variability, so it is only appropriate where the method itself is designed around an average result, and individual results should still be reported. Statistical outlier tests are generally not appropriate for chemical tests with low variability; they may have a place for inherently variable biological assays when defined in advance.

The investigation report should state clearly which hypotheses were tested, the evidence, the root cause (or that none was found), and the batch disposition decision. If the OOS result is confirmed, the batch fails to meet specification and must be rejected or handled under your procedures; impact on other batches must be assessed and documented.

Corrective and preventive actions should address the root cause, with owners, due dates and an effectiveness check. For US-marketed distributed products, a confirmed or unresolved OOS result can also trigger a Field Alert Report within the regulatory timeframe — check this against your regulatory obligations.

Common pitfalls inspectors find

• Results invalidated as “analyst error” without supporting evidence.
• Retesting until a passing result appears, or averaging an OOS result away.
• Phase II closed without reviewing manufacturing or other batches.
• Trial injections or unofficial tests run before the reportable test.
• Investigations open far beyond the time set in the SOP, without justification.
• CAPA raised but never checked for effectiveness.

Swat iQua Suite manages OOS investigations on the same platform as its LIMS, CAPA and batch release, designed to comply with 21 CFR Part 11 and EU GMP Annex 11 and following ALCOA+ principles.

• OOS investigation in up to three levels. Level 1 always runs; Levels 2 and 3 are set when required, so you can map them to your Phase I and Phase II procedure. The system works out the current level — users cannot skip ahead.
• Each level has its own review and approval, with segregation of duties (the approver cannot be the creator or that level’s reviewer) and optional e-signatures per site.
• A finalise step locks the investigation; controlled print is only available after finalising, and every print is logged.
• When CAPA is required, one CAPA is raised and approved together with the final level, then tracked through verified actions and optional effectiveness review.
• Links to the testing sample or instrument maintenance record behind the result, and lab quality event records for OOT results, incidents and other laboratory events.

Data integrity behind the result

• Results lock once submitted; send-back to the analyst is the only route to change them, and the change is recorded with a reason.
• Controlled result import from Excel/CSV and instrument files through approved profiles, with the original file kept as raw data and a SHA-256 checksum.
• Specification versions frozen at sample registration, so the result is always judged against the limits that applied.
• Instrument calibration status checked before use, and second-person raw data review where a finding raises a deviation.

The result, the investigation, the CAPA and the batch release decision share one audit trail, so the full story is a few clicks away when an inspector asks.