Captioning pharmaceutical training video under GxP: FDA 21 CFR Part 11 validation requirements for caption files, CAPA training documentation, and vocabulary accuracy for GMP and GLP content

The GxP captioning landscape: why pharmaceutical training is different

Every employer with 15 or more employees is subject to ADA Title I and must provide accessible training materials for employees with disabilities. Pharmaceutical and biotech companies are no different from any other employer in this respect. What makes pharmaceutical training different is the additional regulatory layer that attaches to training for regulated activities: FDA quality system regulations, EU GMP requirements, and ICH guidelines that treat training records — including the documents that accompany those records — as part of the quality management system.

The term “GxP” covers the family of “Good Practice” regulations that govern pharmaceutical and biotech manufacturing and research: Good Manufacturing Practice (GMP), Good Laboratory Practice (GLP), Good Clinical Practice (GCP), Good Distribution Practice (GDP), Good Pharmacovigilance Practice (GVP), and Good Automated Manufacturing Practice (GAMP). Each has a specific regulatory home: GMP for 21 CFR Parts 210/211 (drugs) and 820 (devices), GLP for 21 CFR Part 58, GCP for ICH E6, GDP for EU GDP Guideline 2013/C 343/01. All of these require documented training. All of those training records are electronic records if they are generated, maintained, or transmitted electronically. And any caption file that is part of an electronic training record is itself an electronic record.

The compliance picture is therefore a two-layer structure. Layer one is ADA Title I: accessible training video for all employees, WCAG 2.1 AA accuracy standard, regardless of whether the training is for regulated activities. Layer two is 21 CFR Part 11 (and its EU equivalent, Annex 11 of the GMP Guidelines): electronic record requirements for any document that is part of a GxP training record. Both layers apply simultaneously to GxP training video, and they have different quality requirements:

• ADA Title I requires: Caption accuracy at or above 99% (WCAG 2.1 AA) for all employee training video, measured by the DCMP error rate formula.
• 21 CFR Part 11 requires: Validated system for generating caption files, version control for all caption file revisions, audit trail for all modifications, access controls limiting who can alter the authoritative caption file, and retrieval capability on demand for regulatory inspection.

A pharmaceutical L&D team that addresses only the ADA layer — by captioning all training video to WCAG 2.1 AA accuracy — has satisfied the accessibility obligation but has not addressed the quality system obligation. A team that addresses only the 21 CFR Part 11 layer — by validating its captioning tool and controlling the caption files — but produces captions at 85% accuracy has satisfied the documentation requirement but not the accessibility standard. Satisfying both simultaneously requires a captioning programme designed with both frameworks in mind from the start.

The size of the pharmaceutical organisation matters for how urgently this gap becomes visible. Late-stage biotech companies with 100–300 employees preparing for their first GMP inspection face a specific challenge: they have typically been producing training video for 2–4 years before the inspection, none of it managed as electronic records in the formal sense, and the inspection will reveal the absence of validation, version control, and audit trails for a training documentation corpus that the organisation considered complete. The post on building a caption compliance programme from scratch provides the general framework; this post adds the GxP-specific layer on top of that foundation.

What is a GxP training video? GMP, GLP, GCP, GDP, and CAPA

Not all training video at a pharmaceutical company is GxP training video. A new employee orientation video that introduces the company’s mission, a benefits enrollment video, a cybersecurity awareness training, a leadership development programme video — these are not GxP training video even though they are produced by a pharmaceutical company. They are subject to ADA Title I but not to 21 CFR Part 11 electronic records requirements (unless they are, for some reason, part of a quality system record).

GxP training video is training video for regulated activities — activities that fall within the scope of the Good Practice regulations. The dividing line is not always obvious, but the following categories clearly fall within GxP scope:

Good Manufacturing Practice (GMP) training

• Standard Operating Procedure (SOP) training videos — any video that explains how a specific GMP procedure is to be performed, including manufacturing, packaging, labelling, quality control testing, cleaning validation, and equipment operation
• GMP orientation and refresher training — annual GMP training, GMP basics, aseptic technique, gowning procedures, cleanroom behaviour
• CAPA system training — videos explaining the CAPA process, how to write a CAPA, how to perform a root cause analysis, how to assess CAPA effectiveness
• Deviation and non-conformance training — how to identify and report a deviation, deviation investigation procedure, non-conformance classification
• Change control training — how to initiate a change control, change control classification, validation requirements for manufacturing changes
• Batch record training — how to complete a batch record, batch record review, batch record common errors
• Out-of-specification (OOS) and out-of-trend (OOT) investigation procedure training
• Product release and certificate of analysis review training

Good Laboratory Practice (GLP) training

• GLP study conduct training — protocol adherence, data integrity requirements, raw data documentation standards
• Laboratory notebook and data recording training
• Test article and control article handling procedures
• Quality assurance inspection preparation for GLP studies

Good Clinical Practice (GCP) training

• ICH E6 GCP training for clinical site staff, monitors, and clinical operations teams
• Informed consent process training
• Protocol adherence and protocol deviation reporting
• Case report form completion and source document verification
• Investigational product accountability

Good Distribution Practice (GDP) training

• Cold chain management and temperature monitoring procedure
• Distribution record requirements and falsified medicines regulation compliance
• Returned goods procedure and suspected falsified product reporting

Any video in these categories generates a GxP training record when a regulated employee completes it. That training record includes the video content, the caption file if one exists, the completion timestamp, and any attestation (electronic signature) that the employee completed the training. All of these elements together are the electronic training record, and they fall within the scope of 21 CFR Part 11.

The non-GxP training video at a pharmaceutical company — HR, IT security, general professional development — generates an employment training record that does not fall within 21 CFR Part 11 scope (assuming it is not also a GMP or regulatory requirement). These videos still require WCAG 2.1 AA captions for ADA Title I compliance, but they do not require the additional quality system controls described in this guide. The practical challenge for pharmaceutical L&D teams is maintaining the distinction: as more companies use the same LMS for both GxP and non-GxP training, the temptation is to apply a single captioning workflow to all content, which either over-controls the non-GxP content or under-controls the GxP content.

FDA 21 CFR Part 11 and caption files as electronic records

FDA 21 CFR Part 11 “Electronic Records; Electronic Signatures” was issued in 1997 and establishes the criteria under which the FDA will accept electronic records and electronic signatures as equivalent to paper records and handwritten signatures. The regulation applies to records required by FDA regulations — including training records required by 21 CFR Part 211 (GMP for finished pharmaceuticals), 21 CFR Part 820 (Quality System Regulation for medical devices), 21 CFR Part 58 (GLP), and other FDA regulations.

What makes a record subject to 21 CFR Part 11?

Under 21 CFR Part 11, a record is subject to the regulation if it:

1. Is required by an FDA regulation or is submitted to the FDA;
2. Is created, modified, maintained, archived, retrieved, or transmitted under any records requirements set forth in FDA regulations; and
3. Is created, maintained, or transmitted in electronic form.

