EMA发布欧盟药品监管数据质量框架ADR报告章节草案，公众咨询至2026年12月14日
Draft Data Quality Framework (DQF) for EU medicines regulation: application to collection, management and submission of Adverse Drug Reaction reports
EMA发布ADR报告数据质量框架草案（EMA/87692/2025），公开征求意见至2026年12月14日。草案适用于人用和兽用药品ADR报告的收集、管理和提交，按可靠性、充分性、一致性、及时性和相关性五个维度梳理数据质量问题并提出改进建议。文件明确为描述性草案，不替代现有法规或指南，也不增加额外要求。
PDF 文字版；图形和原始排版请参阅官方 PDF。
第 1 页
1 15 October 2026
2 Data Analytics and Methods Task Force
3 European Medicines Agency
4 EMA/87692/2025
5 Data Quality Framework (DQF) for EU medicines
6 regulation: application to collection, management and
7 submission of Adverse Drug Reaction reports
8 Draft
Draft review – Network Data Steering Group 17 July 2026
Network Data Steering Group comments addressed 29 September 2026
Start of public consultation 19 October 2026
End of public consultation (deadline for comments) 14 December 2026
9
Comments should be provided using this EUSurvey form. For any technical issues, please contact
the EUSurvey Support.
10
11
Keywords Adverse Drug Reaction, Adverse Event, EudraVigilance, EVVET
12
13
Data Quality Framework (DQF) for EU medicines regulation: application to collection, management and submission of Adverse Drug
Reaction reports
第 2 页
14 Table of contents
15 1. Executive Summary ................................................................................. 3
16 2. Abbreviations .......................................................................................... 4
17 3. Scope and purpose of the ADR DQF ......................................................... 6
18 3.1. Contextualisation of the EMRN DQF framework to ADR reports ................................... 6
19 4. Background on ADR collection, management and submission ................. 9
20 4.1. ADR guidelines and legislation ............................................................................... 9
21 4.2. Considerations for ADR terminology ...................................................................... 12
22 4.3. Processes and systems underpinning ADR reporting ............................................... 14
23 4.4. Current features of ADR data ............................................................................... 17
24 4.5. Considerations for use of ADRs in regulatory activities ............................................ 18
25 5. Managing challenges in ADR reporting .................................................. 21
26 5.1. Reliability .......................................................................................................... 21
27 5.2. Extensiveness .................................................................................................... 24
28 5.3. Coherence ......................................................................................................... 26
29 5.4. Timeliness ......................................................................................................... 29
30 5.5. Relevance ......................................................................................................... 30
31 5.6. Special considerations for specific ADR data sources ............................................... 31
32 6. Recommendations to improve DQ in ADR reporting............................... 34
33 6.1. Education .......................................................................................................... 34
34 6.2. Integration of clinical pharmacology in ADR processes ............................................ 36
35 6.3. Improvement of standard implementation ............................................................. 36
36 6.4. Early implementation of data validation processes .................................................. 37
37 6.5. Implementation of feedback mechanisms to allow for data corrections. ..................... 38
38 6.6. Maintain immutability of data at the point of transmission ....................................... 38
39 6.7. Routine duplicate detection and elimination by databases ........................................ 39
40 6.8. Effect of recommendations on DQ per DQ dimension .............................................. 40
41 7. Concluding remarks ............................................................................... 42
42 8. References ............................................................................................ 43
43 9. Glossary ................................................................................................ 47
44 Appendix I – Overview of guidelines ......................................................... 48
45
Page 2/49
第 3 页
46 1. Executive Summary
47 This document presents the Adverse Drug Reaction (ADR) chapter of the Data Quality Framework
48 (DQF) for medicines regulation published by the European Medicines Agency (EMA) [1] (hereafter
49 referred to as European Medicines Regulatory Network (EMRN) DQF). It is an extension of the
50 EMRN DQF that specifically addresses collection, management and submission of ADR1 reports in both
51 the human and veterinary domains.
52 The data quality concepts discussed in this chapter apply equally to medicinal products for human and
53 for veterinary use. Nevertheless, the legal landscape in the veterinary field contains certain
54 particularities related to terminology and roles and responsibilities in pharmacovigilance (PV) which
55 differ from the legal provisions applicable to medicines for human use. Such differences have been
56 flagged, where applicable, in this document.
