Regulations and compliance · Practical guide

GVP duplicate cases: investigating similar reports without losing cases

Two reports name the same medicine, the same event and a similar patient age. They may be duplicates, but merging them too early can conceal a separate case. In European pharmacovigilance for human medicines, document the initial suspicion, confirmation and subsequent handling as separate decisions.

Table of contents

At a glance

When can we merge two safety reports as duplicates?

Merge reports only when the assessment sufficiently demonstrates that they concern the same individual case. A search hit or similarity score raises a suspicion. Compare identifiers, event history, sources and contradictions, and resolve remaining questions. For confirmed duplicates, link relevant information and references in the master case. Unresolved similarity does not justify discarding a report or missing an applicable reporting deadline.

  • Similarity is a search criterion, not yet confirmation of duplicates.
  • A master batch must integrate information and traceability, not merely reduce data sets.
  • The decision to oppose a consolidation also requires a justifiable basis.

1. What a Fall Dublette Really Means

In the context of EU Human Pharmacovigilance, a duplicate refers to multiple reports relating to the same individual case. However, the same patient may experience different events at different time points. Similarly, two patients may receive the same medicinal product and exhibit a similar reaction. Such scenarios must be carefully distinguished and separated initially.

GVP Module VI Addendum I distinguishes searching for, confirming and managing duplicates. ICH E2D(R1) supplements the current case-handling framework. The additional general EU transition period ended on 18 September 2026, after which the definitions and guidance in E2D(R1) apply. Where E2D(R1) refers to regional requirements, GVP Module VI remains relevant in the EU. Address specific implementation arrangements separately, including ongoing support programmes and technical E2B values. Our example tables prescribe neither a new technical procedure nor a universal similarity threshold.

Sources: [1] [3] [4]

2. The search must allow for differences

A previously processed report may use different terms or more detailed patient data than a new report. A search that requires exact matches can miss potential duplicates. GVP Module VI Addendum I identifies information such as patient, source, medicinal product, event and known case identifiers.

Check how missing values, inconsistent coding, and later diagnosis refinements are handled. An unknown age entry does not constitute positive evidence of different patients. Similarly, a different MedDRA term does not necessarily indicate a different adverse event. The specific search combinations supported by the system and the reliability of those combinations must be validated for the actual dataset.

Sources: [1] [3]

3. A justified decision instead of a similarity score

The following internal matrix focuses on decision-relevant differences. It does not assign equal weight to all fields. A confirmed discrepancy in patient identity may be more significant than multiple superficial agreements. Conversely, inconsistent information should not be rejected solely because it originates from a less detailed report.

Record the evidence and open questions alongside each result. When information is limited, a targeted follow-up query may be more appropriate than immediate consolidation. Responsibility for the decision-making process should be clearly defined in the procedure; the table does not imply any specific job title.

Own inspection aid prior to duplicate confirmation
CharacteristicAssess validityPossible consequence
Known foreign identity Does it provide verifiable reference to the same case? Strong assignment document; source of identification received
Patient information Does the same person hold the position or only similar characteristics? When similarity is present alone, further evaluation is required.
Temporal progression Was is the same incident or a new episode? Evaluate chronology and repetitions separately
Deviation in diagnosis Clarification of the same event or contradiction? Medical evaluation and, if applicable, demand
Joint study or batch Shared context, but different patients? Integration may be feasible; consolidation is not automatically performed

Sources: [1]

4. Example: Two related reports and one standalone case

In the example, three synthetic reports A, B, and C are present. Reports A and B initially describe a similar incident occurring in the same practice. Only a documented feedback from the source confirms that A and B pertain to the same patient and the same incident. For report C, the source explicitly confirms that it involves a different patient.

The target state thus consists of two individually assessable cases. Three inputs remain traceable; inputs A and B are grouped under a master case, while input C remains standalone. The calculation of three inputs minus one confirmed duplicate counts results in two cases. This does not represent a measured duplicate rate and provides no information on the frequency of an adverse reaction.

Example: filled assignment decision
ReportDocumented resolutionExpected handling
A Patient P-01; Event on 2 October; Source: Clinic X Basis for Master M-01
B Source confirms P-01 and the same incident; includes additional timeline. Link to M-01; adopt new relevant information
C Source confirms P-02; similar event in the same facility Independent case M-02; no consolidation
After processing Three entry points, two individual cases All origin references to be included; analysis to use appropriate valid versions

Sources: [1] [3]

5. Check the content and history of the master case

A master record may originate from an existing record or be constructed anew from the verified duplicates. Critical factors are the complete assignment of references and the most current possible representation. Missing information in a report is not automatically interpreted as an explicit denial.

Review structured data, narrative, assessments, and source references following consolidation. Contradictory information requires clarification or clear indication of uncertainty. Inactivated sub-cases remain retained for traceability; simple deletion impedes the ability to demonstrate why a reduced number of cases is present in subsequent analysis.

Sources: [1] [5]

6. Database release decision and external reporting alignment

An internal consolidation does not automatically update the status in EudraVigilance or with a partner. Therefore, verify which reports have already been sent by which organization and identify any additional actions required in accordance with applicable regulations. Case identifier, sender-specific identifier, and other known identifiers serve distinct functions.

If the identity of a batch has not been sufficiently established prior to the expiry of a relevant reporting deadline, the suspicion of a duplicate shall not be used as a ground for unjustifiably withholding an otherwise valid report. The investigation continues. A nullification is furthermore not an arbitrary cleanup action; it must pertain to the specific case report and must align with the actual sender and recipient configuration.