GxP training records meet all three criteria: they are required by FDA regulations (21 CFR 211.68, 211.68, 820.25, 820.100, among others), they are maintained and retrieved as part of ongoing quality system management, and they are created and transmitted electronically when stored in an LMS or electronic document management system. The training record package for a GMP procedure training typically includes the training assignment record, the completion record, and any training material referenced in the SOP or training curriculum.

Does a caption file become part of the training record?

Whether a caption file is part of the training record depends on how the caption file is treated in the training system. There are three scenarios with different compliance implications:

Scenario A: Caption file is embedded in the video file. When a caption file (SRT or VTT) is burned into the video or delivered as a sidecar file that the LMS attaches to the video record, the caption file becomes part of the training artefact. If the training artefact is part of the training record, the caption file is part of the training record. This is the most common scenario for LMS-delivered GxP training, and it means the caption file is an electronic record subject to 21 CFR Part 11.

Scenario B: Caption file is a separate, referenced document. Some LMS configurations link the caption file as a separate document reference. In this case, the caption file may be separately classified in the document management system and subject to document control requirements (change history, version control, approval workflow). This is the most controlled scenario and the one that most clearly treats the caption file as a controlled document.

Scenario C: Caption file is not part of the training record. If an organisation explicitly excludes caption files from the training record scope — for example, by treating the LMS-delivered training module as a self-contained record and the caption file as a supplementary accessibility accommodation rather than a training document — then the caption file may fall outside 21 CFR Part 11 scope. This position is defensible if the training record is complete without reference to the caption file, but it requires a documented quality decision and creates risk: if a regulatory inspector asks for the caption file as part of an accessibility review, and the organisation cannot produce a controlled version, the absence of control becomes visible.

The FDA’s 2003 Guidance for Industry on Part 11, Electronic Records; Electronic Signatures — Scope and Application provides useful context: the FDA takes a risk-based approach to Part 11 enforcement, focusing on records “that are relied upon to perform regulated activities” rather than supplementary documentation that plays no role in the regulated activity itself. The practical question is whether the caption file is relied upon as part of the training process — if employees with hearing impairments use the caption file to complete the required GMP training, and the training system records their completion based on the captioned version, then the caption file is functionally part of the training record that the employee relied upon, regardless of how it is classified in the document management system.

The 21 CFR Part 11 requirements that apply to caption files

For caption files that are electronic records subject to Part 11, the following requirements apply:

The EU equivalent framework is Annex 11 of the EU GMP Guidelines (“Computerised Systems”), which has similar requirements but uses different terminology: “critical computerised systems” must be validated, audit trails must be “computer-generated,” and access must be controlled through a “management system for data and for documents.” The practical requirements are closely aligned with 21 CFR Part 11, and an organisation that satisfies Part 11 for caption files will generally satisfy Annex 11 as well, with minor documentation adjustments.

Validation requirements for the captioning system: IQ, OQ, and PQ

Validation under 21 CFR Part 11 is not a one-time certification — it is a documented process that demonstrates the captioning system performs as intended and produces output that meets the quality criteria required for the intended use. The GAMP 5 framework (Good Automated Manufacturing Practice, 5th edition) provides the most widely used methodology for validating computerised systems used in GxP environments, and most pharmaceutical quality teams will recognise the IQ/OQ/PQ structure.

Installation Qualification (IQ)

The IQ confirms that the captioning software has been installed correctly in the organisation’s environment according to the vendor’s specifications. For a cloud-based captioning tool accessed via browser or API, this means documenting:

• The software version in use and the mechanism for managing version upgrades
• The infrastructure environment (browser versions, operating system, network configuration, API endpoint)
• Any configuration settings applied at installation (accuracy model selection, language settings, output format options)
• The vendor’s technical documentation confirming the software is designed to operate in a GxP-compatible environment
• Security configuration (access credentials, permission tiers, MFA settings if applicable)

The IQ protocol should include a pre-installation checklist and a post-installation verification checklist. The installation must be performed by authorised personnel and signed off by QA. For API-integrated captioning tools, the IQ extends to the API integration layer — the custom code or middleware that connects the captioning service to the LMS or document management system is also a computerised system that requires documentation.

Operational Qualification (OQ)

The OQ confirms that the captioning system performs as intended under expected operating conditions. The OQ protocol should test all functions that will be used in GxP caption production, including:

• Audio file submission: Confirm that audio and video files in all formats used for GxP training content can be submitted to the captioning system without loss of data integrity.
• Glossary application: Confirm that the GxP glossary (CAPA, deviation, OOS, OOT, change control terminology) is applied correctly during ASR processing, with documented test cases showing correct glossary term substitution.
• Output file generation: Confirm that SRT and VTT output files are generated in conformance with the file format standard, with correct timing, character encoding, and file integrity.
• Access control: Confirm that user permission tiers function as designed — that a reviewer-only user cannot approve and release, and that an administrator cannot process without a separate approval step.
• Audit trail generation: Confirm that system-generated audit trail entries are created for all relevant operations — submission, processing, glossary application, review, correction, approval, and release.
• Version control: Confirm that version numbering increments correctly on each revision of a caption file and that prior versions are retrievable.
• Error conditions: Confirm system behaviour under error conditions — rejected file formats, network interruptions, API failures — and verify that error conditions do not result in partial or corrupted caption files being released without detection.

The OQ should include a minimum of three test executions for each critical function to establish repeatability. Test cases should be documented with expected results defined before testing, and actual results recorded against expected results. Any deviation between expected and actual results must be investigated and resolved before OQ completion.

Performance Qualification (PQ)

The PQ confirms that the captioning system consistently performs to the defined quality criteria under actual conditions of use, over a sustained period. The PQ for a captioning tool typically covers:

• Accuracy performance against GxP content: A defined test set of GxP training audio content (minimum 10 clips representing the vocabulary categories used in the organisation’s training library) must be processed, and the accuracy of the resulting caption files must be measured against a reference transcript using the DCMP error rate formula. The pass criterion must be defined in the PQ protocol before testing begins — typically 97% or better at PQ, with 99% as the production target after review correction. See the post on caption quality and the DCMP error rate formula for the measurement methodology.
• Throughput under production conditions: The system must demonstrate the ability to process the expected production volume within acceptable turnaround times. If the organisation produces 20 GxP training videos per month, the PQ should simulate a representative production load and confirm the system meets the throughput requirement.
• Sustained operation: PQ should run for a defined period (typically 30 days in production or in a production-representative environment) with all critical metrics tracked and reported.

The PQ requalification schedule should be defined in the validation plan. A major software update (version change that affects captioning functionality), a change in the underlying ASR model, or a significant change in the types of training content being processed are all change control triggers that should initiate requalification of the affected PQ elements.