57 This chapter builds upon existing guidelines and procedures, identifies gaps and provides actionable
58 and focused recommendations for assessing and improving the Data Quality (DQ) of ADR reports
59 pipeline in order to best-support regulatory decision making. This chapter is intended for all
60 stakeholders involved in ADR collection, management and submission, as identified during the webinar
61 “Multi-stakeholder workshop on Data Quality Framework for Adverse Drug Reaction reporting” [2], and
62 therefore also provides a holistic overview of the overall ADR reporting ecosystem.
63 This chapter specifically provides:
64 • Background on collection, management and submission of ADR reports (guidelines & legislation,
65 terminology, processes underpinning ADR reporting and an introduction to the ADR reports
66 pipeline), Section 4;
67 • Insights into the current challenges of ADR reporting organised by the five DQ dimensions
68 described in the EMRN DQF (reliability, extensiveness, coherence, timeliness, relevance), Section
69 5;
70 • Recommendations to improve DQ in ADR data for systems and processes underpinning data
71 collection, management and submission, Section 6;
72 • Concluding remarks, Section 7.
73 Concepts in this chapter are founded on principles described in the EMRN DQF and require background
74 knowledge of the EMRN DQF (e.g. determinants and dimensions). The chapter provides specific
75 recommendations to improve DQ for human and veterinary ADR reporting.
76 It is important to note that this chapter is not a replacement for existing legislation or guidelines, and
77 it does not supersede any guidance documents.
78
1 For the purposes of this document and in relation to veterinary medicinal products, the term ‘adverse drug reaction’ (ADR)
should be considered equivalent to the term ‘suspected adverse event’ defined in the Veterinary Medicines Regulation (EU)
2019/6
Page 3/49
第 4 页
79 2. Abbreviations
Acronym Meaning
ADR Adverse Drug Reaction
AE Adverse Event
AER Adverse Event report
ANSM French National Agency for the Safety of Drugs and Health Products
AI Artificial intelligence
CIOMS Council for International Organizations of Medical Sciences
CRO Contract Research Organisation
DQ Data Quality
DQF Data Quality Framework
EEA European Economic Area
EMA European Medicines Agency
ENCePP The European Network of Centres for Pharmacoepidemiology and
Pharmacovigilance
EMRN European Medicines Regulatory Network
EPITT European pharmacovigilance issues tracking tool
EV EudraVigilance
EVCTM EudraVigilance Clinical Trial Module
EVPM EudraVigilance Post-Authorisation Module
EVV EudraVigilance Veterinary
HCP Healthcare Professional
ICH International Council for Harmonisation of Technical Requirements for
Pharmaceuticals for Human Use
ICSR Individual Case Safety Report
MAH Marketing Authorisation Holder
MedDRA Medical Dictionary for Regulatory Activities
NCA National Competent Authority
OTC Over-the-counter
PM Post-Marketing
PRAC Pharmacovigilance Risk Assessment Committee
PSP Patient Support Programme
Page 4/49
第 5 页
Acronym Meaning
PV Pharmacovigilance
QA Quality Assurance
QC Quality Control
RA Regulatory Authority
ROR Reporting Odds Ratio
RSI Reference Safety Information
SAE Serious Adverse Event
SOP Standard Operating Procedure
SUSAR Suspected Unexpected Serious Adverse Reaction
UPD Union Product Database: EMA database for veterinary medicinal products
VeDDRA Veterinary Dictionary for Drug Regulatory Activities
VICH International Cooperation on Harmonisation of Technical Requirements for
Registration of Veterinary Medicinal Products
VICH GL VICH Guideline
xEVMPD Extended EudraVigilance Medicinal Product Dictionary: EMA database for human
medicinal products
80
Page 5/49
第 6 页
81 3. Scope and purpose of the ADR DQF
82 Many guidelines, methods and systems are already in place for the purpose of collection, management
83 and submission of ADRs (see Appendix I). The purpose of this chapter is to build upon existing
84 guidelines and procedures across all domains (human and veterinary, post-marketing and clinical
85 trials), to identify gaps and provide actionable and focused recommendations for assessing and
86 improving DQ of ADRs along the pipeline and for regulatory decision making. In order to do so, it
87 extends and contextualises the principles of the EMRN DQF (see section 3.1. ) to ADR reporting.
