Ask a validation lead when Process Performance Qualification succeeds or fails, and most will point to the campaign itself: the three batches, the enhanced sampling, the acceptance criteria. That is where failure becomes visible. It rarely originates there. By the time batch one starts, the outcome is already substantially determined by decisions made weeks or months earlier, in characterization studies, method validation reports, and a CDMO slot booked before technology transfer had actually closed. The pre-execution readiness review is not where readiness gets decided. It is where an earlier decision gets confirmed or exposed. The PPQ program itself sits inside a broader validation master plan, which is where the phase-appropriate rigor and batch-scope logic behind these prerequisites should already have been set, not improvised at the readiness meeting.
This matters because PPQ is expensive to get wrong in a way most validation activities are not. A failed characterization study gets repeated in a lab. A failed PPQ batch, especially one with a narrow CDMO slot and a long-lead-time raw material, can mean months of delay and an uncomfortable conversation with regulatory affairs about a filing timeline. The fix is not better execution. It is a readiness assessment honest about what “ready” means, checked against evidence rather than intention, before the protocol is approved.
This article works through the prerequisites that decide PPQ success before the first batch: process characterization, locked CPPs and CQAs, hold-time strategy, batch size and bracketing, a statistician-built batch-count and sampling rationale, a stated acceptance-criteria and concurrent-release basis, a pre-agreed deviation criticality classification, validated and transferred analytical methods, a frozen batch record, the CDMO logistics that rarely appear on a checklist but decide the date as much as anything else, aseptic-process readiness beyond a point-in-time facility-fit assessment for sterile products, personnel and training readiness specific to the frozen batch record version, raw material and single-source supply risk beyond simple inventory, and the distinction between a deviation and a protocol amendment once the campaign is running.
The readiness question, asked the right way
The wrong question is whether the readiness review passed. A readiness review is a meeting; it can be scheduled to pass. The right question, the one an inspector is implicitly asking when pulling a PPQ report, is how the organization knew, with evidence, that it was ready before committing commercial-representative material to three or more batches. That question only has an answer if every prerequisite below produces documented, dated evidence that existed before protocol approval, not a verbal assurance collected in the readiness meeting itself.
Readiness reviews under schedule pressure tend to converge on optimism. A checklist item gets marked complete because it is expected to be complete by the time it matters, not because it is complete today. That gap, between “will be done” and “is done,” is where PPQ campaigns quietly set themselves up to fail.
Process characterization: “we ran it” is not “we understand it”
Stage 1 process design is supposed to leave PPQ with a control strategy: proven acceptable ranges for critical process parameters, an understanding of which parameters drive which critical quality attributes, and a defined operating range. The test for whether that work is real: can every CPP range in the protocol be traced to a specific study with data behind it, rather than to an engineering run that happened to work.
A small molecule process built on a design-of-experiments study usually clears this bar; the DoE structure forces CPP-to-CQA relationships into the open. Biologics, oligonucleotide, and cell and gene therapy processes fall short more often, because raw material and cell-source variability can dominate over parameter settings in ways a tabletop DoE does not fully capture. For an autologous cell therapy, the honest version of this question is not “did we span three raw material lots,” it is whether the process is robust to donor-to-donor variability or merely untested against it. If the latter, PPQ readiness is not there yet.
The decision rule: if a CPP’s proven acceptable range changed in the months immediately before the PPQ protocol was drafted, and the change was not driven by new characterization data, Stage 1 is not closed. Write the protocol against the range that has data behind it, not the range that feels closer to how the process runs today.
CPPs and CQAs have to be locked, not still moving
Locked means the control strategy, CPP ranges, and CQA acceptance criteria have been stable and approved for a defined period, with no open investigations pending against them. It does not mean the numbers can never be revisited, only that they are not being revisited while the protocol is being authored.
A common false-ready pattern: a CPP range gets tightened in the protocol relative to what characterization showed, because a recent engineering run ran rough at the wider setting and nobody wants to explain that at PPQ. This narrows the target with no data to justify why a wider, characterization-supported range is suddenly unacceptable. Either the characterization data was wrong, meaning Stage 1 needs formal revisiting, or the tightening is cosmetic and will not survive scrutiny of the traceability between risk assessment, characterization report, and protocol.
Before protocol approval, for every CPP and CQA acceptance criterion, can someone point to the specific Stage 1 source, and has it been stable since approval. If the answer requires searching for a justification rather than citing one, the parameter is not locked.
Hold-time strategy needs its own data, not an assumption
Hold times, in-process material sitting between unit operations, a bulk drug substance held before fill, a cell suspension held between processing steps, are a common source of PPQ surprises because they are frequently assumed rather than studied. A hold-time limit of, say, up to 24 hours at 2 to 8 degrees C needs a dedicated study at the worst case, not a number inherited from a platform process or general practice.
This is a particular trap when working across modalities, where a hold-time assumption conservative for one product gets carried into a related but not identical one. RNA-based drug substances and cell therapy intermediates are frequently far more sensitive to hold conditions than the small molecule or antibody intermediates the assumption may have originated from. Before PPQ, every hold step needs an answer to three questions: the maximum hold time and condition the commercial process will use, whether a study exists at that worst case, and whether it showed the CQAs held.
The decision rule: if a hold step exists in the batch record with no dedicated study behind its limit, that is an open prerequisite, not a formality to close later. PPQ batches run under an unstudied hold-time assumption are not qualifying the commercial process; they are qualifying a guess.
