SeltaSquare

1. 
2. 

SeltaSquare identifies industry shifts early and delivers meaningful insights to our clients.

At the Korean Society for Pharmacoepidemiology and Risk Management (KoPERM) Spring Conference, held May 28–29, key issues in drug safety, Signal Detection operational strategy, and region-based active pharmacovigilance systems were explored in depth, covering the latest research trends and practical applications across pharmacoepidemiology and pharmacovigilance.

This content summarizes the latest trends in data-driven Signal Detection and global pharmaceutical companies' operational strategies, curated for Drug Safety and PV professionals.

On May 29, a session titled "Signal Detection Based on Pharmacovigilance Data" featured a researcher from the Korea Institute of Drug Safety and Risk Management(KIDS) Safety Information Management Team and a pharmacovigilance representative from Shinra Zen, who discussed, through real-world case studies, everything from how domestic spontaneous adverse event reports are utilized to the operational Signal Detection strategies of global pharmaceutical companies.

 

Background: Pre-Market Clinical Trials Alone Are Not Enough

Pre-market clinical trials for pharmaceuticals have structural limitations:

• Too few subjects
• A population skewed toward median age
• Too brief a period of observation
• Too simple a design
• Too narrow a range of exposure
• Too indirect a surrogate

This gap widens as the incidence of an adverse drug reaction decreases. Based on a trial of 2,000 subjects, an adverse drug reaction with an incidence rate of 1/50,000 has only a 4% probability of being detected during the clinical trial.

As a result, systematically collecting and analyzing Individual Case Safety Reports(ICSRs) accumulated post-marketing to detect and manage potential safety signals early has emerged as a core challenge in pharmacovigilance. An ideal PV system should be able to promptly and systematically collect and review adverse event and safety information to detect new or changed risks, assess the likelihood of causal relationships with the drug and the clinical significance of the risk based on individual cases and accumulated evidence, characterize or quantify the frequency, severity, and risk factors of the risk when necessary, and translate these findings into risk minimization measures, regulatory actions, and safety communications.

The global pharmacovigilance environment is likewise evolving beyond simple adverse event collection toward the early detection and management of safety signals based on accumulated ICSR data. EMA GVP Module IX presents a Signal Management framework spanning Signal Detection, Validation, Confirmation, Analysis and Prioritisation, and Assessment and Recommendation for Action, treating signal detection not as a standalone analysis but as part of an ongoing safety management process. In Korea as well, products for which an RMP has been submitted are required to periodically report the results of signal analysis on collected safety information.

What Is a Signal, and How Is It Detected?

In 2002, the WHO defined a signal as "information that arises from one or more sources(including observations and experiments), which suggests a new potentially causal association, or a new aspect of a known association, between an intervention and an event or set of related events, either adverse or beneficial, that is judged to be of sufficient likelihood to justify verificatory action." Importantly, a signal does not indicate a direct causal relationship — it is a hypothesis that justifies further evaluation.

The KIDS researcher presented two core approaches to Signal Detection at the session:

• Qualitative Approach (Traditional PV Method): Signals are detected through expert assessment of individual cases or case series. This approach is particularly effective for evaluating cases involving Designated Medical Events(DMEs) — rare and serious adverse events that may be considered potential signals even with just 1–3 reports — and it requires the reviewer's clinical and pharmacological background knowledge.
• Quantitative Approach (Data Mining): Computerized algorithms statistically detect disproportionate drug–adverse event combinations within large-scale safety databases. Representative metrics include PRR(Proportional Reporting Ratio), ROR(Reporting Odds Ratio), IC(Information Component), and EBGM(Empirical Bayesian Geometric Mean), which are used to analyze large public databases such as FAERS and EudraVigilance. However, disproportionality alone cannot establish causality — it must always be paired with clinical review.

Detected signals are prioritized using the WHO-UMC Triage Algorithm, based on criteria such as ADR unexpectedness, ADR seriousness, disproportionality, rate of reporting increase, whether the drug is new, and whether the signal has been identified internationally. Signals are then evaluated through review of the original source data(ICSR and case series causality assessment) and cross-verification against multiple sources, including overseas approval information, the UMC database, and the literature. Significant safety findings are subsequently translated into label changes or continued monitoring.

From 2016 to 2025, KIDS published a total of 18 items in the WHO newsletter and 2 in VigiLyze Signal, providing domestic pharmacovigilance-data-based signal analysis results to roughly 190 member countries of the international drug monitoring program. Through this, Korea has helped fill a gap in ethnicity-specific adverse event information that had been relatively underrepresented in the global Signal Detection system for the Asia region, strengthening the international credibility and representativeness of Korean PV data. Furthermore, given that signals first identified in Korea can lead to early warnings for regulatory authorities worldwide, this demonstrates that Korea's pharmacovigilance capabilities are making a tangible contribution to strengthening global patient safety.

In Part 2, we will cover how global pharmaceutical companies operate their Signal Detection processes.