ICH Q7 clause 8: Production and in-process controls

Weighing and measuring operations are critical quality control points. Issues include: wrong material charged (identity confusion), wrong quantity (weighing error), contamination during weighing (from environment or weighing equipment), weighing equipment inaccuracy. Controls: calibrated balances appropriate for the quantity range, dedicated weighing area with environmental controls, clear container identification, independent verification of weighed quantities.

This is a key error prevention control. Before adding a material to the batch, production verifies: (1) the material label matches what the batch record requires, (2) the quantity weighed matches the batch record requirement, (3) the material is within its retest/expiry period, (4) the material is released (not quarantined). For critical operations, an independent witness or electronic verification provides additional assurance.

Yield is a sensitive indicator of process problems. Significantly low yields may indicate lost material (spills, equipment hangups), incomplete reactions, or wrong reagent additions. High yields may indicate under-reaction products or contamination. Expected yield ranges (typically 90-110% for well-established processes) should be set based on historical data. Deviations trigger investigation, especially at critical steps where yield correlates with quality.

CPPs (temperature, pH, pressure, agitation, addition rates, reaction time) must be monitored continuously or at frequent intervals during production. Modern practice uses automated DCS/PLC systems that log parameters continuously and alarm on excursions. Paper-based monitoring at defined intervals is acceptable but risks missing transient excursions. All CPP data becomes part of the batch record.

In-process testing (IPC) verifies that the batch is progressing correctly before committing to expensive downstream steps. Example: testing for reaction completion before workup, testing particle size before drying, testing moisture before final packaging. Each IPC has: sampling point, test method, acceptance criteria, disposition procedure if out-of-spec. IPC procedures must be QA-approved.

Production deviations are unplanned departures from approved procedures or specifications. Every deviation must be documented and evaluated, with critical deviations investigated thoroughly. The investigation scope must extend to 'associated batches' — other batches potentially affected by the same cause (same equipment, same material lot, same time period). Root cause analysis is mandatory, and CAPA must prevent recurrence.

Planned process changes must go through change control, not informally. Even minor changes can have unexpected effects (e.g., supplier substitution affecting impurity profile). The change control process must: document the change, assess impact on quality, require QA approval, implement the change with verification, and evaluate results. Significant changes trigger revalidation of affected process steps.

Witnessed operations prevent single-person errors on critical steps. Typical critical steps requiring witnessing: charging of starting materials, addition of catalysts/critical reagents, final product isolation, sampling for release testing. Electronic verification (barcode confirmation, automated metering) can replace physical witnessing. The witness signature or electronic verification must be in the batch record at the specific step.

Time limits constrain process steps to prevent quality issues from extended durations (e.g., extended reaction times causing over-reaction, extended hold times allowing degradation, extended drying causing loss of a volatile component). Not all steps need time limits — parameter-based control may be more appropriate (e.g., reaction until IPC shows completion). Where time limits exist, they must be enforced and deviations investigated.

Time limit deviations follow the standard deviation process but with specific focus on time-related degradation risks. Example: if a time limit exists for cooling after a reaction, and cooling takes too long, the investigation must assess whether the extended time at elevated temperature caused degradation, additional impurity formation, or crystal form changes. Investigation may require additional testing beyond routine release testing.

In-process monitoring procedures must be specific and documented. They define what to measure, how often, how much sample, what results are acceptable, and what to do if results fail. Failure handling is particularly important — there must be a clear decision tree: halt process, additional testing, investigate, adjust, continue, or reject. These decisions must be documented and authorized by QA for critical steps.

IPC selection must target parameters that affect downstream quality. Common IPCs: pH of buffered solutions, reaction completion (HPLC or TLC), moisture content (Karl Fischer), particle size (laser diffraction), appearance (visual), density (specific gravity). Acceptance criteria should be derived from process understanding and validation data — not arbitrary. Tighter limits prevent downstream issues; looser limits risk quality impact.

QA approval ensures that production doesn't unilaterally set its own acceptance criteria. QA's scientific review is independent and product-quality focused. Any changes to IPCs (tightening, loosening, adding, removing, method changes) require QA re-approval through change control. This prevents drift in controls that could compromise quality.

Out-of-specification in-process materials must be segregated to prevent accidental use. Quarantine procedures apply the same as for raw materials. Disposition options include: reprocessing (repeating established steps), reworking (using a different step), destruction, or use as a different product. Each disposition requires QA approval and documentation per Section 14.