Batch size: commercially representative, not a scaled rehearsal
PPQ has to run at the scale, and on the equipment train, that will be used commercially, or on a scale-down model with a data-backed, documented justification for why it represents commercial scale. A batch size chosen because it fits equipment currently available, without a firm decision that this is genuinely the commercial batch size, is a common and expensive mistake, one that surfaces as a second qualification campaign once commercial demand outgrows it.
This decision belongs earlier than PPQ, but is worth restating as a prerequisite because it drifts quietly under manufacturing-network pressure. The test: does the protocol’s batch size match the commercial supply plan, and if there is a scale difference, is there a documented rationale with supporting data, not just an assertion that the process scales linearly.
Where a single process spans a family of strengths, fill volumes, or presentations on the same line, bracketing keeps the qualification honest without re-running PPQ on every member: qualify the smallest and largest directly, and extend the conclusion between them only if the CPPs and CQAs scale predictably, with data confirming the assumption rather than just asserting it. A bracket with no data anywhere between its ends is a documented risk, not a closed prerequisite.
Batch count and sampling: what a statistician’s conversation actually decides
“How many PPQ batches” is the question a reviewer asks first, and there is no regulatory requirement for three. FDA’s 2011 guidance deliberately does not fix a number; it ties batch count to Stage 1 process knowledge, product and patient risk, and the variability the campaign needs to demonstrate it can absorb, and EU Annex 15 treats it the same way: a workable default for a simple process, never a substitute for the justification underneath it. A process characterized under a design-of-experiments study, with CPP ranges tied to data, supports that default; a process with multiple raw material lots, donor-to-donor variability, resin lifetime effects, or a high-variability orthogonal CQA method usually needs more, or three batches plus a plan carrying the unresolved variability into CPV. “Three batches per industry standard,” with no risk argument behind it, answers an easier question than the one being asked; the number has to be a conclusion, not an input chosen first and justified afterward. FDA’s May 2026 guidance on CMC flexibilities for cell and gene therapy BLAs makes this explicit for products where a limited number of qualified lots exist at commercial scale: a scientifically justified batch count below the traditional norm is acceptable when the rationale is written down and the residual risk is covered elsewhere, the same discipline this section already describes. See writing the CMC validation summary a reviewer can follow for how that rationale actually gets stated in a filing, and the paired risk-based batch-count justification template for a working document that structures it.
The sampling plan turns on a smaller version of the same question for every attribute: attribute plan or tolerance interval. An attribute plan, the logic behind ANSI/ASQ Z1.4, asks a go/no-go question, does this unit pass or fail a defined criterion, and fits defect counts and other pass/fail attributes. A tolerance-interval approach, the logic behind ANSI/ASQ Z1.9’s variables plans, asks with a stated confidence, say 95 percent, what proportion of the population, say 99 percent, falls within specification; it fits continuous measurements, assay percent, content uniformity, fill weight, where closeness to the limit carries real information. The choice is not a style preference: an attribute plan on a continuous CQA throws away that information, and a tolerance-interval framework on a genuinely binary attribute invents a distribution that is not there.
The decision rule: if the batch count cannot be traced to a written risk assessment, or a sampling point’s sample size cannot be traced to a stated confidence and coverage under a named statistical basis, the prerequisite is not closed, regardless of how experienced the plan’s author is.
A worked example: sizing a fill-weight sampling plan
Take a fill-weight CQA on a lyophilized vial line, a continuous measurement with a specification of 1.00 mL, plus or minus 0.05 mL. Characterization-run data shows the fill process centered near target with a known estimated standard deviation. The statistician chooses a two-sided normal tolerance interval, the logic behind ANSI/ASQ Z1.9’s variables sampling plans, at a stated confidence, commonly 95 percent, and coverage, commonly 99 percent, then pulls the sample size and its matching k-factor from the current published table for that confidence and coverage pair, rather than assuming a round number like 10 or 30 carries the same statistical weight regardless of the underlying variability. The resulting interval, the observed standard deviation multiplied by the table’s k-factor, has to sit inside the 0.05 mL specification with real margin before the sample size is accepted; if it does not clear the specification, the plan needs a larger n pulled from the table, not a looser reading of the result. The exact k-factor for a given n, confidence, and coverage combination is a published table value, never a figure to estimate from memory; pull it from the current ANSI/ASQ Z1.9 or ISO 16269-6 table for the actual n under consideration, and cite the table and edition used in the protocol.
The decision rule underneath the example: if nobody can point to the confidence, the coverage, and the specific table both numbers came from, the sample size is a guess wearing statistical language, not a tolerance interval.
PPQ acceptance criteria: release specification, or something tighter
A readiness review that closes every prerequisite above still leaves one question unanswered if nobody asks it directly: what does the PPQ protocol actually require a batch to hit. The default and most common answer is the commercial release specification and the routine in-process control limits, tested with the enhanced sampling plan from the batch-count conversation above, not a separate, tighter PPQ-only bar invented for the occasion. A batch that passes its own release specification and every in-process control has qualified the process at the level the process will actually be held to for the rest of its commercial life; writing a stricter PPQ-only acceptance criterion the routine process will never be checked against again proves something more demanding than commercial manufacture requires, which is not the same thing as proving the process is ready for commercial manufacture.
