Quality starts with intended use

A successful bioprocess delivers more than a high final concentration. Its output must suit its purpose and be repeatable under understandable conditions. In research, the decisive quality may be a defensible conclusion about metabolism. For a technical enzyme, activity, purity and reproducible processing may matter. A medicinal product adds product-specific quality requirements and a regulated manufacturing framework. The same reactor technology can therefore serve very different quality systems.

The first step is an intelligible target description: what material or knowledge is produced, who will use it and which properties determine suitability? This prevents easily measured quantities from automatically becoming the principal objectives. Biomass can increase while desired product activity decreases. A visually tidy trend can arise from systematically incorrect analysis. Quality therefore connects output properties with the route by which the conclusion is obtained.

Documentation must support the purpose. For an exploratory experiment, precise material identity, assembly, event times, methods and deviations are often more useful than extensive forms lacking scientific content. Regulated manufacturing organises responsibilities, permissions and evidence around applicable requirements. Extra signatures do not automatically turn research equipment into a suitable production installation. Equally, sound scientific records retain value when generated outside a GMP process.

Quality development is consequently a lifecycle: define requirements, investigate relationships, demonstrate suitable equipment, confirm the process and monitor change. Each stage provides different information. A successful equipment check does not establish that product properties remain suitable over many batches. A good product sample does not establish that every measurement chain works correctly. Strength comes from a justified connection between them.

Product requirement → Process understanding → Control strategy → Evidence in operation. lifecycle feedback.
Schematic decision and feedback relationship; not a specific equipment configuration. Scroll sideways on small screens or open to enlarge.

Translate requirements into testable statements

A requirement should describe what a system must achieve under which conditions. “Good pH control” is too vague. A more informative description includes working volume, medium, permitted range, relevant disturbance and required measurement quality. Numerical limits follow the process task. Without that connection, specifications become a list of impressive equipment properties whose importance to later operation remains unclear.

Requirements also cover interfaces: utilities, sampling, data export, cleaning, maintenance and transfer to the next process step. A well-controlled culture can still yield unsuitable information if samples are assigned incorrect times. A suitable vessel can be limited by an inappropriate feed route. Assess the intended assembly including accessories. User requirements are not simply a transcription of the supplier catalogue.

For each significant requirement, define which evidence supports it. Document review can confirm material identity. Functional checking shows whether action and feedback agree. A study under relevant loading examines actual process performance. Different requirements need different forms of assessment. Align the choice with the question so that an expensive experiment does not merely confirm information already established reliably in existing records.

A traceable relationship connects requirement, risk, implementation, test and result. Open items remain visible, including their effect on the next stage. When a requirement changes, identify which earlier findings are affected. A longer maximum run, for example, can influence material exposure, sample count and data retention. This is a scientific change rather than simply a different number in a document.

Collaboration benefits from distinguishing essential requirements, desirable improvements and unresolved development questions. These categories do not judge the people involved; they support realistic decisions about procurement, study scope and readiness. A system may suit a limited development purpose while further evidence remains necessary for later use.

Distinguish product attributes from process parameters

A quality attribute describes a property of the material or product. A process parameter describes a condition or setting during manufacture. Product activity is therefore different from temperature or feed rate. In pharmaceutical usage, a Critical Quality Attribute, or CQA, identifies an attribute whose appropriate limitation matters to the intended quality. A Critical Process Parameter, CPP, is justified by the relevant effect of its variability on a critical quality attribute. A recorded quantity is not critical merely because it is stored.

The relationship comes from knowledge and investigation. Changing temperature can affect product quantity, structure or activity; magnitude and relevant range are application-specific. Statistical correlation provides a clue but needs assessment alongside experimental design, possible common causes and measurement quality. If every poor batch also ran longer and used another medium, the temperature effect cannot readily be isolated.

Hypothetical development example: For an enzyme, activity retained after a defined downstream step is considered an important property. Higher cultivation temperature produces more protein but less activity after processing. A concentration-only objective could then favour the wrong process. Investigation connects temperature, culture state, product analysis and the subsequent operation. Whether this becomes a CPP in a formal sense depends on the actual quality framework and demonstrated relationship.

