thebench.bio #1 Ecosystem for the Bio-Tech Industry
All Process Validation questions

Question 17 of 20 Advanced Practical

How would you design a continued process verification plan?

Short answer

Choose the critical quality attributes and process parameters that the control strategy depends on, plus key material attributes. Trend them with statistical tools at the heightened PPQ level until variability is well estimated, define rules and responses for signals, review regularly with QA and production, and reduce sampling only on evidence.

Why

The FDA's 2011 guidance says Stage 3 needs an ongoing program to collect and analyze product and process data, including incoming materials, in-process materials and finished product, trended statistically and reviewed by trained people. It recommends that a statistician or someone trained in statistical process control design the plan, and that procedures guard against both overreacting to single events and missing real variability. It also says to keep monitoring at the PPQ level until there is enough data for significant variability estimates, then adjust.

Annex 15 asks for ongoing process verification under an approved protocol with a report, using statistics to support conclusions about variability and capability, and feeding the product quality review. A good plan therefore lists each parameter and attribute with the reason it was chosen, the chart type and rules, the sampling frequency and when it can change, who reviews and how often, and what happens on a signal. It should look at within-batch and between-batch variation, as the guidance stresses.

It should also take in softer signals named in the guidance, such as complaints, deviations, yield and operator feedback. A plan fails when it charts everything, so that real signals drown, or when charts are produced monthly and nobody owns the response.

How to work it through

  1. Select CQAs, CPPs and critical material attributes from the control strategy and justify each one.
  2. Set baselines from development and PPQ data and choose chart types and run rules.
  3. Keep sampling at the PPQ level until variability estimates are reliable.
  4. Define the review frequency, owners and response for each type of signal.
  5. Include complaints, deviations, yields and operator feedback as inputs.
  6. Review the plan periodically and change sampling only with documented evidence.

An example

For a monoclonal antibody process, the plan tracks 9 attributes and 12 parameters, including titer, aggregate level, charge variants, host cell protein, final cell viability, Protein A step yield and resin cycle count. Each has an individuals chart with Western Electric rules. For the first 30 batches, sampling stays at PPQ intensity.

A monthly review with QA, manufacturing and process science looks at signals. At batch 30, aggregate level variability is well estimated with Cpk 2.1, so in-process aggregate testing at an intermediate step is reduced, documented through change control.

Where the answer stops holding

Statistical rules assume reasonably stable, independent data, which batch processes with campaigns may not provide. Few batches per year make baselines slow to establish. Charting can become a ritual unless signals reliably trigger action.

A strong answer covers

  • Parameters and attributes chosen from the control strategy
  • Statistical trending with defined rules
  • PPQ-level sampling until variability is known
  • Defined review, owners and responses
  • Within and between batch variation plus soft signals

Common mistakes

  • Charting every available parameter with no rationale.
  • Reducing sampling right after PPQ without variability data.
  • Producing charts that nobody is responsible for acting on.