That said, the release specification is a floor, not automatically the whole answer. Where Stage 1 characterization data supports it, some programs set PPQ in-process action limits tighter than the eventual routine limits, deliberately, while process capability is still being established from a small batch count. The logic: a PPQ campaign with only three to six batches behind it has not generated enough data to defend a limit at the same statistical confidence a mature CPV dataset eventually will, so a tighter provisional in-process limit gives the campaign an early-warning threshold without changing what the batch actually has to hit at release. Whichever approach is chosen, state it explicitly in the protocol and connect it to a rationale, either the release specification because Stage 1 data already demonstrates capability at that limit, or tighter provisional in-process limits during PPQ transitioning to routine limits once CPV data supports it. Never leave the criteria unstated as if using the release specification were too obvious to need a decision.
| Approach | What the batch is actually held to | When it fits |
|---|---|---|
| Release specification plus routine in-process limits, enhanced sampling only | The same numeric limits the commercial process runs against for its full lifecycle | Stage 1 characterization data already demonstrates capability at the release limit; enhanced sampling exists to gather statistical evidence, not to raise the bar |
| Tighter provisional in-process action limits during PPQ | A stricter early-warning threshold than the eventual routine limit, stated explicitly as provisional | Process capability is not yet demonstrated at commercial batch count; the campaign needs an early flag before a marginal trend becomes a release-spec failure |
| A PPQ-specific acceptance criterion with no stated rationale or transition plan | Whatever was written, untethered to either the release specification or a capability argument | Never; this is the false-ready pattern, not a legitimate approach |
Concurrent release: can a batch ship before the full PPQ report is approved
A related question surfaces once the campaign is running rather than before it, but belongs in the readiness conversation because the answer has to be decided in advance, not improvised once batch three is filling: can PPQ batches that individually meet the commercial release specification actually ship before the complete, three-or-more-batch PPQ report is written, reviewed, and approved. The traditional, conservative answer is no, hold everything until the full report closes the campaign. FDA’s May 2026 guidance on CMC flexibilities for cell and gene therapy BLAs states this more flexibly for that modality class: it allows concurrent release of PPQ batches for commercial use after BLA approval, provided each batch independently meets the commercial release specification, an accommodation framed around the reality that cell and gene therapy programs often have a limited number of qualified commercial-scale lots and a patient population that cannot always absorb a long hold. The same guidance permits the batch-count and analytical-method-validation strategy behind PPQ to rely on fewer lots, including a single representative lot in some circumstances, when the rationale is scientifically supported and documented, tying this section directly to the batch-count discussion above rather than leaving it as a separate question.
Concurrent release is a decision that belongs on the readiness review agenda before batch one, not a scramble once a batch is sitting in the warehouse meeting spec while the report is still in draft. If the program intends to rely on it, the protocol should say so explicitly, name the guidance basis where the product qualifies for the cell and gene therapy flexibility described above, and state plainly what triggers a hold, a failing batch elsewhere in the campaign, an unresolved deviation, an unexpected trend, so that meeting spec on one batch is never read as clearance to ignore what a sibling batch in the same campaign is showing.
Deciding deviation criticality before the campaign starts
The concurrent-release logic above turns on one phrase doing a lot of work: “no open deviation, trend, or failing sibling batch anywhere in the campaign.” That phrase only functions as a real gate if the campaign has already agreed, before batch one, what counts as a deviation serious enough to trigger a hold and what does not. A readiness review that reaches batch one with “any deviation” left undefined is not actually protected by the flow diagram above; it has deferred the hardest part of the decision to whichever manager is on call when the first deviation of the campaign happens, under exactly the schedule pressure this article’s other prerequisites are built to resist.
The fix is a pre-agreed criticality classification, written into the protocol or a linked deviation-handling plan before batch one, not improvised deviation by deviation. A workable structure sorts deviations into a small number of classes tied to what the deviation could plausibly affect, not to how alarming it feels in the moment.
| Class | What it means | Effect on batch disposition |
|---|---|---|
| Critical | Directly implicates a CQA, the sterility or potency assurance chain, or a condition the acceptance criteria assume held | Holds the affected batch and, per the flow diagram above, every batch in the campaign until investigated |
| Major | Affects a documented, non-CQA process parameter or a control excursion outside its normal range, with no CQA linkage established yet | Holds the affected batch pending investigation; does not automatically hold sibling batches unless the investigation later implicates a shared cause |
| Minor | A documentation or procedural gap with no plausible product-quality impact, for example a late timestamp on an unaffected step | Does not by itself hold any batch; logged and trended |
The decision rule that keeps this from becoming a loophole: the classification criteria have to be written and approved before batch one, and the person classifying a real deviation during the campaign has to be able to point to which written criterion the deviation matches, not argue the class after the fact based on how much the team wants to avoid a hold. A “major” classification invented in the moment to avoid tripping the campaign-wide hold is the same false-ready pattern this article names everywhere else, just relocated from before the campaign to during it. Where the classification is genuinely uncertain in real time, the conservative default is the higher class, not the lower one, until the investigation resolves it.
This connects directly to the roles table below: Quality Assurance owns the classification decision for any deviation that arises during the campaign, and that ownership is easier to exercise cleanly when the criteria were written and approved in advance rather than negotiated under the pressure of a batch already sitting in the warehouse. A ready-to-use plan for pre-classifying deviation criticality before a campaign starts is at PPQ deviation criticality pre-classification plan.
Analytical methods: validated and transferred, not “almost done”
A PPQ batch generates data only as trustworthy as the methods used to produce it. Method validation, under a framework like ICH Q2 or, for compendial procedures, USP <1225>, has to be complete and approved before PPQ data collection begins, and if testing happens at a lab different from where the method was developed, method transfer has to be closed too. “Validation in progress, expected to complete before results are needed” is not a closed prerequisite; it is a bet that the method will validate as expected, made after commercial-representative batches are already committed.
