Table of contents
At a glance
Which information identifies the data version used for a signal assessment?
Record the data source, assessment cut-off, retrieval time, selection rules and case version. Document the handling of duplicates, follow-up information and exclusions. Keep the analysis linked to its scientific assessment. An unusual numerical value is not a confirmed risk. Different data versions can change a result without a new safety issue having arisen.
- Date cut-off, retrieval date, and case version address different questions.
- Counting rules must distinguish between report versions, cases, and selected cases.
- EudraVigilance is to be classified in the current EU procedure in connection with other data sources.
Signal evaluation and statistical indication differ
This article addresses signal management for authorized human medicinal products in the EU. It does not establish a general statistical method for signal detection. GVP Module IX describes a process in which information is reviewed and assessed. Therefore, a potential association is not yet considered a new drug safety risk.
An unusual result may trigger further investigation. On its own it offers little evidence of causality. Consider data quality, clinical context, chronology and additional sources. The example records make the assessment understandable and do not replace medical signal evaluation.
Three time references to be cleanly maintained
The technical reference date describes which information should be included in the planned assessment. The call-up time indicates when the data were actually taken from the system. The case situation shows which version of a case was available. This information may fall apart, for example if a supplement is received after the deadline or an earlier information is corrected later.
Therefore, define how subsequently received or retrospectively corrected information shall be handled. A requery within the same calendar period may now yield a different inventory balance. For a transparent evaluation, you must be able to demonstrate the applicable snapshot point and provide clear explanations of significant deviations. A universal obligation to adopt a specific snapshot technique does not follow.
Sources: [1]
Selection rules as professional work aid
The following internal matrix captures the decisions established, which would otherwise easily be lost behind a product name in a report. The rules must align with the specific evaluation objective. They are not a universal search algorithm nor a predefined list of medical terms to include in every analysis.
| Decision | What to document | Typical misjudgment |
|---|---|---|
| Source and Time | Source, Date, Actual Retrieval | Retrieval date automatically defines the professional time period |
| Product and Event | Terminology, coding, and inclusion rules used | Identically named exports contained the same selection |
| Case version | Addenda, previous versions and corrections | Each report version is an additional case |
| Dubletten | Testing, Results and Consolidated References | Similar reports have always been the same case. |
| Assessment | Clinical context, additional sources and remaining uncertainty | Anomalous number indicates a risk |
EudraVigilance following the end of the signal pilot
The monitoring of EudraVigilance data is a legal obligation: Article 18(2) of Regulation (EU) No 520/2012, as amended by Regulation (EU) 2025/1466, requires that these data be monitored and jointly utilised with other sources. The EMA Q&A from January 2026 outlines the implementation following the end of the signal pilot. Describe within your own signal management process how the data are incorporated and what frequency aligns with the risk level and the respective action step.
The EMA acknowledges flexibility in the phase at which EudraVigilance is applied. During signal validation and assessment, its consideration is expected; the frequency and approach should align with the risk profile and safety profile. The Q&A also states that the previous practice of independently reporting validated signals via the relevant form is no longer expected. Other applicable regulatory guidelines remain in effect.
Therefore, older pilot instructions from GVP Module IX should not be used in isolation as current procedural requirements. The EMA side continues to reference Module IX as Revision 1 and indicates planned updates. An announced revision does not constitute evidence of a published new module.
Worked example in which twelve report versions yield seven selected cases
Example: For a limited internal evaluation, an export contains twelve report versions relating to nine distinct cases. Three data sets represent previous versions of cases that were already included. According to the rule established for this case, the most current available status is used for each case. Two of the nine cases do not meet the documented exclusion criterion, as they pertain to medicinal products other than those specified.
The calculation is 12 versions − 3 previous levels = 9 cases; 9 cases − 2 out of selection = 7 selected cases. Thus, neither a frequency of adverse reactions nor an incidence is calculated. A denominator for the number of actually exposed persons is missing; moreover, spontaneous registrations are not a complete record of all events.
