Pharmaceutical Deviation and OOS Investigation Workflow

Three Signals, Three Different Starting Points

Teams often lump these together, but each one opens a different path. A deviation is about how the work was done. An OOS is about a result that fails. An OOT result still passes but breaks the expected pattern. Getting the label right on day one decides who investigates and what evidence is needed. Talk through your own triage rules with our quality team.

What the 2006 Guidance and Its 2022 Revision Set Out

FDA first issued "Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production" in October 2006 and published Revision 1 in May 2022. The definition of OOS did not change. The guidance covers traditional lab testing and release, not process analytical technology (PAT) approaches. It sets out a phased path from the first failing number to a final batch decision. Map your current SOP against it in a guided workflow review.

Analyst and Supervisor: Who Checks What

Phase I is a lab investigation, done quickly and without assumptions. The guidance is clear that a result can be invalidated only when a test event is observed and documented. It also says laboratory error should be relatively rare, and frequent errors point to problems such as analyst training or equipment upkeep. Get a sample Phase I checklist from our support team.

No Lab Error Found? The Production Floor Gets Reviewed

A full-scale investigation is led by the quality unit, with production, engineering and the lab contributing. It reviews how the batch was made and sampled, and the written record should state the reason for the investigation, the likely causes, whether the problem has happened before and the actions taken. The MHRA's OOS guidance, first published in August 2013, describes the same step as a manufacturing investigation. Plan the hand-off between lab and production in a Phase II design session.

The Rules for Retesting, Resampling, Averaging and Outliers

Extra testing is where most OOS findings start. The guidance calls "testing into compliance" unscientific and objectionable, and says all results, passing and suspect, should be reported and considered in the batch decision. The matrix shows what each rule allows and how a workflow can hold the line. Check your own retest SOP with our QA specialists.

Four Ways an OOS Case Can End

The quality unit interprets the findings and makes the batch decision. A confirmed OOS means the batch does not meet its specification and should be rejected. When the outcome is inconclusive, the guidance says the QU should always err on the side of caution. Rehearse your closure criteria in an investigation closure demo.

Batch Impact Through Genealogy

21 CFR 211.192 requires a thorough investigation of any batch failure, whether or not the batch has been distributed, and says it must extend to other batches of the same product and other products that may be associated with the failure. Genealogy turns that from a search through paper into a query. Ask our support team how lot links are built from your ERP and batch records.

Close the Loop With Corrective and Preventive Action

ICH Q10 expects a CAPA system that takes in deviations, non-conformances, product rejections and trends, uses a structured approach to find the root cause and scales the effort to the level of risk. An OOS that closes without a linked action, or with a CAPA that is never checked, is a gap. Plan your CAPA links in a CAPA linkage session.

The Fields Every Investigation Record Should Hold

Whatever system you use, the record should let a reviewer follow the case without asking anyone. These are the fields iFactory's template carries by default, and your SOP can add more. Ask our QA specialists for a blank version.

Size the Hours Spent Assembling Each Case

The biggest risk of a weak OOS process is a regulatory finding, which is hard to price and should not be guessed. What you can size is the time spent pulling raw data, batch records and lot histories together by hand. The figures below are assumed inputs to show the arithmetic only. Build the case with your own numbers in a sizing call.

Add the value of fewer late cases, faster batch decisions and cleaner inspection records, but only where your own data supports a figure.

Implementation Guide: OOS Workflow and AI Server in One Package

iFactory delivers OOS Workflow as a turnkey package. A pre-configured NVIDIA AI server ships racked and ready, so investigation records, lab data and genealogy stay on your site and the AI can flag linked lots and repeat failures locally. You rack it, plug in power and Ethernet, and the AI is live. Scope the rollout with our implementation team.

An OOS Workflow That Follows the Guidance

iFactory is built so the phased approach is the easy path, not an extra burden on the lab. See it on your own SOP in a workflow demo.

OOS Investigation Workflow for Plant Managers and QA