This carries extra weight for orthogonal methods common to biologics, oligonucleotide, and cell and gene therapy products: charge variant and glycosylation assays, potency assays with inherent biological variability, identity methods for cell products, where robustness and precision are less forgiving than a compendial small molecule assay. If an investigation later needs to rule out an analytical cause, and method variability was never characterized because validation closed weeks before the batches ran, there is no baseline to work from.
Potency assay design specifically is getting a dedicated reference worth knowing about, even though it is still a draft. FDA issued a draft guidance on August 20, 2026, “Potency Assessment of Active Immunotherapy Products” (docket FDA-2026-D-8561, comments due November 18, 2026), supplementing FDA’s December 2023 draft guidance on potency assurance for cellular and gene therapy products with assay design and validation recommendations specific to active immunotherapy products, ACTIMPs for short: peptide and protein based, vectored, personalized, and cell based products among them. The recommendations are organized by product type because the potency challenge differs by type; a personalized vectored product raises different assay design and reagent availability questions than a non-personalized cell based product, and a team building the potency assay behind a PPQ protocol for one of these modalities gets more use out of reading the recommendation for its own product category than a generic potency chapter. It is still a draft, not yet final, so treat it as a direction to design toward rather than a settled requirement to cite as binding. A worked structure for laying out a multi-assay potency strategy against the biological claim each piece supports is at potency assurance triangulation matrix.
Batch records frozen before batch one
The PPQ protocol drives the enhanced sampling layered on manufacturing, but the batch runs on the master batch record, itself a controlled document under 21 CFR 211.188, and that record has to be approved, effective, and unchanged for the campaign. A record still open for markup, absorbing last-minute edits from an engineering run two weeks before PPQ, is not frozen, and executing against it undermines the traceability between what was reviewed and what ran.
The practical check: confirm the batch record version referenced in the protocol is the exact version manufacturing will use for batch one. If manufacturing plans to use a newer, unapproved draft because it fixes an error, that signals the record was not ready, and the fix needs change control before batch one, not during it.
CDMO slot logistics: the readiness gate nobody puts on the checklist
For organizations manufacturing at a contract development and manufacturing organization, the default for most clinical-stage biotechs, slot logistics are a readiness prerequisite in their own right, and one of the most underestimated. A CDMO slot is booked months in advance, against a calendar that does not wait for transfer, comparability, or material qualification to finish.
The trap is specific and recurring: the slot gets booked to hold the date, transfer and comparability work continue in parallel, and as the date approaches, the sunk cost of losing it outweighs the discipline of confirming the prerequisites are closed. PPQ then runs on a facility’s schedule, not the process’s.
Confirm four things before committing to a CDMO date: the transfer report is closed and approved, not “substantially complete”; comparability between sites is established with data, not assumed from the process looking identical on paper; raw materials, including any with long lead times or a single source, are released with margin against a lot failure; and facility fit, equipment train, environmental classification, utilities, was assessed and closed, not inferred from the CDMO making similar products elsewhere. Most of this evidence should already exist coming out of the CDMO technical assessment that selected the site in the first place; a slot readiness review that cannot point back to that assessment’s findings is starting from scratch on questions that should have been answered months earlier. A person-in-plant presence during the run-up weeks is often what actually catches a facility-fit or material-release gap early enough to act on it, rather than a status update relayed secondhand.
As the date approaches, the decision narrows to three states, and only one of them is actually safe to run on.
There is no fourth option where the gap simply goes unmentioned. A readiness review that reaches the date with an open check and produces neither a moved slot nor a documented risk acceptance has not actually made a decision; it has let the calendar make one by default.
Facility and aseptic-process readiness for sterile products
“Facility fit was assessed and closed” in the CDMO logistics check above covers equipment train, environmental classification, and utilities as a point-in-time assessment. For a sterile or aseptically processed product, that is necessary but not sufficient, because a cleanroom’s classification and an aseptic process’s demonstrated control are two different claims, and the second one has an expiration date the facility-fit assessment does not automatically track.
Aseptic process simulation, a media fill using a microbiological growth medium run through the actual fill line and process steps in place of product, is the direct evidence that the line’s aseptic technique and environment actually exclude microbial contamination under production-representative conditions. EU GMP Annex 1’s 2022 revision expects aseptic process simulation at a minimum of twice yearly for each aseptic process, line, and shift configuration in routine use, with initial qualification typically built on multiple consecutive successful runs and a fresh simulation required after a change that could affect aseptic assurance: a new fill head, a line reconfiguration, a change in gowning practice, or an extended production interruption. A PPQ batch run on a line whose media fill has lapsed past its own qualified interval is not evidence of a validated aseptic process; it is evidence collected on a line whose aseptic status is itself unconfirmed at the moment PPQ ran.
PDA Technical Report No. 22, Process Simulation for Aseptically Filled Products, is the reference framework behind how that media fill itself gets designed and interpreted, covering worst-case process duration, intervention selection, container-closure format, and fill speed, the specifics a media fill protocol has to defend on their own merits rather than copy from a different product’s prior study. PDA issued a full revision in October 2025, the first since the original 2011 edition, written to work through the design and interpretation questions EU GMP Annex 1’s 2022 revision raised for aseptic process simulation, including how a site’s own contamination control strategy should shape which worst-case conditions a given line’s media fill actually needs to simulate. A readiness review citing the 2011 edition by number is citing a document PDA replaced outright, not lightly revised; confirm which edition the facility’s media fill program references before treating that program’s design rationale as settled.