In the status report, the seven cases associated with the decision-making process are linked. One case includes a later received significant addendum. This addendum is not silently incorporated into the already completed evaluation; instead, it is separately assessed for its relevance in determining whether an amendment or re-evaluation is required. The example does not establish a general exclusion rule for subsequent information.
| Step | outset | Withholding | Result |
|---|---|---|---|
| Assign case versions to reporting versions | 12 Versions | 3 previous versions | 9 independent cases |
| Apply documented product criterion | 9 cases | 2 other medicines | 7 selected cases |
| Interpret results | 7 selected cases | No exposure denominator available | No incidence rate calculated |
| Evaluate subsequent addendum | 1 Follow-up after the cut-off date | No silent inventory change | Significance for additional or re-evaluation documentation to be documented |
Associate findings with the evaluation decision.
Document the information that informs the decision and identify remaining uncertainties. This may include unclear exposure levels, inconsistent date references, missing clinical details, or suspected duplicate reports. A corrected dataset may alter the analysis, without automatically implying the emergence of a new signal.
The decision should indicate whether further information is required, whether a re-analysis is warranted, or whether the previous finding supports the intended conclusion. Roles and documentation shall be aligned with the organization's own signal management process. This work guide does not establish a fixed, universal release role nor impose a mandatory obligation to report every numerical discrepancy to regulatory authorities.
Sources: [1]
Checklist prior to completion of an evaluation
Evaluate traceability from the perspective of a colleague who did not originate the query. They should be able to identify the data used and understand the rationale behind a specific conclusion. Not every technical log file is equally relevant; the selection and scope must align with the actual assessment.
- Are source, date, and retrieval date distinguishable?
- Are selection criteria and key rules documented?
- Is it clear whether versions or cases are being counted?
- Are duplicates and conflicting information evaluated?
- Is it clear how later additions are handled?
- Was EudraVigilance considered at the appropriate step in the process?
- Are clinical conclusions and numerical results separated?
- Are open uncertainties and follow-up actions visible?
Monitor changes to data source and analysis methodology
A new coding scheme, a modified filter, or an alternative duplicate handling procedure may affect comparability with previous assessments. Ensure such changes are clearly documented and evaluate their impact on the results used. Supplier or analysis tools shall not be considered appropriate solely because they successfully generate a report.
For small teams, a fixed, short description of the route of evaluation is worthwhile: data source, selection, processing steps and result storage. Additional effort is created especially with several data sources, changing rules and hard-to-understand old data. The professional responsibility remains to be clarified even with outsourced analysis; a technical service does not replace it.
Project experience and collaboration
Identify the data that supported the signal assessment
With signal management and data owners, we trace a selected analysis back to its sources, case versions and selection rules. Which duplicates were accounted for, which follow-ups were still missing and what was the cut-off date? A documented extract may suffice for a defined dataset. Multiple deliveries need traceable reconciliation. We support the technical evidence and its links to operational decisions. Your designated experts retain responsibility for medical signal assessment.
Frequently asked questions about Data versions in signal management
Is the export date equal to the data cut-off date?
No. The export date refers to the technical retrieval, while the professional date denotes the planned temporal boundary. Subsequent amendments or corrections may alter a re-generated export. Therefore, document both dates and the handling of subsequent information. The appropriateness of the technical safeguard for the used version depends on your procedure.
Sources: [1]
May previous versions of the same case be counted multiple times?
Not unintentionally as independent cases. First determine whether your evaluation covers report versions or cases, and apply the appropriate counting rule. For case-based evaluation, prior statuses and amendments must be correctly assigned. The prior status may remain important for traceability, even though it is not counted additionally.
Does a statistical anomaly indicate a new risk?
No. A statistical indication is to be evaluated in the clinical and further technical context. Data quality, possible distortions, duplicates and additional information can change the interpretation. Signal validation and evaluation are therefore more than the assumption of a conspicuous numerical value from a report.
Sources: [1]
Can an incidence be calculated from seven reported cases?
Not solely from case numbers. In this article's example, an appropriate exposure metric is already missing. Furthermore, spontaneous reports do not automatically capture all actual events that have occurred. The seven cases describe a limited, selected dataset and must not be presented as incidence rates in the treated population without additional supporting evidence.