Ongoing monitoring after validation

Validation is not a one-time event. After initial validation, the captioning system must be monitored as part of the quality management system. This means:

• Periodic accuracy spot-checks against the DCMP protocol, with results trending over time
• Prompt investigation of any accuracy decline that crosses the requalification threshold
• Change control review for any system updates, configuration changes, or glossary updates that might affect validated performance
• Annual review of validation documentation currency, confirming that the system description and test records still accurately reflect the system in use

The caption programme annual review process provides a framework for incorporating captioning system performance review into the organisation’s periodic review cadence.

Version control for caption files in a GxP environment

Version control for GxP documents is not a software feature — it is a quality requirement. 21 CFR Part 11 § 11.10(e) requires audit trails that capture changes to electronic records, and 21 CFR Part 211.68(b) requires that computer systems used in drug manufacturing be “validated for their intended use.” The practical implication for caption files is that every version of a caption file associated with a GxP training video must be identifiable, datable, and traceable to the individual who made the change.

Caption file versioning triggers

In a GxP environment, a new version of a caption file is required whenever the caption content changes. Caption content changes for several reasons:

• Initial creation: The first approved version of a caption file for a new GxP training video. This is Version 1.0 (or v1, depending on the organisation’s document numbering convention).
• Accuracy correction after QA review: The captioning tool’s first-pass output is reviewed for accuracy before the caption file is approved. Corrections made during the QA review process that change the text of a caption constitute a revision. If the QA review occurs before the caption file is released to the LMS, the reviewed version is still Version 1.0 (initial approved version). If corrections are made to a caption file that has already been released and used in training, the corrected version is Version 1.1 (or the next minor version).
• Underlying video revision: When the GxP training video is revised — because an SOP has changed, a CAPA procedure has been updated, or a product has been reformulated — the caption file must also be revised to match the new video content. A revised video that has not triggered a caption file revision is a version control failure.
• SOP or procedure change that affects terminology: Even if the video itself has not changed, a change in the organisation’s SOP that changes how a procedure is named or described may require caption file revision if the captions contain the old terminology. This is a subtle versioning trigger that many L&D teams miss.
• Glossary update affecting caption content: If a change to the GxP glossary results in different substitutions being applied to a caption file (for example, a product code has been changed), and the caption file is re-generated or re-processed with the new glossary, this constitutes a revision. The post on caption glossary maintenance workflow covers the governance process for glossary updates.

Version numbering conventions

Most pharmaceutical document management systems use a major.minor versioning convention: Version 1.0 is the initial approved version, Version 1.1 is the first minor revision (typically for editorial corrections or minor content updates that do not change the substance of the training), and Version 2.0 is a major revision (typically triggered by a substantive change to the procedure or training content that requires re-approval). The same convention should apply to caption files:

A critical version control failure that pharmaceutical L&D teams repeatedly encounter: the LMS contains a version of a training module that references a caption file, but the underlying video has been revised and the caption file has not been updated. The learner completes the training, the LMS records a completion, but the caption file they used during training was the v1.0 caption for the old procedure, not the current procedure. This creates a training record that attests to completion of outdated training — a significant finding in a GMP inspection.

Audit trail requirements: who changed the caption file and when

The audit trail requirement under 21 CFR Part 11 § 11.10(e) states that secure, computer-generated, time-stamped audit trails must independently record the date and time of operator entries and actions that create, modify, or delete electronic records. For caption files, this means:

• Every time a caption file is created, the audit trail captures: the date and time, the identity of the system or user that created it, the system version, the glossary version applied, and any input parameters that influenced the output.
• Every time a caption file is modified, the audit trail captures: the date and time, the identity of the user making the change, a description of what changed (or the before-and-after content of the modified caption cues), and whether the change was made in the captioning tool, the LMS, or a separate editor.
• Every time a caption file is approved or released, the audit trail captures: the date and time, the identity of the approver, and the version being approved.
• Every time a caption file is accessed, the audit trail captures: the date and time and the identity of the user (access audit trail, not just modification audit trail).

What is not a compliant audit trail

A manual revision log — a spreadsheet or Word document where the L&D team records changes to caption files — is not a compliant audit trail under 21 CFR Part 11. The regulation specifically requires “computer-generated” and “secure” audit trails, meaning the audit trail cannot be modified by the same users who are modifying the records, and its entries must be generated automatically by the system rather than recorded manually. A spreadsheet can be modified to remove or alter entries; a compliant audit trail cannot.

Similarly, version control in a Git repository satisfies many of the version tracking requirements but may not satisfy all of the access control and audit trail requirements unless the repository is configured with appropriate access controls, protected branches for approved versions, and a mechanism for linking commits to authenticated user identities. Git is a viable component of a GxP document control system if properly configured and validated, but it is not an out-of-the-box compliant solution.

Practical approaches for caption file audit trails

The most practical approach for managing caption file audit trails in a pharmaceutical organisation is to use the organisation’s existing electronic document management system (EDMS) or quality management system (QMS) to manage caption files as controlled documents. Systems like Veeva Vault QMS, MasterControl, Documentum, and OpenText have built-in 21 CFR Part 11-compliant audit trail capabilities. If the caption file is managed as a controlled document in one of these systems, the audit trail requirement is satisfied by the underlying document management system.

The challenge is that most EDMS systems are not designed to preview or play video caption files in their native format — they can store and version-control an SRT or VTT file, but they cannot display it in the context of the video it accompanies. This creates a practical workflow gap: the caption file is controlled in the EDMS, but the QA reviewer who needs to verify caption accuracy against the video must work outside the EDMS. A documented review procedure that bridges the EDMS (for document control) and the LMS or video player (for content review) is required.

CAPA training video: documentation requirements and caption classification

Corrective Action and Preventive Action (CAPA) is among the most heavily scrutinised elements in any GMP inspection. FDA warning letters and 483 observations consistently cite CAPA system deficiencies, and CAPA training records are among the first documentation packages an inspector reviews when assessing whether the quality system is functioning as designed. The logic is straightforward: if employees are not trained on the CAPA procedure, they cannot be expected to use it correctly, and CAPA procedure failures trace directly back to training failures.

Why CAPA training captions are high-stakes

CAPA training video has two properties that elevate the compliance stakes for caption files:

First, CAPA training is required, not optional. Under 21 CFR 820.100 (Quality System Regulation for medical devices) and the equivalent GMP requirements for drugs, the CAPA system must be a documented procedure, and personnel who perform or verify activities for CAPA must be trained on those procedures. This means the CAPA training record is not a convenience — it is required evidence that the regulatory requirement has been met. If an FDA inspector requests evidence that the CAPA procedure has been communicated to quality personnel, and the only evidence is an LMS training record, and the training record contains a captioned video that the inspector reviews, the quality of the caption file is directly visible in the regulatory inspection context.