The practical readiness check, parallel to every other prerequisite named above: before protocol approval, confirm the specific line and shift PPQ will run on has a current, passing media fill within its own qualified interval, that environmental monitoring trends for that line show no unresolved excursion or adverse trend, and that operator gowning qualification for everyone scheduled on the PPQ batches is current. Where a CDMO is the manufacturing site, this evidence sits inside the CDMO’s own contamination control strategy and environmental monitoring program rather than the sponsor’s files, so confirming it is current has to be an explicit, documented request during CDMO slot logistics, not an assumption that a qualified facility stays qualified indefinitely between assessments.
The decision rule: if the media fill covering PPQ’s specific line, shift, and configuration is not both current and passing, or if environmental monitoring shows an open, unresolved adverse trend on that line, that is an open prerequisite regardless of how long ago the facility-fit assessment itself was closed. A stale media fill discovered after batch one starts converts a schedule problem into a data-integrity one: every batch run since the interval lapsed now carries an unanswered question about the aseptic state of the line that made it. The PPQ readiness review checklist scores this alongside every other prerequisite above as its own line item, not folded silently into the CDMO facility-fit row.
Personnel and training readiness: qualified on this batch record version, not just the role
A readiness review that confirms “operators are qualified for this unit operation” against a job-role qualification matrix has answered an easier question than the one PPQ actually needs answered. Role-level qualification, a qualified cell culture operator, a qualified aseptic fill operator, confirms someone can perform the general unit operation. It does not by itself confirm that operator is current on whatever is specific to this batch record’s approved, frozen version: a new in-process control step added for PPQ’s enhanced sampling, a tightened hold-time check, a revised sampling instruction that did not exist the last time that operator trained on the process.
This gap shows up in a specific, recurring way. Training happens against a draft batch record during rehearsal runs, which is reasonable and often unavoidable given the schedule. The final, approved version then introduces a last-minute change, commonly an additional in-process sample point added to support the enhanced sampling plan from the batch-count conversation above, and that change never gets folded back into a documented retraining before batch one. The operators scheduled on batch one are current on the process in general and are executing a specific procedural step nobody has confirmed, with a dated record, they were ever trained on.
The test, parallel to every other prerequisite above: for every operator and analyst scheduled on PPQ batch one, does a dated training record exist confirming both the general unit-operation qualification and specific training on the exact approved batch record version PPQ will run under, with no gap between the version trained on and the version executed. This applies as much to QC analysts running the enhanced sampling plan’s methods as to manufacturing operators; a method validated and transferred per the analytical-methods prerequisite above still needs the specific analyst executing it on PPQ batches to be confirmed current, not assumed current because the method itself passed validation. For aseptic processes, this sits alongside, and is distinct from, the gowning qualification already named under facility readiness: gowning currency covers one specific aseptic skill, while batch-record-version training covers every procedural step in the record, sterile or not.
The decision rule: if a PPQ-specific procedural element, a new sample point, a tightened in-process limit, a changed hold-time check, was introduced into the batch record after an operator’s last training cycle, that operator’s readiness is an open prerequisite until a dated retraining record exists, regardless of how experienced that operator is on the broader process. A ready-to-use verification record for closing this gap before batch one, with a filled specimen, is at PPQ personnel and training qualification verification record.
Raw material and single-source risk: beyond “released with margin”
The CDMO logistics check above states that raw materials, including any with long lead times or a single source, have to be “released with margin against a lot failure” before committing to a PPQ date. That line is correct and belongs where it sits, but it answers only the inventory question, is there enough released material on hand right now, and not the risk question underneath it: what happens to the PPQ schedule if the lot on hand fails release testing, or if the next lot needed for a later batch in the campaign does.
Closing that gap starts with a specific inventory, not a general awareness that supply risk exists somewhere. For every raw material, resin, buffer component, or biological ancillary reagent feeding the process, identify whether it has one qualified source or more than one, and for every single-source item, state the lead time to obtain an emergency lot from the same source or to qualify a backup source, and whether that lead time fits inside the schedule margin the PPQ campaign actually has. A single-source chromatography resin with a 14-week replacement lead time, committed to a campaign with 10 weeks of schedule margin before the booked CDMO slot, is an open prerequisite the moment that math is done, not a risk to discover if the resin lot on hand happens to fail.
This connects back to process characterization in a way worth stating explicitly rather than leaving as two separate conversations. A raw material’s lot-to-lot variability, a resin’s binding capacity, a biological raw material’s potency, that was characterized across only a single lot during Stage 1 is not just a supply-chain exposure; it is an open characterization gap in its own right, the same pattern named under process characterization above, where raw material and cell-source variability can dominate a biologics, oligonucleotide, or cell and gene therapy process in ways a tabletop study does not fully capture. A material flagged as single-source and also tested across only one lot in characterization carries both risks stacked on each other, and closing the supply-chain question alone does not close the characterization question underneath it.
The fix is a written contingency plan for every material identified as single-source or long-lead and critical to the process, covering the backup source or emergency-lot path, its lead time, and the specific trigger, a failed incoming lot, a disclosed vendor supply disruption, a quality agreement termination, that invokes it. The decision rule: if a single-source or long-lead material has no documented contingency and no schedule margin to simply absorb the lead time if the primary source fails, that is an open prerequisite, regardless of how reliable that supplier’s track record has been to date. A ready-to-use contingency plan for identifying and closing this gap before batch one, with a filled specimen, is at single-source and long-lead raw material contingency plan.