Some process steps can't be visually assessed for completion — chemical testing is required. Common examples: HPLC to confirm reaction completion, TLC for quick spot checks, NMR for functional group transformations, chiral HPLC for stereoselectivity. Testing must occur BEFORE the next step to avoid wasting subsequent steps on incomplete material. Results are documented in batch records.

Residual materials from earlier steps can contaminate intermediates and APIs. Controls include: washing steps, drying steps, crystallization for purification, distillation for solvent removal, chromatography for impurity removal. IPCs should monitor residuals that could affect final quality, especially toxic components (genotoxics per ICH M7, heavy metals per ICH Q3D). Limits are set based on toxicological assessment.

Same-product cleaning between batches has different requirements than product-changeover cleaning. For dedicated equipment making the same API, some residue buildup is acceptable if it doesn't accumulate to levels affecting quality. Between-batch cleaning may be simpler (rinse + visual) with periodic comprehensive cleaning. Validation demonstrates that the approach maintains quality across a campaign.

This clause distinguishes two concepts often confused: (1) combining fractions of a single batch into a homogeneous whole — this is normal production, not blending, and (2) combining separate batches into a larger lot — this is blending and requires specific controls. The definitions are important because they trigger different documentation and validation requirements. Most within-batch combining is acceptable as normal production.

This is an absolute prohibition — using blending to 'average out' failed batches is a serious compliance violation. Every batch entering a blend must individually meet all applicable specifications. Blending can be used to: create larger lots for convenience, homogenize subtle batch-to-batch variation within specification, or combine batches for downstream processing. It cannot be used to rescue failed batches.

Blending validation proves that the physical mixing actually produces a uniform result. Validation tests include: sampling from multiple locations within the blended material, analyzing for key quality attributes, and demonstrating that results are similar across all sample locations. Validation also determines acceptable blend sizes (too large may not mix well, too small loses the benefit of blending) and assesses blend stability.

Blended batches need their own batch records showing the composition. Key information: identity and batch number of each component, quantity of each, date and time of blending, equipment used, blending operator, resulting blend batch number, and any IPC results on the blend. This creates a clear audit trail from blend back to individual components, enabling impact assessment if problems arise later.

Blending is often used to achieve uniform physical properties, so testing must verify success. Common tests: particle size distribution (laser diffraction), bulk/tap density (for flow properties), moisture content (Karl Fischer), homogeneity via content uniformity (multi-point HPLC assay). Tests selected based on which physical properties matter for the API's intended use.

This is a conservative rule to protect patients. Since the blend contains material as old as its oldest component, the expiry must match that oldest material. You cannot 'reset' the clock by blending old material with new — the old material is still present and still aging. This rule sometimes makes blending unattractive from a shelf-life perspective, which is by design.

Blending typically averages impurity levels across components, but unexpected interactions can occur — e.g., if one component has a high level of a reactive impurity, it could react with components of another batch during blending. Testing the blended material's impurity profile confirms no unexpected changes. For blends with diverse component sources, comparison to individual component profiles is particularly important.

Blending doesn't reduce testing obligations. The blend must meet all specifications that individual components had to meet. This prevents using blending as a way to skip testing. Some efficiencies are possible with validation — e.g., reduced testing if statistical sampling from the blend is equivalent to testing each component. But the default is full testing of the blend as a new batch.

Unlike cross-product carryover, same-product carryover is sometimes acceptable and even beneficial (avoiding material waste). However, controls must ensure that: (1) degradation products don't accumulate beyond limits, (2) microbial contamination doesn't grow, (3) impurity profile remains within specifications. This requires campaign length validation and periodic comprehensive cleaning. Not all same-product carryover is acceptable — depends on stability and growth characteristics.

Contamination can come from many sources: personnel, equipment, environment, materials, cross-contamination. Controls must address each source. Risk assessment identifies highest risks and prioritizes controls. Typical controls: dedicated facilities for high-risk products (e.g., beta-lactams), air handling for cross-contamination prevention, gowning for personnel contamination, closed systems for sensitive products.

Monitoring confirms that contamination controls are effective. Typical monitoring includes: environmental microbial monitoring (air samples, surface swabs, personnel gloves), particulate monitoring for clean areas, cleaning verification at product changeovers, periodic testing of finished APIs for expected contaminants (e.g., cleaning agents). Monitoring frequency is risk-based — more frequent for high-risk, less for low-risk.