88 Moreover, this chapter provides an overview of all stakeholders involved (as identified during the
89 workshop [2]), their interactions and their roles in ADR collection, management and submission and
90 the overall pipeline as identified during interviews with stakeholders from each domain, in view of
91 understanding DQ issues related to the larger ADR reporting ecosystem.
92 The chapter is aimed at addressing DQ aspects from the perspective of the use of ADR data in the
93 regulatory process. It is applicable to all stakeholders that share ADR data with EudraVigilance (EV) or
94 EudraVigilance Veterinary (EVV) (the databases and data processing networks for ADRs managed by
95 EMA): national Competent Authorities (NCAs) in European Economic Area (EEA) Member States,
96 marketing authorisation holders (MAH), sponsors of clinical trials, Ethics Committees and other
97 regulatory bodies. This chapter focuses on both human and veterinary ADR reporting arising from all
98 sources, independent of design or purpose, unsolicited and solicited reports including, but not limited
99 to, interventional clinical trials, non-interventional studies, registries, patient-support programmes,
100 medical literature and patient-reported ADRs.
101 This chapter is descriptive, not prescriptive. This chapter does not supersede or replace existing
102 guidance, and it does not place additional requirements on stakeholders.
103 3.1. Contextualisation of the EMRN DQF framework to ADR reports
104 This section covers the main concepts of the EMRN DQF, contextualising them to ADR reports and
105 introducing specific terminology. In particular, it introduces determinants, DQ dimensions and the data
106 lifecycle.
107 3.1.1. DQ determinants
108 In order to take a systematic approach to dissect DQ issues, the EMRN DQF partitions them in
109 foundational, intrinsic and question-specific determinants (See Figure 1).
110
111 Foundational determinants are defined as “everything that impacts DQ but is not related directly to
112 the dataset and does not depend on any specific research question”. Foundational determinants are
113 referred to as the characterisation of the systems and processes that have an impact on DQ. In
114 most locations where ADRs are captured, managed, or processed, foundational determinants for ADRs
115 are:
116 • Database systems;
117 • Software with an interface of entry form(s);
118 • SOPs and work instructions;
119 • Medical and product dictionaries available for coding.
Page 6/49
第 7 页
120 Figure 1: Determinants of data quality in ADR reporting
121
122 Intrinsic determinants are defined as DQ aspects that can only be observed based on a given
123 dataset without information about how the data was captured, or about its intended use. In this
124 chapter, intrinsic determinants for ADR are for instance:
125 • The data model;
126 • Required and optional values;
127 • The way of reporting values (e.g. coding) as well as the structure of the data (e.g. structured
128 narrative);
129 • Medical dictionaries (MedDRA & VeDDRA) and product dictionaries (Extended EudraVigilance
130 medicinal product dictionary (xEVMPD) and the Union Product Database (UPD)) in use.
131
132 Question-specific determinants are defined as aspects of DQ that cannot be assessed independent
133 of the research question. In realm of ADRs, questions are generally structured as inherently related to
134 the safety of products, for example
135 • Specific queries (e.g. What is the overall risk-benefit profile of the drug? Is there an ADR possibly
136 caused by the drug? Are the Risk Management Plans and/or Risk minimisation measures [3]
137 working?);
138 • Specific analyses (e.g. Is there a subset of the patient population particularly affected?).
139 Therefore, question-specific determinants in ADR can be considered in terms of assessing the
140 suitability of a dataset for the different types of ADR question2.
141 3.1.2. Data lifecycle for ADR reporting
142 The EMRN DQF introduces an abstract data lifecycle, whose phases for ADR reporting can be defined as
143 follows:
2 This fundamentally maps to a higher level of the question-specific maturity model in the EMRN DQF, as we are considering
classes of related questions.
Page 7/49
第 8 页
144 • Definition of data requirements: what data are sought and what their characteristics should be.
