Summary
- The problem
- Module 3 tables and narratives are assembled by copying results out of laboratory, stability and manufacturing systems. When a timepoint arrives or a specification changes, nobody can easily show which submitted content used which data cut, or what else it affects.
- Seal’s approach
- CMC content keeps a reproducible connection to the protocols, execution records, analytical results, specifications and process definitions behind it. A table uses a defined source population, cutoff and transformation, and each submitted version keeps the data that produced it.
- Neil
- Seal’s AI agent prepares the candidate data cut, a cited draft that keeps open results visible, and the list of content a method or specification change touches. Scientists and regulatory owners decide what the content claims and release it. About Neil.
- What changes
- A refresh is prepared as a new candidate data cut beside the prior submitted artefact, with pending or excluded results still visible. When a source changes, candidate content and market impacts go to accountable owners to assess.
- What it covers
- Recipes, process development, tech transfer, stability, specifications, analytical methods, deviations, post-approval change and regulatory submissions. All capabilities.
- Where to start
- One CMC table or section that changes often, such as a stability table awaiting its next timepoint, with the sources behind it. Book a demo.
1A document system stores the dossier. Seal keeps it tied to the evidence.
Module 3 tables and narratives are usually assembled by copying results out of laboratory, stability and manufacturing systems. When the next timepoint arrives or a specification changes, nobody can easily show which submitted content used which data, or what else the change affects.
| A document-based system | Seal | |
|---|---|---|
| Stability table | Results copied from exports | A data cut of reviewed results |
| New timepoint | The table retyped | A proposed refresh, reviewed |
| Pending result | Easy to miss | Shown as pending |
| What was submitted | A file to find | Kept with its data cut |
| Method change | Found by searching | Affected content listed |
| Process description | A separate paraphrase | Prepared from the process |
| Markets | A copy edited per region | Shared evidence, own positions |
| Authority questions | Reconstructed | Answered from the sources |
| AI | Text to paste elsewhere | Neil drafts with sources |
Most CMC work is supported by a document system for the dossier and separate LIMS, stability and batch systems for the data behind it, joined by copy and paste. The document system records what was filed. The link between a submitted statement and the evidence behind it becomes a matter of memory.
Chemistry, Manufacturing and Controls content describes what the product is, how it is made and how it is controlled. Its tables and narratives draw on stability studies, analytical results, specifications, characterisation and process definitions, and each of those keeps changing after the content is first written.
Seal connects CMC work to the protocols, execution records, analytical results, specifications and process definitions behind it. A CMC table uses a defined source population, cutoff and transformation. Narrative drafts retain the references used. Because the source records and the content share one platform, when new results arrive an author can ask Neil to draft a refreshed table or narrative, then review and approve it as a new version while the submitted one stays fixed.
A shared ontology defines how drug substances, drug products, specifications, methods and studies relate. A specification links to the methods that measure it, the stability data that support it and the characterisation that justifies it. CMC content follows those relationships to the selected source versions behind each claim, and authors assess scientific meaning and limitations before releasing it.
2Refresh the data. Keep what was submitted.
A new timepoint is prepared as a new data cut beside the submitted one. A later result must not silently rewrite what was previously submitted.
- Protocol
- STB-201 v02
- Method
- AM-022 v03
- Timepoint
- 12 months
- Result
- Not released
Pending in cut 02. No earlier value carried forward.
| A data cut keeps | In the example |
|---|---|
| Protocol | STB-201 v02 |
| Population | DP-201, DP-202 and DP-203 |
| Data-lock point | Cut 02 |
| Results | Reviewed results only |
| Method | AM-022 v03 |
| Pending | DP-203 at 12 months, shown |
Stability tables reproduce from versioned protocols, population rules, data-lock points, approved results and transformations. Each submitted version keeps the data cut that produced it, so the table can be reproduced as it was submitted. Module 3 asks for stability results to be presented with information on the analytical procedures used to generate them,¹ so each result keeps its method version as well as its value.
For a product with a shelf life of at least 12 months, long-term stability testing is normally every three months over the first year, every six months over the second and annually after that.² A stability table therefore changes several times while an application is under review and after approval. Each refresh is prepared as a new candidate data cut, and the content it affects is assessed before anything is released.