Protocol amendment versus deviation: changing the plan mid-campaign
Every prerequisite above, and the deviation criticality framework in particular, rests on one assumption: that the protocol itself stays fixed while the campaign runs, and what varies, batch to batch, is whether execution matched that fixed plan. A deviation is exactly that, an unplanned departure from an unchanged, approved protocol. A protocol amendment is a different thing entirely: a planned, documented change to the protocol itself, made mid-campaign because something learned during execution shows the original plan, not just one batch’s execution of it, needs to change.
The distinction matters because burying a needed protocol change inside a deviation investigation produces a worse outcome than either sorting it correctly would. A deviation investigation is built to explain why a single batch departed from an unchanged plan and to decide that batch’s disposition; it is not built to carry a decision about whether the plan’s acceptance criteria, sampling point, or in-process limit was itself wrong for every remaining batch in the campaign. Treating a design flaw as a string of individually-investigated deviations, one per batch, leaves the protocol looking unchanged on paper while every batch quietly worked around a known problem, which is precisely the kind of gap a records-focused inspection is built to find. The opposite mistake carries its own risk: treating every deviation as grounds to amend the protocol erodes the protocol’s value as a fixed reference point mid-campaign, and leaves a later reviewer unable to tell what was actually qualified from what was adjusted on the fly.
The test that sorts the two: does the issue call into question whether one batch departed from the plan, or whether the plan itself, as written, would produce the same issue on every batch that follows if nothing changes. If every remaining batch would hit the same wall unless the protocol changes, that is an amendment question, not a deviation to be investigated and closed batch by batch while the same flaw waits for batch four.
A protocol amendment needs the same governance weight as the original protocol, not a lighter one because the campaign is already underway. It needs validation-lead authorship, Quality Assurance approval, and a regulatory affairs review of whatever the amendment does to the filing’s description of the PPQ campaign, before the next batch runs under the changed version. It needs to document the specific data, often a deviation investigation’s root cause, that showed the original approach would not hold, and it needs to state explicitly whether batches already run under the prior version remain valid toward the campaign’s batch count or whether the campaign’s counting restarts from the amendment forward. This also reaches the concurrent-release logic discussed above: a protocol amendment mid-campaign is itself an event that should pause any concurrent-release activity until its impact on already-released batches is assessed, a trigger the deviation criticality plan’s “no open deviation, trend, or failing sibling batch” language does not by itself cover, since an amendment is not a deviation. A ready-to-use justification and approval form for this exact decision, with a filled specimen, is at PPQ protocol amendment justification form.
Roles and responsibilities
Readiness is not one person’s judgment call; it depends on several roles each closing their own piece with evidence, and on Quality Assurance refusing to approve the protocol until every piece is actually closed.
| Role | Readiness accountability |
|---|---|
| Validation lead | Owns the readiness review itself; confirms every prerequisite below has dated evidence, not a verbal assurance, before recommending protocol approval |
| Process development / manufacturing science | Owns process characterization, CPP and CQA locking, and hold-time study data; the first call when a range or limit looks unsupported |
| Statistics | Owns the batch-count risk rationale and the sampling plan’s attribute-versus-tolerance-interval basis; a sampling plan with no named statistician behind it has no defensible basis |
| Analytical / QC | Owns method validation and transfer closure; confirms no PPQ-critical method is still in progress when the protocol is approved |
| Manufacturing / site operations | Owns the batch record’s frozen, effective status and confirms the version cited in the protocol is the version that will actually run |
| Training / qualification owner | Confirms every operator and analyst scheduled on PPQ batches holds a dated training record for the exact approved batch record version, not just the general unit operation |
| Supply chain / procurement | Owns the single-source and long-lead raw material inventory, lead-time assessment, and contingency plan for any material identified as critical to the campaign |
| CDMO relationship owner or person-in-plant lead | Owns the four CDMO logistics checks above; escalates a slot-versus-readiness conflict before it becomes a forced choice, not after |
| Quality Assurance / Microbiology | Owns the deviation criticality classification during the campaign and confirms media fill currency and environmental monitoring trend status for the specific line and shift PPQ will run on |
| Quality Assurance | Approves the protocol only once every prerequisite shows dated evidence; the last check against the optimism bias readiness reviews tend toward under schedule pressure; approves any protocol amendment with the same rigor as the original protocol |
| Regulatory affairs | Confirms the batch-count, acceptance-criteria basis, and any concurrent-release or flexibility-guidance basis matches what the filing will say, so the readiness review and the regulatory narrative do not diverge later; reviews any protocol amendment for its effect on the filing’s description of the campaign |
A readiness scorecard
Each row should have a named owner and dated evidence, reviewed before protocol approval, not assembled the week of the readiness meeting.