145 For ADR reporting this is defined by international standards, such as ICH E2B(R3) [4] or VICH GL
146 35 [5] & GL 42 [6];
147 • Data collection or generation: gaining data that reflect the observed reality. This represents the
148 data fields in the ADR entry forms, the ADRs reported by the primary source to
149 MAHs/NCAs/Sponsors and ADRs or AEs identified by MAHs. The act of reporting an ADR to an
150 MAH/NCA/Sponsor is referred to as ADR reporting;
151 • Data management and processing: including data transfers, normalisation, and cleaning. This
152 is the ADR reporting and submission pathway with multiple reporting and submission options;
153 • Data submission: making data available to stakeholders. In the ADR reporting pipeline, this is
154 when reports are submitted to the EV databases (EV & EVV). An organisation performing this
155 submission is sometimes referred to in this document as a ‘data submitter’;
156 • Data procurement and aggregation: sourcing data from one or more stakeholders. For
157 example, EMA shares partially redacted versions of all EEA ADR reports with WHO-UMC to record
158 and analyse the data in VigiBase (see the EV Access Policy [7]);
159 • Testing and acceptance: assessing the suitability of the procured data for intended needs.
160 Testing is performed before an organisation is permitted to send ICSRs to the live production
161 database and acceptance is performed automatically by databases assessing whether the ICSRs
162 pass their business rules each time an ICSR is submitted;
163 • Delivery for consumption: using data to support a specific activity, e.g. analysis. The ADR
164 reports are forwarded from EV to NCAs and can be downloaded from EV by MAHs & NCAs for
165 additional analysis, and archiving.
166 3.1.3. DQ dimensions
167 The EMRN DQF introduces five DQ dimensions:
168 • Reliability: Are data faithfully representing what they are meant to be?
169 • Extensiveness: Are data sufficient?
170 • Coherence: Are data analysable as a whole?
171 • Timeliness: Are data available at the right time?
172 • Relevance: Are data of the right kind?
Page 8/49
第 9 页
173 4. Background on ADR collection, management and
174 submission
175 4.1. ADR guidelines and legislation
176 For collection, management and submission of ADRs related to both human and veterinary medicines,
177 many guidelines and rigorous procedures are in place to promote the highest possible quality. An
178 overview of the guidelines for the characterisation of systems and processes underpinning data for
179 ADR collection, management and submission is listed in Appendix I.
180 In order to provide a better understanding of the DQ issues related to the larger ADRecosystem, we
181 summarise in this section the documents that are of particular relevance in supporting good DQ of ADR
182 reports. As a starting point, it is important to refer to the relevant international ICH and VICH
183 guidelines:
184 • ICH Harmonised Tripartite Guideline: Clinical Safety Data Management: Definitions and Standards
185 for Expedited Reporting E2A [8];
186 • ICH Implementation Guide for the Electronic Transmission of Individual Case Safety Reports
187 (ICSRs) - E2B(R3) Data Elements and Message Specification [4];
188 • ICH Harmonised Tripartite Guideline: periodic benefit-risk evaluation report (PBRER) E2C(R2) [9];
189 • ICH Harmonised Guideline: Post-approval safety management: definitions and standards for
190 management and reporting of individual case safety reports E2D(R1) [10];
191 • ICH Harmonised Tripartite guideline: Development Safety Update Report E2F [11];
192 • ICH Harmonised Guideline: A selective approach to safety data collection in specific late-stage pre-
193 approval or post approval clinical trials E19 [12];
194 • ICH Harmonised Guideline: Guideline for good clinical practice E6(R3) [13];
195 • ICH M1 MedDRA - Medical Dictionary for Regulatory Activities;
196 • VICH GL24: Pharmacovigilance of Veterinary Medicinal Products: Management of adverse event
197 reports (AERs) [6];
198 • VICH GL30: Pharmacovigilance of Veterinary Medicinal Products: Controlled list of terms [14];
199 • VICH GL35 (R1): Pharmacovigilance of Veterinary Medicinal Products: Electronic standards for
200 transfer of data [5];
201 • VICH GL42 (R1): Pharmacovigilance of Veterinary Medicinal Products: Data elements for
202 submission of adverse event reports (AERs) [15].