Pending, missing or excluded results remain visible with their status and rationale; they are not silently replaced with an earlier value. The table uses each reviewed result with the method version that produced it, and does not substitute a later record. Where no reviewed value is available, the gap stays visible and no earlier timepoint is carried forward.
Figure 1 compares an assay table through nine months with a candidate twelve-month refresh. Two of three twelve-month results are reviewed; DP-203 remains under review, and its result is not included as a reviewed value. This is a fictional excerpt, not a complete stability assessment or a shelf-life conclusion. Assay alone does not establish shelf life.
3Neil drafts from the evidence. Your team decides what it says.
Neil, Seal’s AI agent, prepares the candidate data cut, the draft and its sources. Scientists and regulatory and quality owners decide what the content claims, and release it.
Prepare the stability update. Can we say all three batches have reviewed 12-month results?
Neil
- Not from cut 02: DP-201 and DP-202 are reviewed; DP-203 is pending.
- Draft: reviewed 12-month assay results for DP-201 (98.7%) and DP-202 (98.4%).
- DP-203 is stated as under review.
- Sources: DP-201-12M, DP-202-12M, DP-203-12M.
- Sources (met)Cited, with versions
- DP-203 (not met)Awaiting QC review
- Release (not met)For your team
| Neil prepares | Your team decides | |
|---|---|---|
| Stability refresh | The candidate data cut | Whether the results are suitable |
| Narrative | A cited draft and its limits | What the content claims |
| References | Versions that do not match | Which version is right |
| Method change | The content it touches | What changes, and when |
| Markets | Markets with affected content | Which changes apply where |
| Authority question | The claim and its sources | The response |
Asked whether all three batches have reviewed 12-month assay results, Neil answers from the proposed refresh: DP-201 and DP-202 are reviewed, and DP-203 is still pending. The draft says so. At data cut 02, reviewed 12-month assay results were available for DP-201 (98.7%) and DP-202 (98.4%); the 12-month result for DP-203 remained under review. Each statement cites the result it came from, and the open question stays attached to the text your team reviews.
The next step belongs to QC: resolve the outstanding result, then reassess the table and narrative. Neil can help retrieve permitted evidence, identify inconsistent references and prepare cited drafts or follow-up work. Scientists review source suitability and interpretation. Regulatory and quality owners review the position and release of content. AI output does not approve a result, shelf life or submission.
In line with the EU’s draft GMP Annex 22 on AI, Neil decides nothing in GMP work. It helps set up configuration, which your team verifies and releases under change control. Customer data is not used to train AI models, and Neil reads only what the person asking may read. See how Neil fits Annex 22 and more about Neil.
4Change the method. Find the affected work.
A proposed method revision can affect a description, results already reported and future testing differently. Seal lists each connection for its owners to assess before anything changes.
v03v04
- 3.2.P.5.2 (not met)Review the next version
- 3.2.P.8.3 (met)Keeps v03, as tested
- STB-201 (not met)Agree when it applies
| Affected work | Decision | Review with |
|---|---|---|
| Method description | Review the next content version | Analytical development, CMC |
| Reported results | Keep v03, the method used | Stability, CMC |
| Upcoming testing | Agree when the change applies | QC, stability, quality |
The method description in 3.2.P.5.2 is a proposed reference. Compare the proposed procedure with the method described in the current content, and identify changed instructions, the supporting development evidence and any validation work needed before the reference can change. Retain the procedure comparison, change rationale and supporting analytical assessment. A change to an analytical procedure is supported by a bridging strategy that shows the revised procedure remains fit for its intended purpose.³
Stability results already reported in 3.2.P.8.3 keep their original method reference. They were generated with AM-022 v03; preserve that reference and the submitted data cut. A proposed v04 does not relabel earlier results; assess comparability if later results use a different procedure. Retain the result-to-method links, the submitted table and any justified comparability assessment.
Upcoming testing under STB-201 needs an authorised method and an agreed effective date. Check which future tests could use the revised method. Agree the applicable protocol, training and effective date through change control, and keep the existing execution instructions in force until the change is authorised. Retain the impact assessment, protocol decision, training assessment and implementation approval.
