EVDAS and Signal Detection in Pharmacovigilance

The EudraVigilance Data Analysis System (EVDAS) is the analytical layer used to interrogate safety information held in EudraVigilance. It allows authorised users to move from a large regulatory ICSR repository to structured outputs that can support signal detection and evaluation.

EVDAS should not be understood as a machine that produces regulatory signals. Its outputs include statistical and case-level information that help a reviewer decide where scientific attention is warranted. The medical and regulatory significance of an observation still depends on case quality, novelty, alternative explanations, background risk, exposure, class information and evidence from sources outside EudraVigilance.

That distinction has become more important since Commission Implementing Regulation (EU) 2025/1466 changed the MAH monitoring framework. All MAHs with medicinal products authorised in the EEA must now monitor EudraVigilance data and use those data together with other sources. The previous limited pilot and the standalone MAH validated-signal notification form are no longer the governing model.

This article focuses on how EVDAS works as an analytical tool. The broader legal and governance framework is discussed in [[eudravigilance-signal-detection]], while the statistical foundations are developed further in [[disproportionality-analysis]].

What EVDAS Provides

EMA describes EVDAS as supporting pharmacovigilance safety monitoring with a main focus on signal detection and evaluation of ICSRs. Its principal outputs include:

• electronic reaction monitoring reports (eRMRs);
• line listings of individual cases of suspected adverse reactions; and
• individual case report forms.

EVDAS also calculates a measure of disproportionality based on the reporting odds ratio (ROR). Together, these functions allow the user to move between three levels of evidence:

aggregate reporting pattern → case list → individual clinical report.

That progression is fundamental. The aggregate layer identifies a pattern; the line listing shows whether the cases are homogeneous enough to support a meaningful clinical question; the individual case form provides the detail needed to evaluate chronology, diagnosis, alternative causes and other medically important features.

EudraVigilance versus EVDAS

The two names are often used interchangeably, but the distinction is useful:

EVDAS does not create additional safety evidence. It reorganises and analyses the evidence already available in EudraVigilance.

The Electronic Reaction Monitoring Report

The eRMR is a structured EVDAS output used to review product-event combinations. It provides a reproducible way to scan a large number of reported reactions and identify combinations that may deserve case-level review.

A reviewer should not read an eRMR as a binary list of “signals” and “non-signals”. Instead, each row is a prompt to ask whether a reporting pattern is sufficiently unusual, clinically important or changing to justify further analysis.

Depending on the report configuration and access level, relevant information may include counts, seriousness, temporal information and disproportionality output. The exact displayed fields and interface can change over time, so controlled procedures should describe the analytical intent rather than depend unnecessarily on one screen layout.

What should be retained?

For a material decision, the organisation should preserve enough information to reconstruct what was reviewed. This commonly includes:

• date of extraction;
• substance or product selection;
• MedDRA term or grouping;
• relevant filters;
• eRMR or equivalent output;
• line listing or case forms where reviewed; and
• the resulting decision and rationale.

This is recommended operational practice, not an EMA-mandated file-naming convention.

Reporting Odds Ratio

The reporting odds ratio compares the odds that a particular event is reported with a product of interest with the odds that the event is reported with other products in the comparison dataset.

Using a conventional 2 × 2 table:

ROR = (a/b) ÷ (c/d) = ad/bc

An ROR above 1 means the event is reported proportionately more often with the product than with the comparator set. Confidence intervals help express statistical uncertainty, particularly when case counts are small.

This does not mean that the product causes the event more often in treated patients. The denominator is composed of reports, not exposed patients. ROR therefore measures a reporting pattern rather than incidence or relative clinical risk.

What Creates an SDR?

A signal of disproportionate reporting (SDR) is produced when a predefined detection algorithm identifies a product-event combination that meets its statistical and other criteria. The algorithm may use case counts, ROR confidence limits, exclusions or additional rules.