203 Moreover, relevant guidelines for human ADR reporting in the EEA are:
204 • Detailed guide regarding the EV data management activities by EMA [16] to help stakeholders
205 understand the processes of their data management activities;
206 • Guideline on good pharmacovigilance practices (GVP) Module VI – Collection, management and
207 submission of reports of suspected adverse reactions to medicinal products (hereafter referred to
208 as GVP module VI) for the collection, management and submission of ADR reports [17];
209 • Guideline on good pharmacovigilance practices (GVP) Module IX – Signal management (hereafter
210 referred to as GVP Module IX), which provides general guidance and requirements on scientific and
211 quality aspects of signal management [18]. It also describes roles, responsibilities, and procedural
Page 9/49
第 10 页
212 aspects in the setting of the EU signal management process overseen by the Pharmacovigilance
213 Risk Assessment Committee (PRAC);
214 • EMA Good practice guide on recording, coding, reporting and assessment of medication errors
215 [19];
216 • EU Individual Case Safety Report (ICSR) Implementation guide [20];
217 • EudraLex Volume 10, which contains multitude of guidelines and Q&As related to Regulation
218 536/2014 [21]
219 For clinical trials on human subjects, in addition to the aforementioned guidelines, sponsors and health
220 authorities may use a Risk proportionate approach in clinical trials [22]. This approach should be
221 adapted to the risk to the subject and/or trial integrity of the research carried out, as well as to the
222 risk related to the reliability of trial results.
223 Reporting requirements may vary based on safety surveillance stage of the licensed medicine based on
224 legislation as depicted in Table 1. The reporting timelines are relevant in the context of the DQ
225 dimension timeliness as discussed in Section 5.4. The Clinical Trials Coordination and Advisory Group
226 has produced guidance on the reporting of ADRs in clinical trials governed by Regulation (EU) No
227 536/2014 [23] in the form of a Question & Answer document [24] and Recommendations on the use of
228 Auxiliary Medicinal Products in Clinical Trials [25].
229 Relevant guidelines for veterinary AE reporting in the EEA are:
230 • Guideline on veterinary good pharmacovigilance practices (VGVP) Module: Collection and recording
231 of suspected adverse events for veterinary medicinal products [26];
232 • Veterinary Union Pharmacovigilance Database – Best Practice Guide [27];
233 • Guidance notes on the use of VeDDRA terminology for reporting suspected adverse events in
234 animals and humans [28].
235 • EU VICH adverse event report implementation guide [29]
236 • Guideline on veterinary good pharmacovigilance practices (VGVP) Module: Signal Management
237 [30].
238 For veterinary ADR reports, the reporting of suspected adverse events (AEs) is governed by Regulation
239 (EU) 2019/6 [31], particularly section 5 Pharmacovigilance, and Commission Implementing Regulation
240 2021/1281 of 2 August 2021 [32]. In addition, the requirements for the Quality management system
241 are outlined in Chapter 2, Article 4 of the EU Commission Implementing Regulation (EU) 2021/1281.
242 Veterinary ADR reports are used for signal management, compliance monitoring, and publication of
243 data. Guidance at VICH and EU level is also available for the management of AEs [26, 27, 33 & 34].
Page 10/49
第 11 页
244 Table 1: Overview of (study) phase and submission requirements according to applicable legislation
(Study) Submission timelines Applicable legislation Module for
phase ICSRs/AERs
Phase 1-3 Fatal/life-threatening SUSARs <7 Regulation (EU) No EVCTM
clinical trials days 536/2014), Clinical trial
(Human) Other SUSARs <15 days regulation [23]
Unexpected events*: CTIS
All SAEs: Annual Safety Report/
Development Safety Update Report
Phase 4 Fatal/life-threatening SUSARs <7 Regulation (EU) No EVCTM
studies days 536/2014), Clinical trial
interventional All SUSARs <15 days regulation [23]
(Human) Unexpected events*: CTIS
All SAEs: Annual Safety Report/
Development Safety Update Report
Phase 4 Design based on primary data Directive 2001/83/EC [35] EVPM
studies non- collection: Serious ADRs <15 days Regulation (EU) No
interventional Design based on primary data 726/2004) [36]
(Human) collection: Non-serious EEA ADRs Commission Implementing
<90 days Regulation (EU) No
All ADRs: Periodic safety update 520/2012 [37]
reports, frequency dependent on
European Union reference dates list
Post- Serious ADRs <15 days Directive 2001/83/EC [35] EVPM