Sources: [1] [2] [5]

7. Incorrect consolidation of controlled corrections

Later it will become evident that the consolidated reports relate to different cases; in such instances, a clear and traceable correction is required. Retain the previous decision, the new evidence, and the assignment of information. Review the affected analyses and already dispatched batches, rather than merely resetting a technical status.

The specific traceback is a system and process issue. This guide does not claim to provide a universal function for splitting a master case. Describe the expected outcome: each case contains the associated data, no finding is lost, and recipients receive the required corrections. Unclear origin of individual entries is treated as an open point.

Sources: [1] [2]

8. Duplicate management in a small team to be safeguarded

The own checklist can be used for a single decision and for the examination of your procedure. A small inventory justifies simple tools, but not an unexplained decision. A central list of open suspicious cases with responsible processor and next action is particularly helpful.

Review changes to the search logic to ensure both identified duplicates and visually similar, distinct cases are detected. Example C must be retained. Add cases with missing values or inconsistent coding. A successful standard case does not validate that the search identifies all relevant differences.

  • Suspicion and confirmed result separately documented?
  • Have key identifiers and prior reports been incorporated?
  • Missing values not treated as proven difference?
  • Are sources and contradictions in the master record traceable?
  • Are standalone cases still retained despite similarity?
  • Reviewed for impact on reports and analyses?
  • Open clarification with reporting deadlines and accountability assigned?

Sources: [1] [2] [3]

Project experience and collaboration

Assess similar reports before deciding whether to merge cases

With case processors and the system team, we compare controlled reports containing similar information at different levels of completeness. Which features support the same case, and which conflict? A clear match may permit straightforward linking. Unclear provenance or several reporting organisations require further expert assessment. We document the decision, identifiers and links between original reports. A similarity score alone must neither confirm a duplicate nor cause a distinct case to disappear.

Frequently asked questions about Duplicate pharmacovigilance cases

Does the same patient already have a duplicate record?

Not on that basis alone. Establish whether the reports describe the same case or incident. A patient can experience several distinct events. Check the chronology and content before merging records that may represent separate episodes.

Sources: [1]

Does a high automated similarity score suffice?

He initially identifies a potential match. The Guidance requires professional confirmation of potential duplicates. The threshold employed by your system is not a regulatory truth value. Document the key similarities, differences, and any subsequently requested information.

Sources: [1]

Must case identifiers exactly match?

No. The same case may have different local identifiers assigned by various organizations. Therefore, external identifiers should be matched and linked. Conversely, a wrongly imported identifier must not override any other evidence. Please verify the origin documentation.

Sources: [1] [2]

Does a different diagnosis establish that the cases are distinct?

No. A subsequent report may provide a more precise description of the same course. The medical concepts and temporal relationship must be evaluated. A simple string-based check is often insufficient for this purpose; it may at best support only part of the search.

Sources: [1]

Can subordinate cases be deleted after a merge?

The original reports and their consolidation must be retained. The addendum addresses secondary cases that remain traceable within the database. Therefore, a reduced evaluation dataset is not equivalent to the removal of the original case history.

Sources: [1]

What should be done in case of conflicting information?

Try to clarify the differences and ensure the origin is documented. If the discrepancy persists, the representation of uncertainty must be understandable. Do not invent a seemingly definitive value and do not accept all provided information as simultaneously accurate facts without verification.

Sources: [1]

May we withhold the report due to suspicion of a duplicate?

An unexplained suspicion does not invalidate applicable reporting deadlines. The Guidance outlines the handling of a otherwise valid case where identity clarification remains pending. The investigation continues and the outcome is subsequently implemented through the appropriate procedure.

Sources: [1]

Is every consolidation associated with a nullification?

The required reporting steps depend on the existing shipped inventory and the sender configuration. An internal master training is not automatically equivalent to an external nullification. Review the applicable EudraVigilance guidelines and the actual affected reports.

Sources: [2] [5]

What do we assess following modification of the search algorithm?

Use appropriate known exceptions: confirmed duplicates, independent similar cases, and variants with missing information. Evaluate new misclassifications and missed hits. The addendum recommends validating search algorithms and assessing their need for adjustment over time.

Sources: [1]

Must a decision against a merge be documented?

Yes, a traceable investigation should also document the finding that no duplicate exists. This enables a later identified match to be meaningfully contextualized. The rationale should refer to specific information and not solely rely on the status 'no duplicate'.

Sources: [1]

Sources and review date

Source date: 7 October 2026. GVP VI Addendum I and Module VI are to be read together with ICH E2D(R1) and the EU Implementation Strategy. The specific system behaviour must be assessed separately. Tables and figures are intended as guidance only.

  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, pp. 4–9
  2. EMA: GVP Module VI – Collection, Processing, and Reporting of Suspected CasesRevision 2, dated 28 July 2017 and applicable from 22 November 2017. Sections VI.B.4–5 and VI.C.6.2.2.6–9, pages 17–18 and 58–63.
  3. EMA/ICH: E2D(R1) – Post-approval safety dataStep 5, 18.09.2025; EU Efficacy 18.03.2026; § 6.6, Page 22
  4. EMA: EU Implementation Strategy for ICH E2D(R1)EMA/11141/2026, 26.02.2026; PRAC 20.02.2026; § 2, S. 2
  5. EMA: EU ICSR Implementation GuideEMA/51938/2013 Rev. 2, March 2021; I.C.6.1.1–.3, p. 98–99

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Traceable decisions and their data implications

We assist you in translating search logic, technical validation, and implications for reports into clear requirements and actionable "should" statements.

  • 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.

Send a project enquiry Customer testimonials →