How to investigate a bioreactor process deviation

Start with the observed event

A deviation investigation starts by defining what changed, when it changed and which process or quality outcome may be affected. For example, a dissolved oxygen excursion is a different question from an unexpected viable cell density trend. Record the run, phase, expected range, first observed time and immediate operational response before exploring causes.

Follow the site's approved quality procedures for any regulated event. The workflow below describes analytical support; it does not replace required deviation handling or release decisions.

1. Verify the signal and its provenance

• Inspect original historian or instrument records, units, calibration status and any missing samples.
• Check time zones, sampling rates, sensor maintenance and transformations before comparing traces.
• Separate measured assay results from modeled estimates or imputed values.
• Preserve the original record and the reason for any exclusion or correction.

2. Align the process timeline

Place feed changes, inoculation, sampling, antifoam additions, controller actions, alarms and operator notes on one time axis. A plausible cause must precede its proposed effect. Keep material lots and equipment configuration attached to the run so that a similar trace from another vessel is not treated as equivalent without review.

3. Compare meaningful reference runs

Select comparable runs by process version, product, cell line or strain, scale, vessel and relevant operating conditions. Show why each reference was included. Compare the shape and timing of the excursion, not only end-of-run averages, and look for any prior runs where the same event occurred without the same outcome.

4. Test hypotheses without losing uncertainty

• List competing explanations such as sensor fault, feed timing, material variation or biological response.
• For each explanation, identify expected evidence and evidence that would contradict it.
• Use models to explore plausible consequences only inside their evaluated range, and show uncertainty.
• Escalate unresolved or high-impact questions to process, engineering and quality experts before action.

5. Produce a reviewable record

A useful report identifies the selected run and time window, source records, calculations, exclusions, comparison runs, hypotheses, model version and limitations. It separates observations from inference and proposed action. Capture who reviewed the evidence and which follow-up test or control change was approved.

After the investigation, record whether the hypothesis was confirmed and whether the new evidence changes the model or process understanding. That feedback prevents the same issue from being rediscovered in a later campaign.

Common questions

Sources and scope

These references inform the terminology and evaluation approach. They do not certify any software or replace a process-specific validation plan.

• FDA Process Validation: General Principles and Practices
• FDA data integrity guidance
• ICH Q9(R1) quality risk management