The important conceptual distinction is:

ROR = statistic
SDR = output of a detection algorithm
validated signal = clinical/scientific judgement after review

These three concepts should not be collapsed into one another. A company may use internal filters or prioritisation rules, but no single internal threshold should be presented as the universal regulatory definition of a signal.

Reading an eRMR Scientifically

A useful EVDAS review moves in layers rather than jumping from a high ROR to a regulatory conclusion.

Start with the event concept

Confirm that the coded term represents the clinical phenomenon of interest. Related MedDRA Preferred Terms can split one syndrome across several labels, while one broad term can combine clinically different diagnoses. When necessary, inspect related terms or groupings rather than assuming a single PT captures the safety question.

Examine the number and quality of cases

A high disproportionality statistic based on a very small number of reports deserves different interpretation from a stable pattern supported by many clinically coherent cases. Review whether the cases contain enough information to establish diagnosis, timing and relevant alternatives.

Case quality can matter more than case count. A small group of well-documented events with a characteristic latency and plausible dechallenge may be more informative than a much larger set of poorly documented reports.

Assess seriousness and clinical consequence

Seriousness is relevant to prioritisation but does not by itself establish a signal. Consider severity, reversibility, preventability, affected population and the clinical consequences if the association proves causal.

Check whether the observation is already known

Compare the event with current product information, the safety specification where applicable, previous signal evaluations and class information. A labelled adverse reaction can still warrant further review if the new evidence suggests a change in frequency, severity, phenotype, outcome or risk factors.

Look for reporting artefacts

EudraVigilance data are influenced by the conditions under which reports are generated. Important possibilities include:

• publicity or regulatory communication;
• stimulated reporting after a label change;
• changes in product utilisation;
• confounding by indication;
• co-prescribed medicines;
• differential reporting between countries;
• duplicate or follow-up reports;
• changes in MedDRA coding; and
• masking by highly reported products or events.

A sudden eRMR change should therefore be interpreted in its historical and regulatory context.

Moving From eRMR to Case Review

Once a product-event pair warrants further evaluation, line listings and individual case forms help determine whether the statistical pattern represents a coherent clinical observation.

A line listing can reveal whether the apparent cluster is concentrated in a particular age group, sex, indication, dose, country, reporting source or time period. It can also expose obvious alternative explanations, such as a common co-medication.

Individual case forms then allow closer review of:

• temporal relationship;
• diagnostic certainty;
• treatment duration and dose;
• dechallenge and rechallenge;
• concomitant medicines;
• underlying disease;
• laboratory, imaging or pathology evidence;
• outcome; and
• reporter assessment where available.

The reviewer should avoid converting missing information into negative evidence. Absence of a documented rechallenge, for example, does not mean that rechallenge was negative.

Integrating EVDAS With Other Sources

Commission Implementing Regulation (EU) No 520/2012 now explicitly requires MAHs to use EudraVigilance data together with other available sources. EVDAS therefore sits inside a larger evidence network.

A product-event pair first identified in EVDAS may be checked against:

• the MAH's global safety database;
• clinical-trial safety datasets;
• scientific literature;
• observational studies or registries;
• real-world healthcare data;
• product exposure and utilisation;
• non-clinical evidence;
• regulatory actions in other jurisdictions; and
• class effects.

Conversely, a signal first detected in literature, trials or internal cases should ordinarily be evaluated against relevant EudraVigilance data when that information can materially inform validation or assessment.

The correct source mix depends on the question. A suspected acute anaphylactic reaction may rely heavily on detailed cases and mechanistic plausibility. A concern about a common cardiovascular outcome may require denominator-based epidemiology before the magnitude of risk can be characterised.

Monitoring Frequency After the 2025 Regulatory Change

The former pilot created a strong association between EVDAS and scheduled primary screening for a limited substance list. The current framework is more flexible but broader in scope.

EMA's 2026 Q&A states that MAHs should monitor EudraVigilance with a frequency proportionate to the product's risk, known safety profile and characteristics. The MAH decides where in its established signal process EV data are used. EudraVigilance may be used as a primary detection source on a defined schedule, but it is also expected to support validation and evaluation.