Classification remains open to new knowledge. A material property initially considered unimportant may become relevant at larger scale. A tightly controlled quantity may prove to have little influence within an investigated range. Record changes and their rationale. A risk register should structure knowledge rather than permanently preserve early assumptions.

These terms are explained in ICH Q8(R2) on pharmaceutical development. The enzyme examples here apply the reasoning about relationships; they do not automatically classify any technical fermentation as a pharmaceutical process.

Prioritise risks with their evidence and uncertainty

Quality risk assessment asks what could impair the intended attribute, how that could occur and what evidence is available. Product impact is distinguished from likelihood and detectability. A rare event can remain important if consequences are substantial or detection is late. Frequent small variation may instead remain within a well-understood suitable range. One combined score explains these differences only imperfectly.

Tabulated assessments help when their basis remains intelligible. Classification supported by several suitable studies differs from an initial expert estimate. Record source data, assumptions and unresolved uncertainty. Calculations using subjective point scales do not create measured probabilities. They can support priorities but should not imply physical accuracy absent from their inputs.

A practical register connects failure mechanism, possible effect, existing control, unresolved question and planned investigation. For feeding, the question might be whether actual delivery variation affects the relevant product attribute. Pump percentage alone is then insufficient. For sampling, the question might concern whether delay shifts the quality result. Assessment consequently produces targeted information needs rather than generic demands for more control.

Risks change through the lifecycle. A new supplier, larger batch or different duration can challenge previous assumptions. Successful experience is also information when its conditions are known. Many unremarkable runs across a narrow operating range do not establish suitability beyond that range. Update assessment following relevant modifications and new findings.

Effort follows the decision. A straightforward, well-understood question may need a short rationale. Complex interactions with possible product effects require deeper assessment and suitable expertise. The decisive issue is whether the assessment plausibly explains development, testing and monitoring scope. An extensive form without an identifiable consequence adds little quality.

Build development studies into usable knowledge

Development investigates not only whether a setting works but why and across which range. Select relevant influences and possible interactions. Changing one factor at a time can reveal some relationships well while missing others. Multifactor experimental designs can help when question, permitted range and analysis suit them. Choose the method for its information value rather than because a statistical label sounds more professional.

Describe the investigated conditions clearly: organism or cell line, starting condition, media, assembly, working volume, operating mode and analysis. Without these details, a successful result is difficult to transfer later. Differences between studies may be biological or arise from differently handled samples. Development-data quality therefore depends strongly on representative sampling and suitable measurement chains.

Unfavourable results also contribute knowledge. They can reveal a boundary, disprove an assumption or expose an interaction. Summarising only successful runs produces a distorted process picture. Each analysis should show which experiments were included, excluded and judged unsuitable, with reasons. A technically failed run and a valid experiment producing unsuitable quality are different cases.

Transfer to larger equipment requires its own rationale. Identical setpoints do not automatically create identical local conditions because mixing, heat removal and transfer can differ. A small-scale model is useful when it represents the relevant question appropriately. State its limits explicitly. Scale-up should therefore connect early with the later quality question rather than beginning only after equipment purchase.

The outcome is a justified process description containing known ranges, important dependencies and open questions. It should explain which controls are needed and why. This knowledge can subsequently support test programmes, working instructions and monitoring. A collection of attractive batch traces without material and method context serves that purpose much less well.

Qualification assesses the specific installation

Qualification concerns a defined subject and intended use. It can cover facilities, equipment and utilities. Common terms DQ, IQ, OQ and PQ organise different questions: is the design suitable, was the intended system installed, do its defined functions operate and does it deliver required performance under suitable service conditions? Boundaries are established in the particular programme; the abbreviations alone do not guarantee uniform depth.

Installation assessment might examine identity, connections, drawings and available documentation. Functional assessment examines actions, feedback, limits and intended fault responses. Performance assessment needs relevant loads and states. A water run without a culture may be an appropriate technical check but does not automatically establish every function during viscous cultivation. Test conditions should meaningfully challenge the requirement.