Second, CAPA training vocabulary is dense and highly specific. A CAPA training video will reference: root cause analysis (RCA), the five whys methodology, fishbone/Ishikawa diagram, CAPA effectiveness check, action item owner, corrective action vs. preventive action distinction, containment action, interim control, risk assessment, FMEA (Failure Mode and Effects Analysis), risk priority number (RPN), CAPA closure criteria, and regulatory citation for CAPA requirements. This vocabulary fails consistently in general-purpose ASR systems. A CAPA training video captioned with general-purpose ASR at 85% accuracy will have specific, inspectable errors in the most technically important terms.

The CAPA caption documentation package

For a CAPA training video in a GxP environment, the complete documentation package that should be maintained alongside the training record includes:

• The SRT or VTT caption file at its current approved version
• The caption file version history and audit trail
• The reference transcript used for caption accuracy verification (the reference transcript is a primary document — see the caption vendor accuracy evaluation methodology post for why this is the step most organisations skip)
• The accuracy measurement record (DCMP spot-check results confirming the caption meets the accuracy standard)
• The approval record for the current caption file version
• The glossary version applied during generation (this is what allows the QA reviewer to reproduce the original output if needed)

This documentation package is what a CAPA system review team, an internal auditor, or an FDA inspector can request when reviewing the CAPA training record. An organisation that cannot produce the caption file version history and accuracy measurement record has a documentation gap in its CAPA training record — the same gap that, in any other part of the CAPA system, would be cited as an observation.

SOP change-notification video: caption as a controlled document

Standard Operating Procedures (SOPs) change constantly in a functioning GMP environment. SOP revisions triggered by CAPA actions, manufacturing process improvements, equipment upgrades, regulatory requirement updates, or audit findings all require that affected personnel be trained on the change before the revised SOP becomes effective. Video has become a common medium for SOP change-notification training: a 3–7 minute video explaining what changed, why it changed, and what employees need to do differently can be assigned in the LMS, completed asynchronously, and documented in the training record with a completion timestamp and electronic signature.

The controlled document parallel

In GMP, every SOP is a controlled document: it has a document number, a version, an effective date, an approval signature, and a document history. When the SOP is revised, the prior version is superseded and the revision history is preserved. The SOP change-notification training video is the communication artefact for that revision. If the training video is the mechanism by which affected personnel receive the content of the SOP revision, then the training video — and its caption file — are functionally aligned with the SOP version it communicates.

This creates a document linkage requirement: the SOP change-notification training video and its caption file should be linked to the SOP version they document. When the SOP is at Version 3.2 and the change-notification video explains the changes in the 3.1 → 3.2 revision, the training record for that video should reference the SOP version, and the caption file should be at the version that accurately reflects the Version 3.2 procedure. If the SOP is subsequently revised to Version 3.3, the change-notification training for 3.2 remains valid as historical training for the 3.1→3.2 change — but the caption file for that training should not be updated to reflect Version 3.3 language unless the 3.2 video itself has been revised.

The vocabulary problem in SOP change-notification video

SOP change-notification videos have a specific vocabulary accuracy challenge: they describe procedural changes using the exact technical language of the SOP, which includes equipment names, material codes, parameter names, step numbers, and specification values. For example, a change-notification video for a revised cleaning validation SOP might say: “The rinse volume for equipment number 7-F-112 has increased from 15 litres to 20 litres, and the acceptance criterion for the swab sample at location S-7 has been tightened from 10 parts per million to 5 parts per million.” Every element of that sentence is a GxP-specific vocabulary item: equipment number format, volume measurement, swab sample location code, acceptance criterion unit. General-purpose ASR has low prior probability on these patterns and will make systematic substitution errors.

The customer glossary architecture post covers how to build a glossary that handles equipment naming conventions, material codes, and procedural parameter vocabulary — all of which are necessary for accurate SOP change-notification captioning.

Deviation investigation training: vocabulary and documentation requirements

Deviation training — training on how to identify, report, classify, investigate, and close manufacturing deviations or protocol deviations (in clinical operations) — is among the most vocabulary-dense GxP training content. Deviations are the real-time quality events that, handled correctly, prevent product failures; handled incorrectly, they become the source of CAPA findings, 483 observations, and in severe cases, consent decrees. Deviation training vocabulary draws from three distinct domains simultaneously: the regulatory framework (FDA guidance on deviation classification, ICH Q10 process performance), the quality system (deviation investigation report structure, risk assessment methodology, escalation criteria), and the manufacturing or clinical operations context (process parameters, equipment, materials, specifications).

Deviation investigation vocabulary that consistently fails ASR

The following terms appear in deviation investigation training video and produce specific, documented ASR failure patterns:

The vocabulary failure pattern in deviation investigation training is primarily acronym substitution — the same pattern documented for medical vocabulary (where drug abbreviations fail) and engineering vocabulary (where technical acronyms fail). See the posts on drug name ASR failures in medical training and HazCom chemical vocabulary failures for adjacent failure profiles. The GxP deviation vocabulary failure set is distinct from both, drawing on regulatory procedure terminology rather than chemical nomenclature or clinical pharmacology.

The GxP vocabulary accuracy problem: why ASR fails on regulatory procedure terminology

Automatic speech recognition systems are probabilistic text generators constrained by an acoustic model (what phoneme sequences occur in speech) and a language model (what word sequences occur in text). The accuracy of an ASR system on a given vocabulary depends on how frequently those words and phrases appear in the training data. General-purpose ASR systems are trained on large corpora of general speech — books, websites, news transcripts, podcasts, conversational audio — where GxP regulatory procedure vocabulary appears rarely or not at all.

The language model component is particularly important for GxP vocabulary. CAPA, OOS, OOT, FMEA, ICH, and GMP are initialisms that appear in regulatory document corpora but are absent from the conversational and media text corpora that dominate ASR training data. The acoustic model may correctly recognise the phoneme sequence, but the language model assigns low probability to the correct expansion because the expansion appears rarely in training data — producing the substitution errors documented in the table above.

Baseline accuracy by GxP content category

Based on documented captioning performance across representative GxP training content types, the following accuracy ranges apply before glossary correction:

Without a GxP-specific glossary, the CAPA training and deviation investigation content categories fall below 86% accuracy — far below the WCAG 2.1 AA 99% standard and, critically, far below the threshold at which the caption file is a reliable record of the training content. A caption that says “a capper procedure requires root cause analysis and containment action” instead of “a CAPA procedure requires root cause analysis and containment action” is not an accurate training record of what the employee watched.

The accuracy gap closes substantially with a well-constructed GxP glossary. The post on why 99% caption accuracy matters covers the absolute error count implications: at 79% accuracy on a 10-minute CAPA training video, there are approximately 290 errors per 1,380-word-equivalent module. At 99% accuracy, there are 14 errors — the difference between a caption file that is functionally unusable for training documentation and one that is a reliable record.