Marketing Non-serious EEA ADRs <90 days Regulation (EU) No
solicited 726/2004) [36]
reporting, e.g. All ADRs: Periodic safety update Commission Implementing
PSP (Human) reports, frequency dependent on Regulation (EU) No
European Union reference dates list 520/2012 [37]
Post- Serious ADRs <15 days Directive 2001/83/EC [35] EVPM
marketing Non-serious EEA ADRs <90 days Regulation (EU) No
unsolicited 726/2004) [36]
reporting All ADRs: Periodic safety update Commission Implementing
(Human) reports, frequency dependent on Regulation (EU) No
European Union reference dates list 520/2012 [37]
Post- Suspected adverse events <30 Regulation (EU) 2019/6 EVVETPROD
authorisation days [31]
reporting
(Veterinary)
Post- Suspected adverse events without Regulation (EU) 2019/6 EVVETPROD
marketing delay and within 30 days from the [31]
surveillance date of the closure of the final
studies study report
(Veterinary)
245 * See 9. Note: Suspected adverse events recorded during clinical trials conducted with unauthorised
246 veterinary medicinal products should not be submitted to EVVETPROD
Page 11/49
第 12 页
247 4.2. Considerations for ADR terminology
248 The concepts that follow are described in a succinct non-exhaustive manner in order to support the
249 understanding of DQ considerations presented in this chapter. The full definitions, guidance and
250 requirements related to ADR reporting are presented, for human post-authorisation activities, in GVP
251 Module VI [17], for Clinical Trials in Regulation 536/2014 [23] and, for veterinary medicinal products,
252 in the VGVP Module: Collection and recording of suspected adverse events for veterinary medicinal
253 products [26].Figure 2 graphically depicts the ADR terminology in the Human domain. An adverse
254 event (AE) is any untoward medical occurrence in a patient or clinical investigation subject
255 administered a pharmaceutical product and which does not necessarily have to have a causal
256 relationship with this treatment [8]. As a general rule for activities defined as Organised Data
257 Collection Systems under ICH E2D(R1), all AEs should be collected in the stakeholder database(s)3.
258 If there is at least a reasonable possibility that the medicinal product is causally related to the AE, this
259 is called an adverse drug reaction (ADR)4 (Figure 2) [17]. For cases reported spontaneously by a
260 primary source, positive causality is implied and all spontaneous AEs should be treated as ADRs, unless
261 causality is explicitly denied by the primary source (see GVP Module VI, section VI.A.1.1). In the
262 human domain, only ADRs should be submitted to EV as an Individual Case Safety Report (ICSR).
263 By contrast, in the veterinary domain there is no distinction between AEs and ADRs. EU veterinary
264 legislation only refers to adverse events and not to adverse drug reactions. All suspected AEs should
265 be submitted to EVV (Figure 3). An ICSR or AER is a document in specific ICH E2B or VICH E2B format
266 for the recording and reporting of AEs/ADRs associated with the use of a medicinal product [17].
267 AEs/ADRs may occur following a single dose or prolonged administration of a medicine, they may
268 result from the combination of two or more medicines or from combinations with food or drinks, and
269 they can occur following use within or outside the terms of marketing authorisation (e.g. off-label use,
270 medication error, misuse, etc.).
271
272
273
274
275
276
277
278 Figure 2: Graphical overview of Human AE/ADR terminology
3
All Human AEs should be collected and recorded except in the context of Non-interventional post-authorisation studies
with a design based on primary data collection, if the protocol provides with a due justification for not collecting certain
adverse events (see GVP Module VI [17], section VI.C.1.2.1.1.), and except in the context of interventional clinical trials
where “The investigator shall record and document all adverse events, unless the protocol provides differently” (Regulation
(EU) No 536/2014 [23], Art 41(2) and ICH E19 [12])
4 An adverse drug reaction can be defined in different ways. Directive 2001/83/EC [35], Article 1(11) defines an Adverse
reaction as “A response to a medicinal product which is noxious and unintended”. ICH E2D(R1) [10] adds “The phrase
“response to a medicinal product” means that a causal relationship between a medicinal product and an adverse event is at
least a reasonable possibility. A reaction, in contrast to an event, is characterised by the fact that a causal relationship
between the medicinal product and the occurrence is suspected.”