Supplier testing can provide valuable evidence. Reuse nevertheless requires assessment of whether configuration, scope and conditions suit the user. Factory acceptance testing often occurs in a different environment from the final installation. Transport, assembly and connection to local utilities can create additional checks. Existing evidence is useful for avoiding unnecessary repetition while deliberately closing real gaps.

Assess adverse findings through their effects. An incorrectly labelled unused display differs from an ineffective interlock or unmet measurement performance. Small findings nevertheless remain traceable. If equipment is used for a limited purpose with open items, define that boundary and responsibility for closure. A blanket “passed” should not conceal the actual situation.

The official EU GMP Annex 15 on qualification and validation addresses these topics in medicinal-product manufacture. Equipment documentation or offered documentation packages do not automatically make the user's entire process GMP-compliant. Suitability comes from equipment, intended use and evidence actually generated.

Process validation is more than a fixed number of good batches

Process validation examines whether a defined process reliably delivers the intended result. It builds on development knowledge and suitable equipment. A series of successful batches can contribute but, without context, cannot answer every variability question. If all runs use one material lot, one operator and nearly identical surroundings, conclusions about other intended conditions remain limited. A supposedly universal batch count cannot replace a justified study strategy.

The FDA's process-validation guidance describes a lifecycle of process design, process qualification and continued process verification. This connects development with confirmed performance and subsequent monitoring. It is grounded in medicinal-product manufacture. The general approach of understanding variation and considering later routine operation also benefits other bioprocesses.

A suitable programme identifies material, configuration, operating ranges, sampling, analysis, acceptance criteria and deviation handling. Establish criteria before the decisive results are available. Otherwise, the standard may be adjusted retrospectively to make existing data appear satisfactory. Samples must also address the relevant question. One final value can conceal temporal or spatial differences significant to quality.

Ongoing observation examines whether the process remains in its understood condition. This includes trends in relevant attributes, materials, process conditions and deviations. An in-specification result can still belong to an adverse trend. Conversely, a statistical change does not automatically mean an unsuitable batch. Interpretation connects patterns with scientific significance and measurement uncertainty. Define appropriate responses for the different situations.

Keep equipment qualification, process performance qualification and product release distinct. Their subjects and decisions differ. A qualified balance supports quantity measurements rather than confirming an entire feed process. A confirmed feed process is itself only one part of manufacture. Precise terminology helps organise existing evidence and identify missing demonstrations.

Hypothetical trend example: An observed quality attribute is 96, 95, 94, 93, 92 and 91 relative units in six successive batches. Assume an agreed suitable range of 90 to 110 for this example. Every result lies within it, yet the sequence moves consistently in one direction. This creates a justified investigation question rather than automatically establishing an unsuitable batch. First examine measurement method, material lots, operating conditions and time assignment. Without knowledge of normal variation and uncertainty, these six numbers alone also cannot establish a defensible statistical alarm rule.

The case distinguishes specification decisions from process observation. A specification addresses a defined suitability question. Trending seeks indications that the underlying state is changing. They may trigger different responses and should not be confused. Prematurely widening limits would not explain the suspected cause. Equally, ignoring the numbers solely because none exceeds a limit would be inappropriate.

Monitoring therefore defines metric, source, period, context and expected response together. The quantity must suit the question: concentration, total amount and activity can show different developments. Changing working volume can influence concentration without changing total product proportionally. The correct scientific reference turns a numerical sequence into useful quality information.

Analysis and PAT must support the quality decision

A quality attribute can be assessed only as reliably as sampling and analysis permit. The method must examine the correct material within a suitable range. Precision alone is insufficient when preparation causes systematic loss or another constituent changes the signal. Define measurand, matrix, range, relevant interference and sample state together. One method may suit broad development trends while being insufficient for a narrow release decision.

PAT can provide earlier and denser information when measurement connects with process understanding. A spectrum initially measures optical properties; an estimated product or substrate quantity needs a suitable model. Good fitting to known batches does not automatically ensure reliable performance with new medium or another scale. Reference data, independent assessment and application-range monitoring remain decisive.