Per-category vocabulary analysis: CAPA, batch record, change control, risk assessment, international regulatory

The GxP vocabulary failure set is large enough that building the glossary requires systematic category-by-category sourcing rather than term-by-term discovery during QA review. The following categories should be used as the glossary seed framework for pharmaceutical and biotech training content.

CAPA and quality event vocabulary

• Acronyms: CAPA, RCA, DIR, FMEA, RPN, OOS, OOT, CAPAs (plural), CAPA ID, CAPA effectiveness check (CEC)
• Procedure names: corrective action and preventive action, deviation investigation report, non-conformance report (NCR), quality event, product complaint, field alert report (FAR), adverse event report
• Methodology terms: root cause analysis, five whys, fishbone diagram, Ishikawa diagram, assignable cause, contributory cause, systemic root cause
• CAPA system steps: containment, investigation, root cause, action plan, effectiveness check, closure
• Classification terms: critical deviation, major deviation, minor deviation, unplanned deviation, planned deviation

Batch record and manufacturing vocabulary

• Document names: batch record (BR), batch manufacturing record (BMR), batch packaging record (BPR), completed batch record (CBR), executed batch record
• Review terms: batch record review, batch record errors, reconciliation, yield calculation, in-process control (IPC), critical process parameter (CPP), critical quality attribute (CQA)
• Release terms: batch disposition, product release, certificate of analysis (CoA), certificate of conformance (CoC), quality release, commercial release, quarantine status
• Manufacturing terms: granulation, blending, compression, coating, lyophilisation (freeze-drying), aseptic fill-finish, visual inspection, in-process testing

Change control vocabulary

• Document types: change control request (CCR), engineering change order (ECO), change control board (CCB), change control committee
• Classification terms: major change, minor change, administrative change, retrospective change control
• Regulatory submissions: prior approval supplement (PAS), changes being effected in 30 days (CBE-30), annual product quality review (APQR)
• Technical terms: revalidation required, comparability study, bioequivalence, pharmaceutical equivalence, process validation impact assessment

Risk assessment vocabulary

• Frameworks: ICH Q9 (quality risk management), FMEA, HACCP (hazard analysis and critical control points), fault tree analysis (FTA), HAZOP (hazard and operability study)
• Scoring elements: severity, occurrence, detectability, risk priority number (RPN), risk score, risk ranking, residual risk
• Control terms: critical control point, preventive control, detective control, risk mitigation, risk acceptance, risk communication

International regulatory vocabulary

• ICH guidelines: ICH Q1, Q2, Q3, Q6, Q8, Q9, Q10, Q11, Q12, Q13 (each a specific guideline with a specific topic)
• EU GMP: Annex 1 (sterile manufacturing), Annex 11 (computerised systems), Annex 15 (qualification and validation), Annex 16 (release for sale)
• FDA guidance: Process Validation (2011), Pharmaceutical Quality System (21 CFR 211), Quality Systems for the 21st Century (QSIT)
• Organisation names and acronyms: FDA, EMA (European Medicines Agency), MHRA (UK Medicines and Healthcare products Regulatory Agency), PMDA (Japan), TGA (Australia), WHO, PIC/S
• Inspection and audit: GMP inspection, Form 483, warning letter, consent decree, import alert, close-out meeting, PAI (pre-approval inspection)

GLP and GCP-specific vocabulary

• GLP: principal investigator (PI), study director, quality assurance unit (QAU), test facility, test article, control article, raw data, study plan, final report, GLP compliance statement
• GCP: investigator, monitor, CRA (clinical research associate), CRO (contract research organisation), IRB (institutional review board), IEC (independent ethics committee), informed consent form (ICF), case report form (CRF), source data verification (SDV), essential documents

Building the glossary from these categories at the start of a GxP captioning programme — rather than waiting for QA review to surface individual term failures — prevents the compounding problem where every new training video starts at a low baseline because the glossary was not seeded for the content type. The caption feedback loop post covers the compounding accuracy improvement that occurs as the glossary is iteratively refined from production experience.

Glossary architecture for pharmaceutical training content

A pharmaceutical training glossary must be more than a flat list of terms. The GxP vocabulary set contains terms that are context-dependent, terms that have organisational-specific forms, and terms that change with SOP and procedure revisions. A well-architected GxP glossary has four tiers:

Tier 1: Stable regulatory vocabulary

Regulatory terms, ICH guideline references, FDA form numbers, CFR citations, and EU directive and guideline references. These terms are externally defined and do not change based on the organisation’s internal decisions. They form the stable foundation of the glossary and should be entered once and maintained by reference to current regulatory publications. Examples: CAPA, OOS, OOT, ICH Q10, 21 CFR Part 211.68, Annex 11, Form 483.

Tier 2: Organisation-specific procedure vocabulary

Internal document names, procedure titles, system names, and role titles as they are used in the organisation’s SOPs. These terms are defined by the organisation’s document management system and quality system procedures. Examples: the specific name the organisation gives to its CAPA form (which may be “Quality Event Report” at one company and “Corrective Action Request” at another), internal system names (Veeva Vault, TrackWise, QUMAS), role titles (Quality Assurance Manager, Site Quality Head, Qualified Person). These terms are owned by the quality department and should be updated in the caption glossary whenever the SOP or procedure that defines them is revised.

Tier 3: Product and process vocabulary

Product names (INN, brand, internal codes), manufacturing equipment names and codes, facility names and codes, material codes, specification values and units, and process parameter names. These terms are owned by manufacturing, R&D, and regulatory affairs and change with product portfolio changes, facility modifications, and manufacturing process updates. This is the highest-maintenance tier of the glossary because it is directly coupled to the product lifecycle. See the caption glossary maintenance workflow post for the governance process for managing product vocabulary updates across the glossary.

Tier 4: People and organisation vocabulary

Speaker names, site names, department names, and external organisation names (regulatory agencies, contract manufacturers, suppliers). Speaker names matter in GxP training video because the training often features subject matter experts who are identified by name in the video — a quality manager introducing the CAPA procedure, a regulatory affairs director explaining an ICH guideline. Incorrect speaker names in the caption file are a visible accuracy failure in an inspection setting where the inspector may know the individuals.

Glossary version control

The glossary is itself a controlled document in a GxP environment. Every change to the glossary — adding a term, removing a deprecated term, changing a term definition — is a controlled change that should follow the same version control process as SOP revisions. The glossary version that was applied during caption generation should be recorded in the caption file’s audit trail, so that a QA reviewer can reproduce the original generation with the same settings. When a glossary update changes the substitution for a term that appears in an existing approved caption file, the organisation must decide whether to reprocess that caption file (creating a new version) or accept the existing version as accurate for the period it was generated. This is a documented quality decision.

For organisations with mature quality management systems, the glossary change control process can be integrated into the standard document change control system: glossary updates are treated as SOP revisions, with a change control request, impact assessment (which caption files are affected), and approval workflow before the new glossary version is deployed to the captioning system.

