FDA PAIFDA Pre-Approval Inspection

A Pre-Approval Inspection (PAI) is the FDA inspection that verifies a manufacturing facility can produce, in commercial reality, the drug product described in a pending application — a New Drug Application (NDA), Abbreviated New Drug Application (ANDA), Biologics License Application (BLA), or a manufacturing supplement to any of those. It is performed by ORA investigators, often supported by CDER or CBER subject-matter experts, and it gates the approval decision. A No Action Indicated (NAI) or Voluntary Action Indicated (VAI) classification typically clears the path to approval; an Official Action Indicated (OAI) classification typically does not — the application sits in Complete Response Letter territory until the cited issues are resolved.

A PAI is not a routine cGMP surveillance inspection. Its scope is narrower (the exhibit batches and the data in the application), its depth is greater (investigators will trace claims in the application back into the raw data), and its consequence is sharper (the application's approval is on the line). The same facility may have passed three NAI surveillance inspections in a row and still earn an OAI on a PAI because the surveillance scope did not include the application's biobatch data.

FDA Compendium Manual 7346.832 codifies the PAI as a three-objective inspection. Each objective has its own evidence focus and its own failure pattern. Knowing which objective the investigator is testing at any moment is the difference between a fluent response and a stumbling one.

Objective 1 — Readiness

Can this facility actually manufacture this product at commercial scale, as described in the application? Investigators look at the facility, the utilities (water, HVAC, compressed gases), the equipment, the cleaning programme, the change-control history on critical equipment, the personnel qualifications and the quality system. The question is forward-looking — will commercial-scale manufacturing be in control — and the evidence is the validation history, equipment qualification status, and the quality system's track record.

Objective 2 — Conformance

Does the data submitted in the application accurately reflect what the firm actually did during development and exhibit-batch manufacture? Investigators trace specific claims in the application backwards into the source records: the batch records of the biobatch, the analytical raw data, the deviation investigations, the change controls, the stability programme. They look for unreported deviations, batch records that do not match the master, yields that do not reconcile, and OOS results that disappeared.

Objective 3 — Data integrity

Are the records contemporaneous, attributable, original, complete and consistent with the rest of the firm's recordkeeping? Data integrity has become the dominant PAI axis over the past decade. A single uncontrolled spreadsheet, an audit-trail gap in the chromatography software, a 'trial injection' that was never reported — any one of these can convert an otherwise clean PAI into an OAI.

A PAI focuses on the exhibit / biobatch — the validation lots, primary stability lots and any batch whose data is referenced in the application. For an ANDA this is typically the bioequivalence batch (the pilot- or commercial-scale batch on which the BE study was performed) plus the three exhibit batches. For an NDA it includes the Phase 3 clinical batches relevant to the registered manufacturing process. For a supplement it includes whatever batches support the change being filed.

• Exhibit batch records (compared against the approved MMR for any discrepancies).
• Analytical raw data for every release and stability test on the exhibit batches.
• OOS investigations on any exhibit-batch result — disproportionate scrutiny here.
• Change controls between exhibit-batch manufacture and the filed process.
• Deviations on exhibit batches and how they were dispositioned.
• Stability data — including any out-of-trend (OOT) or accelerated-station observations.
• Container-closure integrity testing and packaging configuration.
• Cleaning validation supporting the equipment used.
• Supplier qualification for the API and critical excipients.
• Computerised systems used to generate or store any of the above.

What is generally not in scope of a PAI: products other than the application, routine production outside the exhibit batches, and quality-system workstreams unrelated to the application. Investigators stay within scope, but if they find evidence of broader data-integrity or cGMP issues during the PAI work, the inspection scope can be widened mid-stream.

FDA's 2018 Data Integrity guidance, the MHRA 2018 GxP Data Integrity guidance and PIC/S PI 041 converge on the same model: data must be Attributable, Legible, Contemporaneous, Original (or true copy), Accurate — the ALCOA principles — and Complete, Consistent, Enduring and Available. Investigators test each principle with concrete evidence.

The most common single PAI data-integrity finding is the chromatography audit trail. Investigators ask to see the audit trail for the analytical run that produced the assay result in the application; they look for sample injections that were aborted, results that were reprocessed without justification, and 'trial' injections that should have been reported. Any uncontrolled chromatography environment is now a PAI risk.

A typical PAI lasts 3–8 days on site, with a defined opening meeting, a focused work phase, and a closing meeting where any 483 observations are issued. Knowing the rhythm lets the firm sequence its responses.

1. Opening meeting: investigators present their credentials, state the scope (the application, the products, the inspection objectives), and request opening documents (organisation chart, site master file, list of products, list of changes).
2. Document review: investigators settle into a room with the exhibit-batch records, master records, deviations, change controls, OOS investigations, validation reports and stability data. They identify what to walk down on the floor.
3. Floor walkdown: investigators see the actual equipment, the actual operators, the actual records being made. They observe a batch in progress where possible, look at cleaning, and compare what they see to what the master record says.
4. Lab walkdown: investigators observe analytical methods being run, ask to see audit trails on instruments, review notebooks, and trace specific application results back to the raw data.
5. Quality system review: investigators sample CAPA, change control, deviation, OOS and supplier-qualification workstreams. The depth depends on what they have found so far.
6. Daily wrap-ups: investigators raise emerging issues; the firm has a chance to respond or commit before they harden into 483 observations.
7. Closing meeting: any 483 observations are issued in writing; the firm responds with commitments and timelines.