The example shows three connections, not a complete impact assessment. Your team also considers affected products, specifications, sites, markets and commitments. Where established conditions have been agreed, any change to them needs a submission to the regulatory authority.⁴ Regulatory owners determine the reporting route and implementation conditions; a linked record is not approval to use a new method.
Authority questions are answered from the same records. Bring the submitted claim, its source records and any new evidence into one response, citing the data cut, source versions and review behind the original content rather than a reconstruction. Authorship, review and the final response stay with the work.
5Keep specifications, processes and markets consistent.
Each Module 3 section is prepared from the versioned records it describes. The laboratory, the submission and each market’s approval can hold different versions, and Seal keeps them distinct.
- Laboratory (met)v3.2 effective
- US (met)v3.2 approved
- EU (not met)v3.1 approved, v3.2 submitted
| Section | Content | Prepared from |
|---|---|---|
| 3.2.S.2, 3.2.P.3 | Manufacture | Process definitions and runs |
| 3.2.S.3 | Characterisation | Structure, impurities, properties |
| 3.2.S.4, 3.2.P.5 | Control | Specifications, method versions |
| 3.2.S.6, 3.2.P.7 | Container closure | Packaging and E&L studies |
| 3.2.S.7, 3.2.P.8 | Stability | Protocols, data cuts, results |
Link specification limits and method references to the source versions used. Specification history is preserved, and intended references can be checked to resolve to the same approved version before content approval, so an inconsistent reference is raised for review instead of assuming “latest” is correct. A drug substance release specification revised on process validation data keeps its earlier versions and the evidence for each.
Prepare process descriptions from selected process definitions and execution evidence, with technical review of terminology and meaning. Module 3 describes a process by its steps, process parameters and equipment.¹ In Seal the description uses the same unit operations, parameters and equipment as the process that is run, rather than a separately maintained paraphrase, and scientists check the meaning and any differences from the submitted description.
When the process itself changes, the description, specifications and stability evidence change with it. For a biological product, a comparability exercise shows that the quality attributes of pre-change and post-change product are highly similar, not necessarily identical, and that existing knowledge is sufficiently predictive that any differences have no adverse impact on safety or efficacy.⁵ Keep the pre-change and post-change populations explicit, with the batches, results and criteria behind the comparison. The post-approval change blueprint follows that work market by market.
Operational, submitted and approved states should remain distinct. A specification can be effective in the laboratory while the submitted version is still the previous one, and a market can still be operating under an earlier approval. Keeping those states separate is what makes a later question answerable: which data supported the claim, when and in which market. A site change, specification update or process modification remains traceable to the approval it modifies.
Reuse common evidence across markets while retaining each market’s requirements, regulatory position, source versions, review and submission history. Regional differences can affect substance and implementation, not just formatting. In the US, each change to a condition established in an approved application is reported to FDA, by supplement or in the annual report depending on the change;⁶⁷ the EU classifies variations by their potential impact on quality, safety or efficacy.⁸ Regulatory owners assess which changes apply to each market.
6Start where the evidence is produced.
Run protocols and review results in Seal, or connect records from the systems you already use. CMC content keeps the link to the execution instead of starting again from a copied table.
| Stage | Where it runs | What it holds |
|---|---|---|
| Study execution | In Seal or connected | Protocol, method, results |
| Controlled content | In Seal | Data cut, table, review |
| Publishing | Your publishing tools | Assembly and transmission |
Configure sampling, test methods, calculations and review workflows in Seal, as for stability studies, and keep the protocol version, actual results and scientific review with each record. Select the source population and data cut, and prepare the table, narrative and scientific review together as controlled documents. Release reviewed content to your publishing workflow and submission operations, keeping assembly, technical validation and transmission distinct.