This means that a universal rule such as “all products monthly” or “mature products quarterly” should not be presented as a regulatory requirement. An organisation may adopt such frequencies internally, but it should be able to explain why they are suitable for the portfolio and how exceptions are managed.

A risk-based operating model

A practical model may consider:

• time since authorisation or launch;
• amount and quality of accumulated exposure;
• important identified and potential risks;
• uncertainty in the safety profile;
• additional-monitoring status;
• product complexity and population vulnerability;
• major recent regulatory or scientific developments; and
• whether other data streams already provide intensive surveillance.

The resulting schedule is an organisational decision. The regulatory requirement is that EudraVigilance data are actually monitored and used coherently with other safety information.

Validation, Assessment and Documentation

When EVDAS contributes to a potential signal, the signal record should make clear what role the system played. It may have initiated the issue, strengthened a hypothesis detected elsewhere, provided cases for validation or contributed quantitative context during assessment.

A concise but reconstructable record should identify:

• the safety question;
• the relevant EVDAS output and extraction date;
• important case-level observations;
• the statistical interpretation;
• other evidence sources reviewed;
• the validation or assessment conclusion; and
• resulting actions.

The record does not need to reproduce every screen or every case field. It needs to preserve the evidence necessary to understand the decision.

Governance and Quality Controls

EVDAS governance should support both scientific flexibility and reproducibility. The objective is not to force every product through an identical analytical recipe, but to make clear who performs the work, what methods are used, how important decisions are reviewed and how changes to the process are controlled.

Useful controls include defined access roles, documented monitoring strategies, training appropriate to analytical and medical responsibilities, change control for internal procedures and signal tools, and a signal record that links EVDAS evidence to the wider assessment.

Where EVDAS data are downloaded into internal analytical systems, the organisation should also understand data lineage, transformation rules, versioning and validation of any code or software that materially affects the result.

Role of the QPPV

The QPPV's responsibility is oversight of the pharmacovigilance system rather than routine execution of EVDAS queries. Oversight should provide visibility of material open signals, major process deviations, overdue assessments, important methodological changes and significant regulatory actions.

The QPPV should be able to understand how EudraVigilance is used within the signal-management process and whether the process has been adapted to the post-2025 legal framework.

Potential Failure Modes

The following are illustrative failure modes rather than reported inspection findings.

eRMR review becomes a tick-box activity

A file is downloaded and marked “reviewed” without evidence that the underlying cases or scientific context were considered. The control demonstrates document handling, not signal detection.

ROR is interpreted as relative risk

A reporting odds ratio reflects relative reporting in a spontaneous-report database. Treating it as incidence or clinical relative risk can produce materially misleading conclusions.

Every statistical flag is escalated

A process with no effective clinical triage may generate large numbers of low-value assessments. The purpose of an SDR algorithm is to focus attention, not remove scientific judgement.

Thresholds are changed without impact assessment

Changing minimum counts, filters or prioritisation logic can alter which observations are detected. Material methodological changes should be controlled and their effect on ongoing surveillance considered.

The former pilot list remains embedded in procedures

The pilot ended in 2025. Limiting EV monitoring to those historical substances no longer reflects Article 18(2).

EV evidence is absent from signals detected elsewhere

If a significant signal originates from literature or internal cases but the organisation never considers relevant EudraVigilance data, the current expectation to use EV together with other sources may not be met.

Standalone signal notification is still treated as mandatory

The former Article 21(2) MAH notification requirement was deleted. Procedures should instead direct resulting actions through the appropriate current regulatory mechanism.

Inspection Considerations

Illustrative inspection questions include:

• How is EVDAS positioned within the company's signal-management process?
• How was the process revised after the termination of the MAH pilot?
• How is monitoring frequency selected and reviewed?
• Which eRMR fields or algorithms are used for screening and why?
• How are SDRs distinguished from validated signals?
• How are line listings and case forms incorporated into medical review?
• Can a historical signal decision be reproduced from the retained EVDAS evidence?
• How are changes to internal analytical methods controlled?
• How are EVDAS data used when a signal originates from literature, clinical trials or the company database?
• What does the QPPV see about EVDAS-related signal performance and significant issues?