LMS considerations for GxP training: Veeva Vault, ComplianceWire, Cornerstone Life Sciences

The LMS landscape for pharmaceutical GxP training is distinct from the general corporate L&D LMS market. Pharmaceutical companies use specialised platforms built with GxP compliance requirements in mind, or configure general-purpose enterprise LMS platforms with GxP validation packages. Each platform has a different approach to caption file management.

Veeva Vault Training

Veeva Vault Training is the dominant GxP training platform for pharmaceutical companies with 500+ employees and is increasingly used by late-stage biotech companies approaching their first GMP inspection. Veeva Vault is a validated cloud platform with native 21 CFR Part 11 compliance for all documents stored and managed within the system. Caption files stored as document attachments to training content within Veeva Vault inherit the Vault audit trail and version control capabilities — the Vault system generates the computer-generated, time-stamped audit trail entries required by § 11.10(e) automatically, without requiring additional infrastructure from the L&D team.

The practical limitation for caption management in Veeva Vault: the Vault system is primarily a document management and training assignment system; it is not a video editing or caption review environment. Caption files must be created and reviewed outside Vault, then uploaded as controlled documents within Vault. The workflow is: (1) GxP training video produced by L&D, (2) caption file generated by validated captioning tool, (3) caption file QA reviewed for accuracy in a separate environment, (4) reviewed caption file uploaded to Vault as a controlled document, (5) approved version attached to training content in Vault, (6) training assigned and completed in Vault with audit trail. Steps 1–4 occur outside Vault and require their own documentation; only step 5 onward is managed within Vault’s Part 11-compliant environment.

ComplianceWire

ComplianceWire is a GxP learning management system built specifically for pharmaceutical, biotech, and medical device companies. It supports 21 CFR Part 11-compliant electronic signatures and training records. ComplianceWire manages training assignments, completion records, and electronic signatures natively but has more limited native caption support than general-purpose LMS platforms. Caption files for ComplianceWire-delivered GxP training are typically managed as document attachments within the ComplianceWire content management layer, with the document control infrastructure managed by the organisation’s quality team rather than the ComplianceWire system itself.

Cornerstone OnDemand (Life Sciences configuration)

Cornerstone OnDemand is used by some pharmaceutical companies for both GxP and non-GxP training, with GxP training configured in a validated partition of the system. Cornerstone’s native caption support allows SRT and VTT files to be uploaded as sidecar files for video content. In a validated Cornerstone configuration, the caption file version and the video version should be tracked as linked items — a revision to either the video or the caption file triggers a change control review for both. See the post on enterprise LMS caption audit methodology for how to audit a Cornerstone configuration for caption compliance gaps.

MedBridge and HealthStream

HealthStream and MedBridge are primarily used for clinical and nursing staff training but are also used for GxP-adjacent training in hospital-based clinical trial sites and academic medical centres. Caption support in these platforms is less developed than in the dedicated pharmaceutical GxP platforms, and the validation infrastructure for caption file management typically requires a supplementary document control layer rather than native platform capabilities.

The SCORM packaging problem for GxP training

Many pharmaceutical companies deliver GxP training as SCORM packages built in Articulate Storyline, Rise, or Adobe Captivate, and launched from the LMS. SCORM packages present a specific challenge for caption version control: the caption file is typically embedded within the SCORM package rather than managed as a separate sidecar file. When the caption file needs to be revised — for an accuracy correction or a procedure change — the entire SCORM package must be rebuilt and republished, which constitutes a change to the training module that triggers the LMS content change control process. The post on SCORM and xAPI caption delivery tracking covers the technical mechanics of caption delivery in SCORM packages and the options for managing caption version control at the package level.

For GxP training delivered as SCORM, the most manageable approach is to use xAPI (Tin Can) rather than SCORM where the LMS supports it, because xAPI allows more granular reporting of learner interaction with caption tracks and enables more flexible content management at the LMS level without requiring full package replacement for caption-only revisions.

Building a caption programme for a GxP environment

The general framework for building a caption compliance programme is covered in the post on building a caption compliance programme from scratch. The GxP-specific additions to that framework are the quality system integration elements: the validation protocol, the document control integration, and the change control process for the captioning system itself. This section covers the GxP-specific elements in the order a late-stage biotech or pharmaceutical company approaching its first GMP inspection should implement them.

Step 1: Classify your training video library

Before any captioning work begins, the L&D team must work with the quality team to classify the training video library by GxP scope: which videos are GxP training content (subject to 21 CFR Part 11 caption controls), and which are non-GxP (subject only to ADA Title I). The classification should be documented and maintained as a controlled document — a training content registry that maps each video to its GxP classification, the applicable regulation, and the caption file status.

Step 2: Select and validate the captioning tool

Select a captioning tool that supports the technical capabilities required for GxP use: glossary application, API access for system integration, audit trail output, version management, and access control. Then execute the IQ/OQ/PQ validation protocol described in the validation section above, with QA participation in the protocol design and approval of the validation results.

Step 3: Build the GxP glossary

Using the category framework in this guide, build the initial GxP glossary and validate its effect on caption accuracy as part of the PQ protocol. The PQ test set should include representative samples from each GxP content category (CAPA, deviation, SOP change-notification, OOS/OOT, batch record, change control) to confirm that the glossary achieves the accuracy improvement required for both WCAG 2.1 AA compliance and reliable training record documentation.

Step 4: Establish the caption file document control procedure

Write an SOP (or SOP section) that covers: caption file classification (GxP vs. non-GxP), the captioning workflow for GxP content (generation, review, approval, release), version control triggers and numbering conventions, audit trail maintenance, access control assignments, the DCMP-based accuracy measurement protocol for QA review, and the change control triggers for the captioning system and glossary. This SOP should be approved by QA before GxP caption production begins.

Step 5: Backlog remediation for existing GxP training video

Most pharmaceutical companies that implement a GxP captioning programme have a backlog of uncaptioned or under-controlled GxP training video. The backlog remediation approach for GxP content is risk-stratified:

• Priority 1 (immediate): GxP training video for procedures where an active CAPA, an OOS investigation, or an audit finding identified training as a root cause. These are the highest-risk items and should be captioned and brought into document control before the next inspection.
• Priority 2 (30–60 days): Required GxP training for all current manufacturing or operations personnel — GMP orientation, CAPA procedure, deviation investigation, SOP training for current effective SOPs. These are the videos that inspectors are most likely to request during a training record review.
• Priority 3 (60–90 days): GxP training video for discontinued procedures that still appear in the training record (because personnel who completed them before the procedure was discontinued still have the training record). These should be captioned and archived as controlled historical documents.
• Priority 4 (ongoing programme): All future GxP training video goes through the caption control programme before release.

The large-scale backlog remediation framework is covered in the post on caption backlog remediation; the GxP-specific addition is the risk stratification by regulatory exposure rather than by content age or usage frequency.