Source records can remain in your document, laboratory or manufacturing systems. Agree the connection, permitted records and actions for each source, and preserve source identifiers, versions and review states when preparing content in Seal. Test the configured workflow against its intended use before relying on it.
| Where the record lives | What CMC content uses | Connections include |
|---|---|---|
| LIMS | Samples, tests and results | LabWare, LabVantage |
| Analytical systems | Original files and results | Waters Empower, Chromeleon |
| MES and historians | Batch steps, process values | Tulip, Emerson DeltaV |
| Document systems | Controlled documents | Veeva Vault, SharePoint |
| ERP | Materials and lots | SAP, NetSuite |
Available APIs, files and source permissions determine the implementation, and you do not need to replace these systems to start. See all integrations.
Module 3 content is composed from CTD sections, each linked to the source data it uses. The content carries the selected source versions, review and publishing handoff. It remains controlled content, not the submission package itself.
7Keep specialist publishing in place.
Seal supports evidence, controlled content and review. Dossier assembly, technical validation and transmission remain part of your publishing workflow.
Agree the output format and handoff with regulatory operations; a draft table is not a submission package. Released content carries its selected source versions and review, so its basis can still be shown after it is filed.
Start with one CMC table or section that changes often, such as a stability table awaiting its next timepoint, with the sources behind it. Reproduce the current version from its data cut, prepare a candidate refresh and review the difference before extending the approach across Module 3.
References
- 1ICH M4Q(R1), The Common Technical Document for the Registration of Pharmaceuticals for Human Use: Quality (2002). Section 3.2.P.3.3 describes the process by its steps, process parameters and equipment; 3.2.P.8.3 presents stability results with the analytical procedures used to generate them. ICH
- 2ICH Q1A(R2), Stability Testing of New Drug Substances and Products (2003), section 2.2.6: long-term testing normally every 3 months over the first year, every 6 months over the second year and annually thereafter. ICH
- 3ICH Q14, Analytical Procedure Development (2023), chapter 7: a bridging strategy demonstrates that a revised or new procedure remains fit for its intended purpose. ICH
- 4ICH Q12, Technical and Regulatory Considerations for Pharmaceutical Product Lifecycle Management (2019), section 3.2.1: established conditions are legally binding, and any change to them necessitates a submission to the regulatory authority. ICH
- 5ICH Q5E, Comparability of Biotechnological/Biological Products Subject to Changes in Their Manufacturing Process (2004), section 1.4. ICH
- 621 CFR 314.70, Supplements and other changes to an approved NDA: the applicant must notify FDA about each change in each condition established in an approved NDA beyond the variations already provided for in it. eCFR
- 721 CFR 601.12, Changes to an approved application: the applicant must inform FDA about each change in the product, production process, quality controls, equipment, facilities, responsible personnel or labeling established in the approved license application. eCFR
- 8Commission Regulation (EC) No 1234/2008, concerning the examination of variations to the terms of marketing authorisations, Article 2: a minor variation of type IA has a minimal impact, or none, on quality, safety or efficacy; a major variation of type II may have a significant impact. EUR-Lex
ACapabilities
| Capability | What it covers |
|---|---|
| Limits that reference their evidence | A recipe step's limit references the study and records it rests on, pinned to the version it was set on. Anyone with access opens the same fields, and sees what references them. |
| Answers that link their records | Neil answers CMC questions from the records the person asking can read, and links the records and versions it read, for you to open. What they don't show, it says. |
| One question, asked of every row | A Neil column asks one question of every study, lot or deviation in a view, and of the text of its files, by a rule anyone can read. Each answer opens to the field, or the file and page, in its own words. |
| Files from before Seal | Reports come in attached to the records they belong to, or are captured from SharePoint, Box, Google Drive or Veeva Vault at their version. Neil reads their text; a Neil column can't read a scan. |
| Module 3 sections, drafted | Neil drafts a section, a table or a response from the records, with their sources, for your people to edit, sign and send. Publishing stays in your own tools. |
| Versions and their history | Recipes, specifications, methods and studies are versioned. A change goes through a changeset, a change to signed content keeps its reason, and every lot keeps the versions it was made to. |
| Stability studies | The protocol, each physical pull and the original result, together for the life of a stability study. Neil prepares the next timepoint’s work. |
| Technology transfer | The receiving site starts from a copy of the configured process, not a document package. Neil prepares the gap assessment and the local changes. |
