Question 1

What is pharmacovigilance and what are its regulatory requirements?

Accepted Answer

Pharmacovigilance is the science and activities relating to the detection, assessment, understanding, and prevention of adverse effects or any other drug-related problems. Regulatory requirements vary by jurisdiction: in the EU, pharmacovigilance is governed by Directive 2001/83/EC, Regulation (EU) No 1235/2010, and Good Pharmacovigilance Practice (GVP) Modules I–XVI, administered by the EMA and national competent authorities (NCAs). In the US, post-marketing safety reporting is governed by 21 CFR Part 314 and Part 600, with expedited reporting (15-day IND safety reports, 7-day fatal SAR) and periodic safety reports (PSURs/PBRERs). In Australia, the TGA requires adverse event reporting under the Therapeutic Goods (Standard for the Uniform Scheduling of Medicines and Poisons) Order. Marketing authorisation holders (MAHs) must maintain a pharmacovigilance system, appoint a qualified person for pharmacovigilance (QPPV), and submit Periodic Benefit-Risk Evaluation Reports (PBRERs) at defined intervals.

Question 2

What is the EMA DARWIN EU programme and how does it change pharmacovigilance?

Accepted Answer

DARWIN EU (Data Analysis and Real World Interrogation Network) is the EMA's federated network for accessing real-world healthcare data from across EU member states to support regulatory decisions. Established in 2022, DARWIN EU connects electronic health databases from hospitals, primary care, and insurance systems across multiple countries — currently covering over 100 million patients — without centralising data. Instead, analyses are run locally at each data source and results are federated. DARWIN EU enables the EMA to: generate real-world evidence (RWE) to support conditional approvals, evaluate post-authorisation safety signals across large populations, conduct pharmacoepidemiological studies for PASS (Post-Authorisation Safety Studies), and provide background incidence rates for adverse events of special interest (AESIs). The network addresses the ICSR-only limitation of spontaneous reporting and the VAERS/Yellow Card system for detecting rare adverse events in large populations.

Question 3

How are safety signals detected in spontaneous reporting systems?

Accepted Answer

Safety signal detection from spontaneous adverse event reports uses disproportionality analysis — statistical methods that identify drug-adverse event combinations reported more frequently than expected by chance, given the reporting patterns across all drugs and all events in the database. The two principal methods are: Reporting Odds Ratio (ROR) — case-control based; and Proportional Reporting Ratio (PRR) — used by the UK MHRA Yellow Card database. For large databases (FAERS in the US, EudraVigilance in the EU), Bayesian methods are preferred: the Multi-Item Gamma Poisson Shrinker (MGPS, FDA FAERS) and Bayesian Confidence Propagation Neural Network (BCPNN, WHO VigiBase) both apply shrinkage to reduce false-positive signals from rare, low-count combinations. A signal is typically defined as IC025 (BCPNN lower 95% credible interval) or EBGM05 (empirical Bayes geometric mean, 5th percentile) exceeding the alert threshold.

Question 4

What is a Risk Management Plan (RMP) and when is one required?

Accepted Answer

A Risk Management Plan (RMP) is a pharmacovigilance document submitted to EMA alongside a marketing authorisation application that characterises a medicine's safety profile, specifies what is known and not known about safety, and describes what risk minimisation measures will be implemented to ensure the benefit-risk balance remains favourable post-market. RMPs are required for all new medicinal products in the EU and for some established medicines when significant new safety concerns arise. RMPs contain three components: a safety specification (characterising the risks — identified risks, potential risks, missing information); a pharmacovigilance plan (how risks will be monitored post-market); and risk minimisation measures (routine measures — product information/label; and additional measures — educational materials, controlled access programmes, REMS equivalent). US REMS (Risk Evaluation and Mitigation Strategies) serve a similar function — the clozapine REMS being a prominent example.

Question 5

How is AI being applied to pharmacovigilance signal detection?

Accepted Answer

Artificial intelligence is increasingly applied across the pharmacovigilance workflow. Natural language processing (NLP) automates the extraction of adverse event information from unstructured sources — literature, social media, patient forums, and unstructured clinical notes — complementing structured ICSR data. Machine learning classifiers detect safety signals earlier in spontaneous reporting databases by identifying complex multi-drug, multi-event patterns that disproportionality analysis misses. Large language models (LLMs) are being evaluated for case narrative assessment, causality evaluation, and PSUR generation. AI-powered pharmacovigilance platforms (Aetion, Veeva Vault Safety) are reducing the time from raw ICSR data to validated safety signal. The principal challenges are explainability (regulatory agencies require transparency in signal detection algorithms), validation on retrospective gold-standard signals, and reducing false-positive burden that can trigger unnecessary label changes.

Pharmacovigilance World Summit 2027

About the Summit

2027 Programme Highlights

DARWIN EU: 2 Years On

DILI Safety Biomarkers

AI-Assisted Signal Detection

Biologics Immunogenicity

RMP/REMS Reform

Pharmacoepidemiology Methods

Frequently asked questions