Step 6: Integrate caption compliance into the annual training review

GxP training programmes require periodic review as part of the quality management system. The annual review of the GxP training programme should include: a review of caption file version currency (confirming that no caption file has become stale due to a procedure change), a spot-check accuracy review of a random sample of GxP caption files (measuring against the DCMP protocol), a review of the captioning system validation status (confirming no uncontrolled changes have been made to the system or glossary), and a review of any accessibility accommodation requests received in the prior year. The caption programme annual review process post provides the general annual review framework; the GxP-specific elements should be integrated as additional review items in the quality system’s periodic review calendar.

The regulatory inspection scenario: what inspectors look for in caption records

FDA GMP inspections operate through a process called the Quality Systems Inspection Technique (QSIT), which focuses on the interconnections between quality subsystems rather than reviewing individual records in isolation. The training subsystem is a component of the Production Quality System and the Management Responsibilities Quality System under QSIT. Inspectors reviewing the training subsystem are looking for evidence that the training programme is effective — that the right people have been trained on the right procedures at the right time, and that the training records accurately document that training occurred.

When caption files become visible in an FDA inspection

Caption files are not a primary focus of FDA GMP inspections — they become visible in specific scenarios:

• Accessibility complaint or accommodation request review: If the inspection includes a review of quality system documentation related to a complaint or accommodation request by an employee with a disability, the caption records for the training involved may be requested.
• Training record completeness review: When an inspector requests a complete training record for a specific employee or for the CAPA training programme, and the training included captioned video, the caption file and its version history may be part of the requested documentation package.
• CAPA training effectiveness finding: If an investigation has identified CAPA training as a root cause of a quality system failure, the CAPA training records including caption files may be reviewed as part of the investigation documentation.
• Computerised systems review: Under Annex 11 (EU) or the FDA’s approach to computerised systems validation, the captioning tool itself may be reviewed as a computerised system used in a GxP-regulated activity. The validation documentation for the captioning tool (IQ, OQ, PQ protocols and reports) may be requested.

What documentation to have ready

A pharmaceutical quality system that is prepared for caption records review should be able to produce, on request:

• The current approved version of any specific GxP caption file
• The version history for that caption file, including the date and author of each revision
• The audit trail for caption file access and modification
• The accuracy measurement record for the caption file (DCMP spot-check results confirming 99% or better)
• The reference transcript used for accuracy measurement
• The validation documentation for the captioning system (IQ, OQ, PQ protocols and reports)
• The SOP governing the GxP caption production and document control process
• The glossary version that was applied during generation and the change history of the glossary

An organisation that cannot produce this documentation within a reasonable inspection response time (typically 30–60 minutes for a specific document request, same-day for a documentation package request) is in a weak compliance posture for caption records. The documentation should be organised in a location accessible to the QA team during an inspection, not scattered across L&D SharePoint sites, the captioning vendor portal, and individual team members’ desktops.

The caption programme governance policy template provides the policy framework that organises these documentation requirements into a formal quality system document.

Eight failure modes that expose pharmaceutical organisations to regulatory findings

Failure mode 1: Unvalidated captioning tool generating GxP training records

Using a captioning tool that has not been through the IQ/OQ/PQ validation process to generate caption files that become part of GxP training records. This is the most foundational failure: every subsequent quality control effort (version control, audit trails, accuracy measurement) is built on a foundation of unvalidated system output. The fix requires retroactively validating the captioning tool (or selecting and validating a new one), then reviewing the accuracy of caption files generated before validation and remediating any that fall below the accuracy standard.

Failure mode 2: Caption files stored outside the document control system

GxP caption files stored on shared drives, in the captioning vendor’s portal, in the LMS without document control integration, or on individual workstations rather than in the organisation’s validated document management system. Files in these locations do not have the access controls, audit trails, and version control required by 21 CFR Part 11. The fix requires identifying all GxP caption files in uncontrolled storage, importing them into the document control system with a documented justification for the version being imported, and establishing the SOP for all future GxP caption files to go directly into document control after approval.

Failure mode 3: No version linkage between video revision and caption revision

A revised GxP training video has been released to the LMS, but the caption file attached to it is the caption from the prior video version. Employees who complete the revised training using the outdated caption have a training record that attests to completion of the revised training, but they received accurate verbal content from the revised video and inaccurate caption content from the prior version — or they received accurate caption content from the revised video but the caption file record does not match. This failure mode requires a systematic review of all LMS content items to confirm that the video version and caption file version are aligned, which is exactly what the LMS migration caption checklist covers in the post on LMS caption checklists for migrations.

Failure mode 4: Glossary not seeded for GxP vocabulary before production begins

Beginning GxP caption production with a general-purpose or empty glossary, then discovering accuracy failures during QA review or, worse, during an inspection review. The error rate on CAPA and deviation training content without a GxP glossary (75–86%) means every GxP training video produced before the glossary was in place has a caption file that requires retroactive accuracy review and likely revision. The fix is to seed the glossary from the category framework before the first GxP caption is generated in production, and to include GxP glossary performance as part of the PQ protocol.

Failure mode 5: Manual revision logs instead of system-generated audit trails

Tracking caption file revisions in a spreadsheet, a Word document, or a shared project management tool rather than relying on a system-generated, time-stamped audit trail. Manual logs can be altered; system-generated audit trails cannot. The regulatory requirement under 21 CFR Part 11 § 11.10(e) is explicit: the audit trail must be “computer-generated” and “secure.” Any organisation relying on manual tracking of caption file changes does not meet this requirement, regardless of how carefully the manual log is maintained.

Failure mode 6: Caption accuracy measurement using vendor-provided scores rather than independent DCMP evaluation

Accepting vendor-provided accuracy scores as the basis for GxP caption file approval rather than performing an independent DCMP-protocol accuracy measurement. Vendor accuracy claims are subject to all the measurement methodology problems documented in the caption vendor accuracy evaluation post: sampling bias toward easier content, pre-corrected tracks, and self-reporting without a reference transcript. For a GxP caption file, the accuracy measurement must be performed using an independently prepared reference transcript and the DCMP error rate formula, with the measurement record included in the caption file documentation package.

Failure mode 7: Caption review performed by L&D staff without GxP vocabulary expertise

The QA review of a GxP caption file requires someone who knows the GxP vocabulary well enough to identify substitution errors in regulatory procedure terms. A general L&D reviewer who can identify grammatical errors and obvious misheard words may not recognise that “the FMEA shows an RPN of 85” has become “the FME-A shows an R-P-N of 85” in the caption, or that “OOS investigation” has become “loose investigation.” The subject matter expert who owns the procedure being trained on — the Quality Systems Manager who owns the CAPA SOP, the Manufacturing Sciences leader who owns the deviation procedure — must be involved in the GxP caption QA review, not just the L&D team. The caption QA methodology post covers the spot-check protocol and the role of SME reviewers in the QA process.