| Prerequisite | Objective evidence of “ready” | The false-ready trap |
|---|---|---|
| Process characterization | CPP ranges traceable to a specific study; control strategy closed and approved | ”We ran it several times at scale and nothing went wrong” |
| CPPs and CQAs locked | Stable and approved for a defined period, no open investigations against them | A range quietly tightened in the protocol with no new data behind it |
| Hold-time strategy | Dedicated worst-case studies for every hold step, approved before the protocol | Hold limits carried over from a different product or platform |
| Batch size and bracketing | Matches the commercial supply plan; bracket extremes and at least one midpoint have data | Chosen to fit available equipment; a bracket’s middle assumed, never sampled |
| Batch count and sampling | Count traced to a written risk assessment; every sampling point states its attribute-or-tolerance-interval basis and claimed confidence/coverage | ”Three batches” by convention; a sample size carried over from the last protocol |
| Acceptance criteria and concurrent release | Protocol states the release-specification or provisional-limit basis explicitly, and names the concurrent release strategy and hold triggers if the program intends to use it | Criteria assumed to be the release specification with no stated rationale; concurrent release exercised informally once a batch happens to meet spec |
| Analytical methods | Validation reports approved; method transfer closed and successful | ”In progress, expected to close before results are needed” |
| Batch records | Approved, effective, unchanged for the campaign; matches the protocol’s cited version | A record still open for markup when batch one starts |
| CDMO logistics | Transfer report closed; comparability established with data; materials released with margin | A slot booked and held while transfer work continues in parallel |
| Deviation criticality | Classification criteria written and approved before batch one, with the conservative default stated for genuinely uncertain cases | ”Any deviation” left undefined until the first one forces an ad hoc call |
| Facility and aseptic readiness (sterile products) | Media fill current and passing for the specific line, shift, and configuration; environmental monitoring shows no unresolved adverse trend | A media fill treated as evidence that never expires once the facility-fit assessment closed |
| Personnel and training readiness | Dated training record for every scheduled operator and analyst confirming the exact approved batch record version, not just the general unit operation | Role-level qualification checked; version-specific retraining after a last-minute batch record change never confirmed |
| Raw material and single-source risk | Single-source and long-lead materials inventoried, lead time assessed against schedule margin, written contingency plan for any critical item | ”Released with margin” checked as inventory on hand, with no contingency for what happens if the next lot fails |
How PPQ campaigns actually go sideways
None of these name a specific product or company; they are generic patterns that repeat across modalities.
- A CDMO slot booked before technology transfer closed, then run anyway rather than lose the date, with open transfer deviations absorbed instead of resolved.
- A batch size chosen for equipment convenience rather than confirmed demand, forcing a second qualification campaign at launch.
- A batch count fixed at three before the risk assessment existed, with the assessment built backward to fit a number already tied to a CDMO slot.
- A CPP range tightened after a rough engineering run, with no new characterization data behind the change.
- Hold-time limits assumed from a related process, discovered mid-campaign when a hold step drives a CQA toward its limit.
- Method validation closing during, rather than before, PPQ, leaving no variability baseline when an investigation needs one.
- A readiness checklist signed off with open items footnoted “to close by report issuance,” which reads as schedule pressure overriding governance.
- Concurrent release treated as a default once a batch happens to meet spec, with no protocol-stated basis or hold trigger, discovered only when a later batch in the same campaign fails.
- A deviation classified “major” instead of “critical” in the moment specifically to avoid a campaign-wide hold, with no written criterion cited for the call.
- A media fill that lapsed past its qualified interval on the exact line PPQ ran on, discovered only when an investigation asks for the aseptic evidence behind an already-completed batch.
- Operators trained on a draft batch record during rehearsal, with a last-minute sampling addition in the approved version never folded back into a documented retraining before batch one.
- A single-source resin’s replacement lead time never compared against the campaign’s actual schedule margin, discovered only when the lot on hand fails release testing mid-campaign.
- A design flaw in the protocol investigated as a fresh deviation on every batch instead of being recognized and handled as a protocol amendment, leaving the same flaw unaddressed for the next batch in line.
Common inspection citations behind these findings
None of the rows below describe a specific finding at a specific site; they name the class of citation a records-focused inspection tends to write when one of the prerequisites above was not actually closed before batch one.
| Prerequisite gap | Typical citation basis |
|---|---|
| CPP range tightened in the protocol with no new characterization data behind it | 21 CFR 211.100(a), production and process control procedures not adequate to assure the process performs as represented |
| Hold-time limit carried over from a related process with no dedicated worst-case study | 21 CFR 211.111, time limitations for production not established where appropriate to assure quality |
| Batch record still open for markup when batch one starts | 21 CFR 211.188, batch production and control records not an accurate reproduction of the approved master |
| Method validation closing during, rather than before, PPQ | 21 CFR 211.165(e), test method accuracy, sensitivity, specificity, and reproducibility not established and documented |
| CDMO oversight gaps treated as the contract site’s problem alone | FDA’s “Contract Manufacturing Arrangements for Drugs: Quality Agreements” guidance (2016): the sponsor retains CGMP responsibility regardless of how a quality agreement allocates tasks internally |
| A batch released commercially with no documented basis connecting the decision to a stated concurrent-release strategy | 21 CFR 211.22(a), quality control unit responsibilities not exercised to ensure release decisions rest on an adequate, documented basis |
| A deviation classified in a way that avoided a hold with no written criterion behind the call | 21 CFR 211.192, an unexplained discrepancy not thoroughly investigated, extended here to a discrepancy investigated but classified without a documented basis |
| PPQ batches run on a line whose media fill or environmental monitoring status was not current | 21 CFR 211.113(b), procedures designed to prevent microbiological contamination of drug products purporting to be sterile not established and followed |
| An operator executing a batch-record-specific step with no dated training record for that version | 21 CFR 211.25, personnel required to have the education, training, and experience to perform assigned functions, extended to the specific procedure actually being performed |
| A raw material shortfall forces an unplanned process change mid-campaign, with no contingency plan on file | 21 CFR 211.100(a), production and process control procedures not adequate to assure the process performs as represented, where the unplanned change itself was never risk-assessed in advance |
Interview Q&A
“How do you know a process is ready for PPQ, beyond the readiness meeting saying so?” A strong answer describes evidence, not process: every prerequisite traced to a dated document that existed before the protocol was approved, not a verbal confirmation collected in the meeting itself. A weak answer describes the meeting’s agenda and attendee list as if holding the meeting were the same thing as being ready.
