A supplier tells you they are moving production of an intermediate to a different site. Somebody now has to work out what that means.

The assessment that follows is one of the least automated things in regulatory affairs. Which of our products use that site? Which markets are those products registered in? What have we actually filed that names the site, and is any of it still in force? And then the harder half: what filing does each country oblige — a prior approval supplement, a Type IA variation, a partial change application, a notification?

Most organisations answer this with a senior person, a spreadsheet and institutional memory. That works until the person is on holiday, and it never leaves a trail showing how the conclusion was reached.

The mistake is treating it as one question

There are two halves to a change impact assessment. They have different owners, different evidence and different ways of going wrong.

What do we have? A fact about your organisation. Which sites make which substances, which products contain them, which applications they are filed under, what is in force, where you hold an authorisation. It is answerable from your own information — and where it cannot be answered, that itself is a finding.

What does that oblige? A fact about regulation. It comes from variation regulations and agency guidance, and it changes when those change, not when your own information changes.

Software that blurs the two produces the worst possible output: a confident regulatory conclusion resting on information that was silently incomplete. DnXT keeps them apart, and keeps them visibly apart on the page.

The first half: your own information, and nothing else

Ask about a manufacturer, substance or product and the platform works through what you already hold:

  • The supplier relationships — what this site makes, who else makes it, and which supplier is currently named in your filings.
  • The filed documents whose section context names it. Not “we think this site appears somewhere in Module 3”, but the specific sections, in these sequences, that name it.
  • The later filings that superseded some of those documents, so what comes back is what is actually in force — decided by the submission’s own record of what replaced what, never by date order.
  • The dossiers and countries those sequences were filed in.
  • The market registrations that say where the affected products are actually authorised.

Every line is labelled: known where the platform could trace it, or gap where it cannot say — and a gap always explains why.

Why the gaps are the most valuable output

A manufacturer with no country, no FEI and no DUNS number on file cannot be matched to a regulatory establishment. A dossier that has never been indexed cannot contribute filed evidence. A product with no registration cannot tell you which markets are affected. A country with nothing filed and nothing registered is not evidence of no impact — it is evidence of no information.

In every one of those cases, a naive system returns a shorter list and looks more confident. That is exactly backwards. An incomplete answer that presents itself as complete is how a change gets under-filed, and under-filing is what turns a routine variation into a compliance finding.

So the assessment names its own blind spots, and each gap names the thing that would close it. “This site has no FEI or DUNS on file” is a task somebody can pick up. Silence is not.

The second half: a rule library that cites its sources

The obligation side lives in DnXT’s regulatory intelligence capability as a maintained library of change-classification rules — covering the US, EU and Japan for drug-substance site changes, drug-product site changes and drug-substance specification changes, with the alternatives available and the CTD sections each one requires you to revise.

Three principles govern how it answers:

  • Every rule cites its sources. A rule backed by two independent sources is marked verified. One backed by a single source is flagged as such, so whoever reads it treats it as an assertion rather than a settled fact.
  • Open questions stay visible. Where guidance is genuinely ambiguous, the ambiguity is recorded rather than quietly resolved by whoever wrote the entry.
  • No rule means “we do not know”, never silence. If the library has nothing for a given change in a given country, it says so — because a missing rule must never read as an absent obligation. Software that returns an empty answer for both “no filing required” and “we have no rule for this” is worse than no software at all.

Joined together, never merged

The planning workspace puts the two halves together in a “what changes if…” panel: pick a manufacturer, substance or product and a country, and read the assessment.

The labels stay visible throughout. The known lines are your own filings and registrations. The rule lines are what each country obliges — for the country you asked about, and for every other country where you have filed evidence, because that is usually the set somebody had forgotten. The gap lines are where neither half can answer.

Underneath, the in-force evidence is broken out by CTD section, alongside the replacements already recorded and the registrations currently in force.

What it replaces, and what it does not

It does not make the regulatory decision. A change classification is a judgement, it gets signed by a person, and it should.

What it removes is the two days of reconstruction before the judgement — and the quiet risk that the reconstruction missed a market. The assessment starts from the same evidence an inspector would use: what you filed, where you filed it, and what is still standing.

DnXT builds eCTD publishing, submission planning, document management and dossier review software for regulatory operations teams. Book a demo to see change impact assessed from your own filings against your own submissions.