Failure mode 8: Captioning vendor contract without 21 CFR Part 11 provisions

Using an external captioning vendor for GxP training content without a vendor agreement that includes the 21 CFR Part 11 provisions required for a GxP supplier: validation documentation for the captioning system, audit trail output, data handling and deletion terms, and the right to audit the vendor’s quality systems. An external captioning vendor that processes GxP training audio is a supplier of a GxP-regulated service; it must be qualified as a GxP supplier under the organisation’s supplier qualification procedure. A vendor that will not provide IQ/OQ/PQ documentation for their captioning system, that cannot supply an audit trail for the caption processing, or that will not agree to vendor audit rights is not a qualified GxP supplier for caption services. The post on caption vendor SLA and contract review covers the standard contract provisions; the GxP additions — validation documentation requirement, audit trail output requirement, GxP supplier qualification right — should be added as addenda to the standard provision checklist for any vendor used for GxP training content.

FAQ

We use a cloud-based captioning tool from a vendor that provides no GxP documentation. Can we still use it for GxP training?

You can use it if you are prepared to perform the validation work yourselves, but a vendor that provides no GxP documentation is creating significant additional work for your quality team. 21 CFR Part 11 validation is the organisation’s responsibility, not the vendor’s — the FDA does not certify captioning tools as “21 CFR Part 11 compliant.” The validation documentation must be prepared by your quality team, and the evidence on which that documentation is based (system specifications, test records, vendor technical documentation) must be available for regulatory review. A vendor that can provide technical specifications, system architecture documentation, and evidence of testing is easier to validate than one that provides nothing — but neither situation is impossible. The harder question is whether the vendor’s system architecture (particularly around access control, audit trail generation, and data handling) is capable of meeting the Part 11 requirements even in principle. If the vendor uses a multi-tenant cloud architecture where your caption files are co-mingled with other customers’ data without clear data isolation, the access control and audit trail requirements may be structurally difficult to satisfy. This is a vendor qualification question that should be answered before any GxP caption production begins.

Does every pharmaceutical training video require the GxP caption controls, or only the ones explicitly required by regulation?

Only training video for regulated activities requires the 21 CFR Part 11 caption controls. General employee training — HR onboarding, benefits enrollment, cybersecurity awareness, soft skills development, leadership training — does not fall within GxP scope and does not require the additional quality system controls. These videos still require WCAG 2.1 AA-compliant captions for ADA Title I compliance (assuming the organisation has 15 or more employees), but the caption file does not need to be managed as a GxP-controlled document. The practical implementation challenge is maintaining a clear classification: which training content is GxP and which is not. Companies that use the same LMS for both categories must configure the LMS to enforce different content management workflows for GxP vs. non-GxP content.

We have a clinical trial site at an academic medical centre. Does this create GxP captioning obligations for our clinical training video?

Yes. Clinical trial site staff training is GCP training — it is explicitly within GxP scope under ICH E6 (GCP) and 21 CFR Part 312 (IND regulations for clinical trials). Training records for clinical site staff — including training on the specific protocol, on adverse event reporting procedures, on informed consent process, on investigational product handling — are GCP records subject to 21 CFR Part 11 when maintained electronically. If your clinical trial uses an electronic training system (which is nearly universal), the caption files for clinical site training video are electronic records subject to Part 11. The academic medical centre setting does not change the regulatory obligation — the clinical trial site’s obligations under GCP apply regardless of the institutional context.

Our captioning vendor has a SOC 2 Type II report. Is that sufficient for GxP supplier qualification?

SOC 2 Type II is a security and availability certification that covers information security controls, not GxP quality system requirements. A SOC 2 Type II report demonstrates that the vendor has effective controls for data security, availability, processing integrity, and confidentiality — which is relevant to 21 CFR Part 11 access control and data integrity requirements but does not constitute GxP qualification on its own. For GxP supplier qualification, you also need: evidence of system validation capability (can they provide IQ/OQ documentation or support your validation?), audit trail output capability (can the system generate the required audit trail records?), documentation change control (can you ensure the system version you validated is the version you are using?), and the contractual provisions for GxP supplier qualification (audit rights, change notification, quality agreement). SOC 2 Type II is a supporting document in the GxP supplier qualification package, not a substitute for it.

How do we handle caption files for training videos that have been retired and archived?

Caption files for archived GxP training videos must be retained for the same retention period as the associated training records. FDA regulations require training records to be retained for at least 2 years beyond the period during which the trained employee performed the regulated activity, and in some contexts longer (clinical trial records, for example, must be retained for 2 years after an NDA is approved or the study is discontinued). Caption files associated with those training records must be retained and retrievable for the same period. The retention requirement does not change when the video is retired from active training use — the caption file for an archived CAPA training video that employees completed in 2023 must remain retrievable until at least 2025, and longer if those employees continued to work in regulated roles. The caption file should be archived with a status indicating it is no longer in active use but is retained for record purposes.

Our L&D team uses Articulate Storyline to build GxP training modules with embedded captions. Does this create any additional compliance issues?

Yes. Embedding captions in an Articulate Storyline course creates several compliance challenges specific to GxP environments. First, the Storyline course file is the container for the caption content — any revision to the captions requires rebuilding and republishing the course, which constitutes a change to the training module and triggers the LMS content change control process. Second, the caption content is embedded in the published SCORM output in a format that is not easily extracted for independent review or storage as a controlled document — the SRT or VTT file used in Storyline is internal to the course package and not separately accessible in the LMS. Third, caption version history within a Storyline course is not natively tracked in a way that satisfies the 21 CFR Part 11 audit trail requirement — you know which version of the Storyline file is current, but the change history of the caption content within the file requires external documentation. The most manageable approach is to maintain the source SRT or VTT file as a separate controlled document and import it into each version of the Storyline course, so the caption file has its own version history independent of the course file version history.

What happens if we discover that a GxP training video has been in use for 18 months with a caption file at 82% accuracy?

This is a deviation — specifically, a quality system deviation from the document control and accuracy standards required for GxP caption files. The appropriate response is to initiate a deviation investigation: identify the root cause (was the captioning tool unvalidated? was the QA review process not implemented? was the accuracy threshold not defined?), assess the impact (how many employees completed training using the inaccurate caption file, what procedures did those employees work on, and is there any evidence that the caption inaccuracies contributed to procedural non-conformances?), and implement a corrective action (retire the inaccurate caption file, generate and validate a new caption file at 99% accuracy, reassign the training to affected employees if the caption inaccuracies were material to the training content, and document the entire investigation as a CAPA if the impact assessment warrants it). The 18-month timeline also raises a retrospective review question: are there other GxP caption files that were generated by the same uncontrolled process and may have similar accuracy problems? A systematic review of all GxP caption files generated in the same period, against the same DCMP accuracy protocol, is the appropriate follow-up to ensure the deviation was isolated rather than systemic.