Signal detection pharmacovigilance and EU safety assessment
Signal detection pharmacovigilance identifies a possible new medicine-event link. In the EU, validation and PRAC assessment do not prove causation.
Signal detection in pharmacovigilance identifies information suggesting a new potentially causal association between a medicine and an event, or a new aspect of a known association, that warrants investigation. It combines clinical review with statistical screening where appropriate. In the EU, EMA and PRAC treat a detected signal as a question. Detection does not prove that the medicine caused the event. Do not use this as UK or US compliance guidance. Article 19 of Implementing Regulation 520/2012 defines the signal. EMA's signal-management page states the causation limit in plain words.
A safety signal can arise from spontaneous reports, clinical studies or scientific literature. EudraVigilance is an important source of suspected adverse-reaction information, not the whole evidence base. EMA, Member State authorities and marketing-authorisation holders (MAHs) share signal detection and management. PRAC prioritises and assesses signals for both centrally and nationally authorised medicines. EMA sets out those sources and responsibilities.
What is a safety signal in pharmacovigilance?
EMA describes a safety signal as information on a new or known adverse event that may be caused by a medicine and requires investigation. Article 19 asks whether the information suggests a new potentially causal association, or a new aspect of a known one, with enough reason to justify verification. Causation does not have to be established first. The regulation states that test, and EMA explains the adverse-event wording.
An adverse event reported after treatment does not, by itself, show that the medicine caused it. Another illness or another medicine can explain the event. Signal assessment investigates that possible causal relationship. A report is not an assessed conclusion, and a statistical flag is not one either. EMA states that distinction.
Pharmacovigilance signal detection methods
Clinical review and statistical screening complement each other. EU law requires a multidisciplinary approach. Signal detection in EudraVigilance is complemented by statistical analysis where appropriate. The regulation also provides for individual-case review, case-series review, grouping or stratification, and statistical methods. An algorithm alone is not signal management. Articles 19, 20 and 23 describe these approaches.
Clinical and case-based review
Clinical review looks at the evidence behind a suspected association. Article 20 lists clinical relevance, quantitative strength of the association, consistency, exposure-response relationship, biological plausibility, experimental findings, analogies, and the nature and quality of the data. Article 20 sets out that list.
Statistical prioritisation does not remove the need for clinical attention. EMA's designated medical-event list is a safety net for conditions that are inherently serious and often medicine-related, irrespective of the statistical criteria used to prioritise safety reviews. It does not replace case review, so an absent statistical flag is no reason to stop. EMA explains the list on its signal-management page.
Statistical screening and disproportionality
Disproportionality methods compare observed product-event reporting with expected reporting. One method is the proportional reporting ratio (PRR). It compares the proportion of reports mentioning an event for one product with the corresponding proportion for other products. FDA's data-mining white paper explains PRR. That paper is a methodological explanation. It is not an EU legal requirement, and it is not a binding FDA procedure for companies.
A disproportionate reporting pattern is a statistical association. FDA says it cannot be interpreted as causal or related to risk. Don't read a reporting ratio as the probability that an exposed patient will have the event. On drug safety reports, FDA's reporting-system dashboard adds a separate limit: reports can be duplicate, incomplete or unverified, they do not establish causation, and they cannot estimate how often the event occurs. The white paper and the public dashboard are US methodological sources, not EU duties.
Signal management from detection to assessment
For signal detection and management in pharmacovigilance, EU law splits the work into several decisions. Validation asks whether the supporting documentation contains enough evidence of a new potentially causal association, or a new aspect of a known one, to justify further analysis. Confirmation is the regulatory decision that a validated signal requires that further analysis. Neither step is a finding that the medicine caused the event. Article 21 defines those procedural steps, and EMA leaves the causality question to assessment.
| Stage | What it establishes |
|---|---|
| Detection | Information suggests a new potentially causal association, or a new aspect of a known one. |
| Validation | The documentation has enough evidence of that new association, or new aspect, to justify further analysis. |
| Confirmation | A validated signal requires further regulatory analysis. Causation is not proven. |
| Analysis and prioritisation | PRAC does the initial analysis and prioritises confirmed signals. |
| Assessment | The potential causal relationship is investigated, including whether more evidence or action is needed. |
| Recommendation for action | PRAC can recommend no further action at present, more evidence, or regulatory action. |
Who confirms the signal depends on the authorisation route. For a nationally authorised medicine, confirmation is by the competent authority of a Member State where the product is marketed, or by an appointed lead Member State or co-leader. For a centrally authorised medicine, EMA confirms it in collaboration with the Member States. Confirmed signals then go to PRAC for initial analysis and prioritisation. Article 21 sets out that split. A lead Member State can also monitor and validate signals for nationally authorised substances under work-sharing. Where no lead is appointed, the relevant Member States share that monitoring.
The word "confirmed" can mislead.
Read it as the name of a procedural step, not as a causal finding. But signal assessment in pharmacovigilance is a later step: it investigates whether a causal relationship exists. Check that assessment, and the recommendation, before you treat the signal as settled. EMA describes assessment as the investigation of the potential causal relationship.
What PRAC outcomes mean for product information
PRAC can conclude that no further evaluation or action is needed at present. It can also ask for continued monitoring, further analysis or more evidence. Evidence requests can include marketing-authorisation holder data in the next periodic safety update report (PSUR), an ad-hoc PSUR, or a post-authorisation safety study. A signal does not automatically require a label change or a withdrawal. EMA's signal-management page lists those outcomes.
No further action at present is not a finding that the medicine can never cause the event. It means further evaluation is not required now. EMA states that a signal does not directly mean the medicine caused the event.
Regulatory action, when it is chosen, can update the summary of product characteristics (SmPC) and package leaflet, or the risk-management plan, through a variation. Other options are a referral or urgent safety restrictions. EMA does not treat any of these as an automatic result of detection.
For a centrally authorised medicine, a product-information recommendation goes to CHMP for endorsement. By the time EMA publishes that recommendation, CHMP has already endorsed it. For a nationally authorised medicine, the recommendation goes to CMDh for information, not for endorsement. Member State authorities oversee adherence. EMA also publishes the recommended wording in official EU languages, plus Norwegian and Icelandic. The PRAC recommendations page explains this route.
Publication is not proof that every affected leaflet has already changed. MAHs are expected to monitor the published recommendations and take the recommended action. For a product-specific check, read the recommendation, the affected substance or product, the requested action and the timetable, then compare that text with the authorised product information and the applicable implementation guidance. Do not reuse a generic variation category in place of the particular recommendation. Start with the published PRAC recommendations.
The old EudraVigilance pilot has ended
Old pilot lists mislead. EMA states that the MAH EudraVigilance pilot ended when Implementing Regulation 2025/1466 entered into force. The updated requirements apply to all MAHs with medicinal products authorised in the EEA. The old pilot substance list is not the current general rule. EMA's signal-management page explains the change.
On 2 November 2026, EMA still listed GVP Module IX Revision 1, reference EMA/827661/2011, legally effective on 22 November 2017, while saying updates would follow to align that guidance with the amended framework. Current law applies during that gap. Use the current consolidated regulation with EMA's current guidance and Q&As. Do not substitute an old pilot timetable for the timetable in the particular recommendation.
RegAffairs AI has a source-linked answer on signal detection in pharmacovigilance and EU follow-up. For the recommended wording and the translations, use EMA's PRAC recommendations page.