Where process characterization sits (FDA Stage 1, ICH Q8(R2) QbD)

Process characterization is the last and most data-intensive activity of Stage 1, Process Design, in the FDA's 2011 guidance Process Validation: General Principles and Practices. Stage 1 builds process knowledge; Stage 2, Process Qualification, confirms the process at commercial scale; Stage 3 verifies it during routine production. The guidance expects the manufacturer to understand sources of variability and their effect on quality before PPQ, and characterization is how a biologics manufacturer meets that expectation.

ICH Q8(R2) (2009) supplies the vocabulary. A critical quality attribute (CQA) is a property that must stay within a limit to ensure product quality. A critical process parameter (CPP) is a parameter whose variability affects a CQA. The guidance describes an enhanced, quality-by-design approach in which the relationship between parameters and attributes is established by design of experiments, leading to a design space and a control strategy. Process characterization is the experimental core of that approach. ICH Q9(R1) (2023) governs the risk assessments that decide what to study, and ICH Q11 (2012) applies the same thinking to drug substance.

Timing matters. Characterization is usually run after the commercial process is frozen, between Phase 2 and the start of PPQ. Earlier wastes runs on a process that will change; later leaves no time to act on the findings.

Start with risk: CQA list and parameter ranking (FMEA)

A process characterization study is only as good as the risk assessment that scopes it, because a typical mammalian process has 60 to 100 parameters and no study can vary them all. The sequence is fixed: define the CQAs, map each unit operation's parameters against them, score the risk, and study only what scores high.

The CQA list comes from the quality target product profile and from product knowledge. For a monoclonal antibody it typically includes aggregates, charge variants, glycan species such as afucosylation and high-mannose content, fragments, host cell protein, host cell DNA, residual Protein A and potency. Each CQA is scored for impact on safety and efficacy.

Parameters are then ranked with a cause-and-effect matrix or a failure mode and effects analysis (FMEA). In an FMEA each parameter receives a severity score for its potential effect on CQAs, an occurrence score for how likely it is to drift in manufacturing, and a detectability score for how readily that drift would be seen. The product, the risk priority number, sorts parameters into those that enter the DoE study, those fixed by platform knowledge, and those that are non-critical and need only routine control. The rationale for each exclusion must be documented, because reviewers will ask why a parameter was not tested.

Choosing the design: screening versus response-surface

Characterization studies use two classes of design of experiments, and choosing the wrong one wastes either runs or information. Screening designs test many factors at two levels to find which matter. Response-surface designs test fewer factors at three or more levels to model curvature and interactions and to locate edges of failure.

A common structure for a bioreactor step is a fractional factorial over six to eight parameters such as temperature, pH, DO, seeding density and feed rate, followed by a central composite design over the two or three that showed significant effects. Center points are replicated to estimate pure error, and runs are randomized and blocked by bioreactor or by day. Every DoE study should pre-define its response variables, statistical model and significance threshold before the first run.

Qualifying the scale-down model first

A process characterization study is run at bench scale, so its conclusions apply to the commercial process only if the bench system is a qualified scale-down model. Qualification is a documented comparison between bench and at-scale performance at the same set points, completed and approved before the DoE study starts.

Upstream, the model is a 2–5 L bioreactor matched to the production vessel on power per volume, gassing strategy and feed profile. Comparison covers viable cell density and viability, glucose, lactate and ammonia profiles, titer, and the CQA panel. Downstream, the model is a small column at production bed height, residence time and load, or a membrane device at the same flux and load per area. Equivalence is assessed statistically, for example by two one-sided tests against a pre-set margin, or by showing that bench results fall within the range of at-scale batches. A systematic offset is documented and the conclusions adjusted for it.

Outputs: proven acceptable ranges, normal operating ranges, control strategy

The deliverable of process characterization is a set of ranges and a parameter classification, not a collection of models. Three definitions carry through to the regulatory file.

• Proven acceptable range (PAR). The range of a parameter over which the study demonstrated that all CQAs stayed within acceptance criteria while other parameters were held at set point or varied within their own ranges. ICH Q8(R2) defines a PAR as a characterized range, distinct from a multidimensional design space.
• Normal operating range (NOR). The narrower range used in routine manufacturing, set by equipment capability and control precision. The gap between NOR and PAR is the process margin.
• Parameter classification. Parameters whose variation within the studied range affected a CQA are critical process parameters. Those that affected yield or a performance attribute but not a CQA are often labeled key process parameters. The rest are non-critical.

These feed the control strategy described in ICH Q10 (2008): which parameters are monitored and alarmed, which in-process controls and acceptance criteria apply at each step, and which attributes are tested at release. Worst-case verification runs, with several critical process parameters set at the edges of their PARs at once, are a common final step to confirm that the ranges hold in combination. The characterization report, with risk assessment, design rationale, data, statistics and resulting ranges, enters Module 3 section 3.2.S.2.6, Manufacturing Process Development.

How characterization data feeds the PPQ protocol

Process characterization decides what the process performance qualification protocol measures and where its acceptance criteria come from. The FDA 2011 guidance makes the link explicit: the approach to PPQ should be based on sound science and the manufacturer's overall level of product and process understanding and demonstrable control.

The characterization report supplies four inputs to the PPQ protocol. The list of critical process parameters defines what is recorded and must stay within NOR in every PPQ batch. The PARs justify the NORs and the alert limits around them. The CQA list and acceptance criteria define the enhanced sampling plan, including in-process samples outside routine release. And the scale-down model qualification lets viral clearance and resin lifetime studies run at bench scale support the commercial process. Where characterization found no sensitivity, the protocol can justify lighter sampling; where it found steep responses, it adds samples and tighter criteria.

The same data set serves the lifecycle after approval. Under ICH Q12 (2019), established conditions and their reporting categories can be proposed from the characterized ranges, so a well-run DoE study reduces post-approval submissions over the life of the product.