Sources: [1]
Must all license holders continue the previous EV pilot procedure?
No. The previous pilot procedure has been terminated. License holders must monitor EudraVigilance data and jointly utilize it with other sources (Article 18 paragraph 2 of Regulation 520/2012 as amended by Regulation 2025/1466). The EMA Q&A clarifies integration into the signal management process, particularly during validation and evaluation of signals. Therefore, previous pilot instructions should not be carried forward unchanged as the current process.
When should EudraVigilance be incorporated into the evaluation?
The current Q&A allows integration that is appropriate to the process and expects to be taken into account in signal validation and evaluation. Frequency and procedure should fit the product and its safety profile. An identical rigid query frequency for all medicines is not derived from this. Concrete regulatory requirements remain to be observed in addition.
Sources: [2]
What happens to an important follow-up after the cut-off date?
Evaluate, in accordance with your established procedure, whether an amendment, re-analysis, or other review is required. The assessment should neither be silently incorporated into a final data state nor be disregarded on a mere date basis without substantive justification. Document the linkage to the original evaluation and the rationale for the decision.
Sources: [1]
Must every analysis be technically exact and fully reproducible?
The underlying data, methods, and decisions must be sufficiently traceable. When source data is time-varying, a subsequent query may yield a different dataset. Appropriate data snapshots, query descriptions, and result documentation support understanding of the original evaluation. The sources do not specify a single universal technical snapshot solution.
Can an external analyst determine the selection criteria?
Tasks can be agreed upon, but the purpose, responsibilities, and professional assessment must remain clear. The marketing authorization holder retains responsibility for the pharmacovigilance system. Vote on which rules the analyst applies, what changes must be reported, and how the resulting data are reviewed for the decision-maker's own evaluation.
Which changes require special attention?
Particularly relevant are changes that may affect the selection, counting, or interpretation: new data sources, different coding, filter modifications, and altered duplicate handling. Assess their actual impact on the evaluation and, if applicable, on comparisons with prior results. Not every technical change automatically results in a complete re-execution of all analyses.
Sources and review date
Source date: 7 October 2026. GVP Module IX Revision 1 will be aligned with the EMA Q&A from January 2026 regarding the EU procedure. The anticipated module update referenced in the EMA Q&A is not to be assumed as already published. Example and selection matrix are standalone reference tools.
- EMA: GVP Module IX – Signal ManagementRev. 1, 12.10.2017; applicable from 22.11.2017; IX.B.3–B.5, p. 8–11; Validation, Assessment and Quality System
- EMA: Questions and Answers on Regulation 2025/1466 and the End of the EV-Signal PilotEMA/243145/2025 Rev. 2, 28.01.2026; Questions 1–5, Pages 2–3; Pilot end, EV usage and amended reporting pathways
- EMA: Good Pharmacovigilance Practices – Current Modules and Planned UpdatesWeb page accessed 07.10.2026; Planned updates and Final GVP modules: Module IX Revision 1
- EMA: GVP Module VI Addendum I – Duplicate ManagementEMA/405655/2016, 28.07.2017; applicable from 22.11.2017; VI.Add.I.2–4.1.1, p. 4–9; Assignment and duplicate management
- EMA: GVP Module I – Pharmacovigilance Systems and Quality Management SystemsEMA/541760/2011; applicable from 02.07.2012; I.B.10–11, p. 9–12; I.C.1.5, p. 18–19
- EU: Implementing Regulation (EU) 2025/1466 amending Regulation (EU) No. 520/2012Article 1, No. 7 amends Article 18(2) of Regulation 520/2012: Monitor EudraVigilance data and jointly utilise them with other sources. Official German text reviewed on 18 October 2026; the EMA Q&A provides guidance on practical implementation.
Sources checked: .
Signal evaluations with traceable data pathways support
We support you in describing data states, selection rules, and appropriate validation evidence. Medical signal assessment remains the responsibility of the respective subject-matter experts.
- Computer system validation →
Bring your GxP systems into validated operation, from laboratory software, MES and eQMS to ERP. We define the scope, plan risk-based tests using GAMP 5 and hand the evidence to QA for release.

