At last month’s 5th Pharmacovigilance & Drug Safety Conference in Madrid, Sandoz researchers took the stage to present something that should stop every drug safety team in its tracks: AI models trained to mine adverse event signals from social media, scanning unstructured, informal, patient-generated content for the kind of real-world safety data that never makes it into a clinical trial report.

The technology is impressive. The problem it reveals is even more important.

“A safety signal only matters if it gets captured, understood, and acted on. Detecting it is only the first step — what happens next determines whether it protects a patient.”

Because here’s what those social media feeds look like in practice: patients in Brazil posting about side effects in Portuguese. Forum threads in Japanese about treatment experiences. Arabic-language health communities discussing symptoms. Turkish patient groups sharing stories that would never appear in a clinical database. Advanced pharmacovigilance (PV) teams are increasingly equipped to detect these signals across languages, but capturing the signal is only step one. Ensuring it is accurately understood, correctly classified, and properly documented in a compliant, regulatory-ready format is where the real complexity begins.

The Real-World Data Challenge Is Multilingual by Definition

The Madrid conference made clear that real-world evidence is no longer a supplementary data source, it’s central to modern pharmacovigilance. Electronic health records (EHR), insurance claims, wearable device data, patient registries: these are the inputs that give regulators and manufacturers a picture of how medicines actually perform in the real world, across real populations.

And those populations are global. An EHR system in South Korea is documented in Korean. Claims data from Brazil is in Portuguese. Wearable data collected in the Gulf region arrives with Arabic-language notes. The data richness that makes real-world evidence so valuable is inseparable from its linguistic diversity.

Consider This Scenario

A drug is approved and marketed across 40 countries. Your PV team detects an adverse event signal from a patient forum in a non-English language — exactly as it should. The report is triaged and passed for translation. But the translation vendor isn’t a specialist in PV. A terminology nuance is lost. The case narrative reaches submission with an inconsistency that only surfaces during a regulatory inspection six months later. The signal was found. The documentation didn’t hold up.

This isn’t a hypothetical. It’s the operational reality for any pharmaceutical company with a global portfolio, and it’s precisely why multilingual capability has moved from a translation line item to a patient safety imperative.

What “Adverse Events in Every Language” Actually Requires

Handling safety content across languages isn’t the same as general translation. Adverse event reports, ICSRs, medical summaries, and safety narratives require terminology precision, MedDRA coding, drug and symptom nomenclature, regulatory-specific language, that demands specialized expertise. A mistranslation in a case narrative isn’t just a quality issue. In a regulatory submission, it can be a compliance failure.

The patients submitting these reports speak naturally, informally, and in their own language. Capturing that signal accurately and converting it into actionable, codable safety information requires human linguistic expertise working alongside the technology, not instead of it.

The languages generating the most real-world safety data today include:

• Mandarin Chinese, Spanish, Portuguese, Japanese, Arabic
• German, French, Korean, Turkish, Hindi, Italian, Russian
• And dozens more across Asia-Pacific, Eastern Europe, the Middle East, and Latin America

Multilingual Isn’t a Feature. It’s the Foundation.

The Madrid conference signaled a clear direction: pharmacovigilance is getting smarter, faster, and more automated. AI helps handle case intake, signal detection, and narrative generation at a scale that was unimaginable five years ago. That’s an enormously positive development for patient safety, but only if the language layer keeps pace.

An AI model that detects signals across social media and real-world sources is a powerful tool. An AI model whose output gets reviewed, translated, and validated by specialists who understand both the language and the regulatory context — that’s a complete safety surveillance capability.

The gap between those two things is where multilingual expertise matters most.

What Welo Life Sciences Brings to the Table

Welo Life Sciences is a specialized language and technology partner for pharmaceutical, CRO, and medical device organizations operating in regulated global markets. We support the multilingual layer that modern pharmacovigilance programs require, from translation and review of adverse event reports submitted in non-English languages, to the precise, terminology-accurate documentation that PV compliance demands: Lab Reports, Periodic Safety Update Reports (PSURs), CIOMS Reports, and Product Safety Reports.