Throughout this document, the term Adverse Drug Reaction (or ADR) should be understood in this sense within the human
domain* – always meaning that there is at least a suspicion of a causal relationship between drug and reaction.
*as described in footnote 1, the term ADR is also used to cover suspected AEs in the veterinary domain.
Page 12/49
第 13 页
279 Figure 3: Graphical overview of Veterinary AE terminology
280 ADRs can be distinguished into two types of reports [10]: solicited reports and unsolicited reports
281 (the latter is also known as spontaneous reports). The definitions of solicited and unsolicited reports
282 are described for the human domain in E2D(R1) [10] and GVP Module VI [17] and for veterinary
283 medicinal products, in the VGVP Module: Collection and recording of suspected adverse events for
284 veterinary medicinal products [26]. As a summary:
285 • Solicited reports are derived from organised data collection systems such as clinical trials,
286 registries, registry-based studies, post-approval patient support programmes, market research
287 programmes, prospective cohort studies or surveys. The data derives from uniform collection
288 methods and collected in a solicited manner;
289 • Unsolicited reports usually rely on the initiative of a patient (consumer), animal owner, healthcare
290 professional (HCP), pharmacist, or lawyer to report an ADR to an NCA (possibly via a regional
291 pharmacovigilance centre) or MAH.
292 • Reports originating from various sources such as medical literature reports or from non-medical
293 sources, such as lay press or digital platforms (e.g. social media channels), can be part of either
294 solicited or unsolicited reports, depending on the protocol for gathering the information (digital
295 platforms) or how the information was gathered by the author (medical literature). Even if they are
296 classed as unsolicited, reports from the medical literature, and other sources where the MAH has
297 identified information rather than having received information, should initially be considered as
298 adverse events, rather than adverse reactions unless the AEs/ADRs are clearly suspected by the
299 article’s author to be associated with one or more medicinal products. As described in ICH E2D(R1)
300 [10], where the author’s suspicion is not stated or not available, MAHs should consider the
301 relationship between products and events in this context to determine if AEs/ADRs from literature
302 should be reported. If an author explicitly states in an article that an event is not associated with a
303 medicinal product, or the event occurred before the patient was exposed to the product, the MAHs
304 should not report it as an ICSR.
305 Although the terminology of the types of reports has been defined in ICH E2D(R1) [10] and in GVP
306 Module VI [17], the correct report type is not always used, leading to potential undersubmission
307 (noncompliance) or oversubmission of ADRs to regulatory authorities.
Page 13/49
第 14 页
308 4.3. Processes and systems underpinning ADR reporting
309 4.3.1. ADR data lifecycle
310 In this section we describe the systems and processes that underpin ADR reporting, framing them in
311 the context of the data lifecycle introduced in the EMRN DQF (see also Section 3.1. ). DQ of ADR data
312 is derived from the interplay of multiple actors, and we introduce a holistic overview of the ADR
313 reporting pipeline, including the main actors involved in the data lifecycle, along with the main
314 components and stakeholders of this flow (Figure and 5).
315 Step 1: data requirements
316 This is incorporated by specific data fields in the ADR reporting forms. These forms have been released
317 by many different organisations at different responsibility levels e.g. regional/national PV centres,
318 MAHs and Sponsors. Variations in data entry forms can lead to DQ issues, including coherence,
319 extensiveness and precision.
320 Step 2: data collection or data generation
321 ADR reports are generated by the primary source (the person who reports the facts about an ADR)
322 such as patients or relatives, animal owners, HCPs, lawyers and, for clinical trials, investigators. The
323 ADR reports can also be identified through another medium such as a digital platform or a literature
324 and in this instance it is not directly generated by the primary source but by an MAH or a regulatory
325 authority (RA) (see ICH E2D(R1) [10]). Lack of education or awareness on ADR reporting for HCPs or
326 training at MAHs, CROs & Sponsors can affect extensiveness, specifically underreporting,
327 incompleteness; and coherence issues such as semantic coherence (inconsistent wording). Primary
328 sources can report to more than one organisation (e.g. to an MAH and to an NCA in parallel) or an
329 article may be identified by multiple MAHs. This is referred to in this document as duplicate
330 submission.
