Active ingredient

Without regulatory affairs at the centre of pharma’s product data infrastructure, the rest of the formulation doesn’t hold

Ask most pharmaceutical companies whether they have a single authoritative source of product data and the honest answer is no.

Every pharma function has traditionally run its own systems, built for its own respective purposes.

Still today, in most organisations, clinical, manufacturing, quality and pharmacovigilance teams each hold a version of the product record that has evolved separately, reconciled manually when necessary and rarely tested against what the neighbouring team actually has.

Take a project to connect regulatory and PV systems – reasonably scoped as a contained data-sharing exercise.

Some way in, though, the team discovers that the product information each side holds bears only a passing resemblance to what the other side is working from.
 
Years of separate maintenance have resulted in two parallel records, neither of them wrong exactly yet irreconcilable without an active decision on which one to trust.

Because PV data feeds into signal detection platforms and aggregate safety reporting, that decision turns out to have consequences across a chain of downstream systems – in some cases running to multiples of ten.

None of this is down to the technology or the people involved. Rather, no function has ever taken responsibility for what the product record should look like; meanwhile no process has kept the answer consistent.

For years, regulatory affairs absorbed these inconsistencies, situated at the end of the data chain, collecting what came in from every function, checking it for coherence and getting it out to a health authority.

The manual reconciliation work may have been considerable, but largely invisible — hidden inside the submission process, surfacing as delay or late-stage stress rather than as a structural problem.

Today, structured submissions are making that problem visible. As agencies move from narrative documents to machine-readable data, algorithmic checks will surface discrepancies that a well-staffed RA team once caught and resolved before anyone upstream had chance to notice it.

EMA’s Product Management Service, the EU’s primary vehicle for IDMP implementation, requires structured product data from this year onwards, starting with products on the Union List of Critical Medicines and extending across non-centrally authorised products shortly after.

EMA has already sought formal assurances from industry associations that the deadlines will be met, suggesting it does not entirely trust that they will.
ISO IDMP provides the technical vocabulary to address the underlying problem, not just the compliance requirement.

IDMP-compliant data carries the same meaning in a quality system as it does in a clinical database or a PV platform – the same substance identifier follows a compound from early development through manufacturing, labelling and adverse event reporting without being re-interpreted at every handover.

Used properly, it provides the architecture for a single product record rather than a collection of functional approximations.

The argument for fixing data anomalies has historically been compliance-led, which has not always garnered real motivation.

A more persuasive argument, at least in organisations with ambitious AI programmes, is that inconsistent product data will make such ambitions fail.

AI tools fed conflicting versions of the same data point do not simply produce imperfect results. They hallucinate – generating outputs that look entirely plausible but are constructed from contradictory source material.

An AI system trained on a product record where the same compound carries different identifiers in regulatory and PV will confidently produce nonsense.

Fixing the data architecture after the tools are deployed is far more expensive than doing it first. For organisations serious about using AI across regulatory intelligence, signal detection or R&D productivity, IDMP-aligned product data is a prerequisite, not an optional upgrade.

Regulatory affairs is uniquely positioned to take this on. If it does not, someone else might fill the gap and probably do a worse job.

After all, no other function assembles as complete a picture of a medicine across its life cycle: RA already collects, standardises and checks product information for external submission.

The extension of that role to internal data distribution is a short step in logic, even if the organisational change required is more significant.

The vital connectors are data stewards, anchored in RA but embedded across adjacent functions – people with the requisite data literacy to spot where the same product element exists in two different forms and understand what that means for every system downstream.

The starting point for transformation will inevitably vary by organisation. For development-focused companies, the most pressing connection is likely to be between RA and PV. For generics manufacturers, the greater affinity may need to be between RA and quality.

Whatever the starting point, a staged approach is generally easier to justify than wholesale transformation.

Before taking action, companies should assess how much of their existing data actually needs to be retained and standardised. Not all product data serves a continuing purpose.

The benefits of doing the data work, and of appointing RA to champion it, extend beyond any individual organisation.

Regulators working from IDMP-structured submissions will be able to share product data across agencies more efficiently, while companies co-producing combination therapies will be able to collaborate more fluidly when both sides’ product records are operating to the same standard.

As the function that already collects and standardises product information for every submission it makes, RA’s opportunity now is to harness that same infrastructure to serve the rest of the organisation with the latest product intelligence in real time, building on compliance investment their companies are already making.