FDA publishes a public 483 database and an annual inspection observations report. Across the last five years of pharmaceutical PAIs, the same themes recur. Designing the quality system to make each one structurally impossible — rather than relying on operator vigilance — is the practical PAI-readiness strategy.

• Unreported deviations on exhibit batches — the BMR says nothing, but the analytical or yield data show something happened.
• Audit-trail gaps in chromatography or other analytical software — disabled, partial, or not reviewed.
• OOS investigations that hypothesise root cause as 'analyst error' without supporting evidence.
• Specifications changed between exhibit batches and commercial filing without adequate justification.
• Cleaning validation that does not cover the worst-case product or the equipment hold-time used in commercial.
• Stability programme with missing data points or unexplained out-of-trend results.
• Supplier qualification dossiers older than the supplier qualification SOP allows.
• Operators executing the BMR who are not on the current training matrix for the SOP referenced.
• Master record and executed batch record diverge — different process parameters, different IPCs, different sign-off points.
• Computerised systems without the validation evidence to support the GxP records they produce.

PAI readiness is not a project that begins when the FDA Form 482 arrives. It is the cumulative state of the quality system at the moment the inspection opens. A well-designed system makes most of the work invisible because it happens during routine operation.

1. Exhibit-batch records are reviewed contemporaneously against the master, not retrospectively when the application is being filed.
2. Deviations are opened in the system within hours, not days; the close-out cycle time is short enough that nothing sits open across the filing window.
3. OOS investigations follow the 2006 OOS guidance with phase 1, 1B and phase 2 documented and signed; 'analyst error' as root cause without supporting evidence is structurally blocked.
4. Audit trails are reviewed routinely — not just at PAI prep — and the reviewer's record is itself audit-trailed.
5. Computerised systems carry validation packages that map to GAMP 5 categories, with periodic review evidence.
6. Cleaning validation worst-case is current; equipment hold times match commercial reality.
7. Stability programme is in routine state, with OOT triggers reviewed by QA.
8. Supplier qualification is on-schedule, with the supplier audit calendar visible.
9. Training currency gates kiosk access — the SOP can only be executed by someone trained on its current version.
10. Mock PAI is run annually, not three weeks before the real one.

Pre-approval inspections are not unique to FDA. EMA inspections via national competent authorities cover the same ground for centralised-procedure applications; MHRA does the equivalent in the UK; Health Canada's establishment licensing carries a similar function; PMDA's GMP on-site inspection in Japan and NMPA's pre-launch inspection in China each have their own dialect of the same logic. Coordinating PAI readiness with international inspectability is the modern requirement — the same exhibit-batch records may be examined by three or four regulators across the lifecycle.

EU GMP Annex 16 (QP certification), Annex 15 (qualification and validation) and Annex 11 (computerised systems) are the EU's equivalent of the PAI evidence stack. ICH Q7 (APIs), Q8 (development), Q9 (risk management) and Q10 (quality system) frame the international common ground.

V5 is built on the premise that PAI readiness is a property of the operating system, not an exercise that begins three weeks before the inspection. The eBMR / eDHR layer captures every step in the exhibit batches with attribution, contemporaneous timestamp, e-signature and audit trail. The deviation, OOS and change-control workstreams write into the same record as the batch they affect, so the cross-references the investigator wants exist by construction rather than by reconstruction.

• Exhibit-batch records are immutable from work-order release; any change is a tracked deviation, not an overwrite.
• Audit-trail review is a routine batch-closure step, not a PAI-prep activity.
• Analytical results flow into the batch record with raw-data provenance; trial injections and reprocessing are captured.
• Master record snapshot at work-order release means the executed BMR cannot diverge silently from the master.
• Training currency gates kiosk access, so operators executing the BMR are demonstrably current on the SOPs referenced.
• Cleaning validation, supplier qualification and stability data are queryable against the affected batch in seconds.
• Mock PAI mode lets the QA team filter the entire batch record by the seven canonical investigator questions and see the responses pre-assembled.

The FDA Form 483 is issued at the closing meeting. The firm typically has 15 business days to respond in writing — the response is voluntary but practically essential. A factual, evidence-supported response that addresses each observation specifically, commits to verifiable remediation and provides timelines has materially changed classification outcomes in many published cases.

The Establishment Inspection Report (EIR) is the formal narrative of the inspection. It is not released until classification, but the firm can request a copy under FOIA after classification. The classification — NAI, VAI, OAI — is communicated to the application's review division and is the determinant for approvability. An OAI does not automatically mean denial; it means the application cannot be approved until the underlying issues are resolved, which may require follow-up inspection.

• Respond to every observation, in writing, within 15 business days.
• For each observation, state factual response, root cause, immediate corrective action, systemic corrective action, and verifiable timeline.
• Provide evidence — not commitments — wherever possible.
• Do not contest observations on factual grounds unless the facts are clearly wrong.
• Track every commitment to closure with QA verification.