“A CDMO slot is booked and transfer work is still open two weeks out. What do you do?” The strong answer treats this as a decision with exactly two honest outcomes, move the slot or run and document the accepted gap in writing, and explains why quietly hoping the gap closes in time is not a third option. A weak answer describes pushing the team to finish faster with no fallback if that does not happen.
“Why would a validated process still need a formal batch-count risk assessment instead of defaulting to three batches?” The strong answer explains that FDA’s 2011 guidance ties batch count to Stage 1 knowledge and risk rather than a fixed number, and that “three” without a written rationale answers a different, easier question than the one an inspector is actually asking. A weak answer defends three batches as industry standard with no product-specific reasoning behind it.
“What is the difference between an attribute sampling plan and a tolerance-interval sampling plan, and when does the choice actually matter?” The strong answer ties the choice to the nature of the data, a go or no-go attribute plan for pass or fail counts, a tolerance-interval plan for continuous measurements where closeness to the limit carries information, and gives a concrete example, fill weight or assay percent, where using the wrong one either discards real information or invents a distribution that is not there. A weak answer treats the two as interchangeable statistical formalities.
“Can a PPQ batch that meets the commercial release specification ship before the full PPQ report is approved?” The strong answer explains that the conservative default is no, and that any exception has to be a decision made before batch one, not an improvisation once a batch meets spec. It names FDA’s May 2026 cell and gene therapy CMC flexibility guidance as the concrete basis for concurrent release in that modality class, ties it to the same limited-qualified-lots reasoning that also supports a reduced batch count, and states the hold trigger, any deviation, trend, or failing sibling batch anywhere in the campaign, that would stop it. A weak answer either says batches never ship early under any circumstances, missing that a real, guidance-based path exists, or treats meeting spec alone as sufficient with no mention of a protocol-stated basis or a hold trigger.
“The concurrent-release strategy holds on ‘any deviation anywhere in the campaign.’ How do you decide, mid-campaign, whether a specific deviation counts?” The strong answer explains that the classification criteria have to be written and approved before batch one, sorting deviations into classes tied to what they could plausibly affect rather than how alarming they feel, with the conservative default applying whenever the classification is genuinely uncertain in the moment. A weak answer treats classification as a judgment call made fresh each time, which quietly reopens the door to a “major” label invented specifically to avoid a hold.
“Facility fit was confirmed during CDMO selection months before PPQ. Is that enough for a sterile product?” The strong answer explains that facility fit is a point-in-time assessment of equipment, classification, and utilities, while a sterile product also needs current evidence that the aseptic process itself is under control: a passing media fill within its own qualified interval for the specific line and shift, and no unresolved adverse environmental monitoring trend, both confirmed as part of readiness rather than assumed to still hold from the original assessment. A weak answer treats the facility-fit sign-off as covering aseptic assurance indefinitely.
“Your operators are qualified for the unit operation PPQ will run. Is personnel readiness closed?” The strong answer says role-level qualification is necessary but not sufficient: the batch record approved for PPQ may carry a last-minute procedural change, a new sample point, a tightened in-process limit, introduced after that operator’s last documented training, so readiness requires a dated record confirming training on the exact approved version, not just the underlying unit operation. A weak answer treats an experienced operator’s general qualification as covering whatever the final batch record happens to say.
“A single-source raw material has a long replacement lead time. What closes that as a readiness prerequisite?” The strong answer goes beyond confirming a released lot is on hand: it names the lead time to obtain an emergency lot or qualify a backup source, compares that lead time against the campaign’s actual schedule margin, and requires a written, triggered contingency plan rather than an assumption the current lot will not fail. A weak answer treats a released lot sitting in inventory as closing the question.
“Midway through a PPQ campaign, you realize an in-process limit in the protocol is wrong for the process as it actually runs. Is that a deviation?” The strong answer says no, and explains the test: a deviation explains why one batch departed from an unchanged plan, while a flaw that would recur on every remaining batch unless the plan itself changes is a protocol amendment, requiring the same authorship, QA approval, and regulatory review as the original protocol, plus an explicit statement of whether batches already run still count toward the campaign. A weak answer investigates and closes each batch’s occurrence individually as a string of deviations while the same flawed limit stays in the protocol.
Monday-morning value
If PPQ batch one is still months out, protect the date in this order: lock the control strategy first, since the protocol, sampling plan, and batch record are all written against it; commission hold-time and method studies with margin, not on the critical path; get the statistician into the batch-count and sampling conversation while the protocol is still a draft, so the risk rationale and the attribute-versus-tolerance-interval choice are decided before anyone is attached to a number; confirm batch size and any bracketing against the supply plan before the record is finalized; write and approve the deviation criticality classification before batch one rather than leaving “any deviation” undefined; inventory single-source and long-lead materials early enough that a contingency plan, not a schedule crisis, is the response to a failed lot; schedule version-specific retraining the moment the batch record is finalized, not assumed to be covered by general qualification; for a sterile product, confirm the media fill covering PPQ’s specific line and shift will still be current on the run date, not merely current today; and treat the CDMO date as a target protected by finishing prerequisites early, not a deadline they bend around.
The single most effective habit is requiring dated evidence for every readiness item, not a verbal yes in the pre-execution meeting. A prerequisite that cannot produce a document with a date on it is not closed, whatever the checklist says.