SeltaSquare

1. 
2. 

• Automating rule-based tasks such as literature search, data entry, and MedDRA coding
• SeltaSquare connects safety data from intake through analysis and reporting
• "Beware of AI false negatives — final judgment still requires PV expert review"

(From left) CEO Shin Min-kyung, PV Center Director Lee Jeong-min, PVX Center Director Han Kyung-hee.

In an interview with Yakup News, SeltaSquare representatives emphasized that alongside the expanding use of AI in pharmacovigilance (PV), final expert review and judgment remain essential.

AI adoption in pharmacovigilance (PV) is expanding primarily in rule-based, repetitive tasks — literature searches, safety data entry, and MedDRA coding. The goal isn't to replace people, but to let PV experts spend more time on signal analysis and medical judgment as the volume of safety information (from clinical trials, post-marketing reports, medical literature, partner data, EMRs, and RWD) keeps growing.

CEO Shin Min-kyung noted that AI currently works best on clearly defined, repetitive tasks — like literature searching and intake of PDF-based safety reports into databases — and that freeing up time from search/entry lets teams focus more on medical review and safety assessment.

Literature monitoring: For products with heavy publication volume (e.g., botulinum toxin), literature review becomes a recurring burden that's hard for one or two staff to manage. SeltaSquare's literature monitoring solution LITUS supports periodic searches, deduplication, results management, and audit trails, with AI detecting safety-relevant information within retrieved papers. PV Center Director Lee Jeong-min explained that high product counts translate into heavy, recurring review workloads. Shin added that a feature letting AI scan uploaded PDFs of paywalled papers for adverse event data is planned for beta release in Q3 2026.

Data accuracy over speed

As AI use expands into safety data entry, MedDRA coding, quality review, and regulatory reporting, data accuracy at the outset matters more than processing speed — inconsistent coding of the same medical concept can distort signal detection and downstream assessments (PSURs, RMPs). To address this, SeltaSquare is building an integrated platform, iVigilance Square, connecting literature monitoring, data standardization, ICSR processing, evaluation, analysis, and regulatory reporting into one continuous data flow. PVX Center Director Han Kyung-hee emphasized that the goal isn't automating individual tasks in isolation, but keeping data connected across the entire PV lifecycle — this "data orchestration" approach supports more consistent analysis and decision-making, freeing experts to focus on risk assessment.

False negatives are the real risk

Unlike false positives (which just add review work), false negatives — AI missing a reportable case — risk both regulatory non-compliance and patient safety gaps. Lee stressed the need to assume AI can be wrong, and that SeltaSquare follows a Human-in-the-Loop approach where PV experts always give final review and judgment on AI outputs, verified against both technical accuracy and PV/medical context. Shin added that what matters most is a team's ability to accurately define real-world problems — not just having PV experts nominally "involved" — since even simple-looking data flows in PV are entangled with regulation and medical judgment at every step.

RWD as a safety evidence source

The next challenge is leveraging real-world data (EMRs, insurance claims) as safety evidence. Unlike efficacy-focused RWD studies, safety information often sits in unstructured free-text clinical notes, requiring dedicated standardization and NLP techniques to extract adverse event/side-effect data and link it to full-cycle PV data — something Shin called a key task for RWD-based safety research going forward.

Beyond outsourcing — a data decision-making partner

Adopting AI tools alone doesn't automatically improve PV capability; companies still need operating standards for what tasks to automate, who reviews outputs, and how errors get corrected. For small/mid-size pharma and biotech companies, running full in-house PV is difficult, but control over safety data and final decision-making should stay in-house even when using external partners. Shin positioned SeltaSquare's goal as being a data-driven decision-making partner rather than a simple task-outsourcing vendor, helping companies build systematic PV foundations and internalize safety management capability over time.

SeltaSquare plans to strengthen its position as an AI/DX specialist in pharma/biotech over the next 3–5 years and expand into global markets, aiming to be a platform that connects information across the full clinical-to-post-marketing lifecycle to support better decisions — not just a repository for accumulated safety data.

The Original Article: [약업신문]셀타스퀘어, AI가 바꾸는 약물감시 반복 업무 줄이고 ‘안전성 판단’에 집중下