Hypothetical case: A model predicts a target concentration close to the decision threshold. Independent reference analysis finds a discrepancy occurring mainly after a material change. Rather than examining only an overall model-performance score, investigate new material variability, spectra, sample timing, reference method and whether the model operates beyond its previous range. Retrospectively shifting all predictions can reduce visible error while leaving the underlying uncertainty unresolved.

Automatic decisions also need a defined response to invalid information. This might involve another reference measurement, a bounded interim strategy or interruption of the relevant action. The appropriate response follows the product and process task. High measurement frequency does not mean every prediction supports unrestricted release. Real time describes availability rather than establishing fitness for purpose.

The model-reference combination should remain understandable throughout its lifecycle. Document training data, version, permitted inputs, changes and decision rule together. This allows later reconstruction of the basis for action. Analysis then becomes an assessable part of the quality strategy rather than another unexplained number in the batch record.

Data integrity covers the complete information route

Data integrity means keeping data complete, consistent and accurate while preserving understandable origin and meaning. In a bioprocess this includes sensor values, recipes, interventions, samples, laboratory results and analyses. An exported block of numbers without units, status or time reference is not a complete process description. Even an attractive PDF plot can omit information needed to examine the original measurement later.

Useful questions include: who or which system generated the value, when did it originate, which processing was applied, was it changed or recalculated, and which original information remains? These questions already help research laboratories. For medicines, the FDA guidance on data integrity and drug CGMP explains metadata, traceable changes and complete records. Purchasing software alone does not satisfy its requirements.

An audit trail records change, but its existence does not answer every question. Establish which events it captures, how they relate to a record and how relevant entries are reviewed. Shared accounts can undermine attribution. Administrators may have capabilities different from process users. Roles, actual permissions and working practices must therefore agree rather than remaining separate only on paper.

Suitable checks demonstrate backup and restoration. A successful-backup message does not automatically establish that data and metadata can be restored and read. Long-term availability also requires appropriate formats, access and ownership. Software changes must preserve the meaning of older records. Altered units or time bases can damage information without technically losing a file.

Keep raw records, processing steps and results connected during analysis. Explain exclusions and repeats. An unfavourable original value does not disappear because a subsequent repeat looks better. This practice protects against unintentional errors and inappropriate result selection while strengthening scientific conclusions.

Connect changes, deviations and closing decisions

A change can look minor while having broad scientific effects. New tubing may influence delivery, cleaning and gas permeability. Longer cultivation increases exposure, sampling consumption and hold times. A software update can alter calculation, display or export. Before implementation, identify affected requirements, risks and previous evidence. Assessment depth follows effect rather than the price of the replaced component.

Distinguish deviation from planned change. A deviation describes a state that did not occur as intended and needs investigation and assessment. Correction addresses the immediate situation; a lasting action addresses its mechanism. If a sample was misidentified, another sample can describe the present condition but cannot automatically reconstruct the past event. Assess consequences for data and material separately.

A meaningful closure connects requirements with evidence actually available. It states what was assessed, which deviations occurred, how they were evaluated and which use is now justified. Conditional or limited use must remain recognisable. Product release, equipment release for an experiment and maintenance closure are different decisions. Combining them into one green symbol loses important meaning.

Periodic review examines developments as well as individual failures: more recalibrations, repeated cleaning, increasing sample delay or gradual movement of an attribute. Such indications can reveal change before a hard limit is exceeded. Response should suit the evidence. Investigate a trend rather than ignoring it because values remain within limits or declaring it a confirmed product defect without context.

For regulated tasks, check current editions and product-specific scope through the official EudraLex overview, which distinguishes relevant areas and links the applicable documents. Practical quality work remains centred on connecting each significant statement with suitable evidence and an understandable decision.

Frequently asked questions

Is a bioreactor with IQ/OQ documentation automatically GMP-compliant?

No. Those records concern a defined assessment subject. Suitability of the complete user process additionally depends on intended use, assembly, methods, organisation and appropriate evidence.

Does every research experiment require formal process validation?

The necessary framework follows intended use and applicable requirements. Sound scientific methods, traceable data and clear limits to conclusions matter outside formal validation as well.

Terms in this chapter

34 terms explained in the glossary