An inspector may also compare written procedures with actual signal files to determine whether the operating model is implemented consistently.

Practical Checklist

Before relying on an EVDAS-based signal process, the organisation should be able to confirm that:

• EVDAS access and responsibilities are current;
• all EEA-authorised products are considered under the current monitoring framework;
• monitoring frequency and the role of EVDAS are risk-based and documented;
• eRMRs are interpreted as screening outputs rather than definitive signals;
• ROR is not described as incidence or causal risk;
• clinically important observations can be escalated even without statistical disproportionality;
• material eRMR findings can be traced to line listings and case-level review;
• duplicate, coding, stimulated-reporting and confounding issues are considered where relevant;
• EVDAS information is integrated with other data sources;
• validation and assessment conclusions remain reconstructable;
• changes to methods, filters or internal analytical software are controlled; and
• regulatory actions use the current legal routes rather than the obsolete standalone signal-notification form.

Key Takeaways

EVDAS is an analytical interface, not an autonomous signal-management system. Its principal value is the ability to connect an aggregate reporting pattern with the case-level evidence needed for scientific interpretation.

The reporting odds ratio identifies disproportionate reporting; it does not estimate incidence or prove causality. An SDR is an algorithmic flag; it is not automatically a validated signal.

Since the 2025 amendment to Implementing Regulation (EU) No 520/2012, all MAHs with medicines authorised in the EEA must monitor EudraVigilance data and use them with other sources. The monitoring model may be proportionate to product risk and characteristics, and EVDAS can be used at detection, validation and assessment stages.

The quality of an EVDAS process is therefore demonstrated less by the number of reports generated than by the scientific trail from screen → cases → integrated evidence → decision → action.

References

1. European Medicines Agency. EudraVigilance system overview — EVDAS, eRMRs, line listings, individual case forms and reporting odds ratio. https://www.ema.europa.eu/en/human-regulatory-overview/research-development/pharmacovigilance-research-development/eudravigilance/eudravigilance-system-overview
2. European Commission. Commission Implementing Regulation (EU) 2025/1466 of 22 July 2025 amending Implementing Regulation (EU) No 520/2012. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32025R1466
3. European Commission. Commission Implementing Regulation (EU) No 520/2012, consolidated version current at 12 February 2026. https://eur-lex.europa.eu/eli/reg_impl/2012/520/2026-02-12/eng
4. European Medicines Agency. Questions and answers on Implementing Regulation (EU) 2025/1466, EMA/243145/2025 Rev. 2, updated 28 January 2026. https://www.ema.europa.eu/system/files/documents/other/questions-answers-implementing-regulation-eu-2025-1466-en.pdf
5. European Medicines Agency. Signal management. https://www.ema.europa.eu/en/human-regulatory-overview/post-authorisation/pharmacovigilance-post-authorisation/signal-management
6. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module IX – Signal management (Rev. 1) and Addendum I. Available from the EMA GVP collection. https://www.ema.europa.eu/en/human-regulatory-overview/post-authorisation/pharmacovigilance-post-authorisation/good-pharmacovigilance-practices-gvp
7. European Medicines Agency. 2025 Annual Report on EudraVigilance for the European Parliament, the Council and the Commission. https://www.ema.europa.eu/en/documents/report/2025-annual-report-eudravigilance-european-parliament-council-commission_en.pdf

Regulatory Note

The requirement for MAHs to monitor EudraVigilance and use its data with other available sources arises from Article 18(2) of Implementing Regulation (EU) No 520/2012 as amended by Implementing Regulation (EU) 2025/1466. EMA's 2026 Q&A provides current practical interpretation while GVP Module IX is being aligned with the amended legislation. Internal eRMR schedules, thresholds, case-review templates, retention models and inspection questions described in this article are recommended or illustrative practices unless specifically required by an authoritative source.