331 Step 3: data management and processing
332 Multiple reporting routes are possible, and it does not necessarily follow a linear path. A distinction is
333 made between reporting routes:
334 1) Post-Marketing (PM) EEA reporting route - the ADR occurred in the EEA;
335 2) PM non-EEA reporting route – the ADR occurred outside the EEA but needs submitting within
336 EEA;
337 3) CT (Human) reporting route
338 For the PM EEA reporting route, ADRs originating from the primary source may be reported directly to
339 an MAH, an NCA or to a regional/national PV centre. Identified literature and digital platform cases
340 (EEA or non-EEA) should be reported by MAHs or, for certain literature and substances in the human
341 domain, by the EMA’s Medical Literature Monitoring (MLM) service. Serious PM ADRs from non-EEA
342 countries may be reported to the MAH, who may send the ADR reports to a third country regulatory
343 authority (RA) who may make these reports available to other MAHs, depending on local regulations.
344 Alternatively, non-EEA ADRs may be reported to a third country RA directly, from whom the MAHs may
345 obtain the ADR reports, depending on local regulations. The serious non-EEA ADR reports will be
346 submitted to EV by the MAH that obtains them, regardless of source or how many steps the report has
347 been through before reaching the MAH. The number of steps in the reporting route can lead to
348 duplication, data loss or data changes, due to corrections or compatibility issues with databases during
349 data management, coding refinements and processing, resulting in a loss or change of information or
350 differences in versioning. By contrast, the CT reporting path is far simpler. An ADR report is submitted
351 by the investigator to the Sponsor & then, following assessments of causality and expectedness, a
352 SUSAR will be reported by the Sponsor to EudraVigilance.
353
354 Step 4: data submission
Page 14/49
第 15 页
355 In this phase reports are submitted to EV/EVV. Only registered organisations (MAHs/NCAs/Sponsors)
356 are permitted to submit ICSRs to EV. Several DQ considerations apply to this step: is the ADR report
357 submitted in a well-formed XML and does it comply with the EV/EVV business rules? Acknowledgement
358 messages are returned to the submitting organisation to inform them whether their report was well-
359 formed and conformed to the business rules. Has the reporting been performed in a timely fashion? If
360 there has been duplicate submission, then this will lead to duplicate reports.
361 Step 5: data procurement and aggregation
362 In this step, the cases are loaded into EudraVigilance and processed to make available for data
363 analysis. The EV and EVV systems do not change the data received, but they do link reported
364 drug/substance to the xEVMPD and the UPD respectively, link ADR reports to cases and clusters and
365 EMA performs duplicate detection and management, manually linking cases identified as duplicates.
366 From EV and EVV the data can be used for regulatory activities or other purposes as secondary
367 research.
368 Step 6: delivery for consumption
369 In this step in the human domain, post-marketing ADR reports can be downloaded from EV by MAHs.
370 ADR reports are also rerouted to NCAs, or are downloaded by NCAs, to collect and analyse. Different
371 stakeholders have different levels of access [7].
372 For EEA ADR reports, EMA copies the ICSRs and changes the sender (EV) and receiver (NCA) and
373 reroutes them to the relevant NCA (or NCAs can download the ADR reports, as they prefer). Moreover,
374 a version of the case is made available to MAHs without identifiable information (e.g. date of birth,
375 initials), and without the narrative and most other free text fields (and a version with the free-text
376 fields can be accessed by MAHs for signal investigation). Furthermore, EMA shares partially redacted
377 versions of all human EEA ADR reports with World Health Organisation - Uppsala Monitoring Centre
378 (WHO-UMC) to collect and analyse the data in VigiBase. Full details of the data available to each
379 stakeholder are in the EV Access Policy [7].
380 There can be a DQ risk in this step, when the only identifier in the ADR report is the date of birth or
381 patient initials: if this identifier is removed for MAHs, the case is no longer valid and may not be able to
382 be processed in the MAH’s database.
383 In this step, in the veterinary domain, the AE reports can be downloaded from EVV by MAHs and NCAs.
384 Different stakeholders have different levels of access [38]. Additionally, NCAs with their own databases
385 may request rerouting of AERs from EVV to their national systems.
386 4.3.2. ADR reporting pipeline
387 The journey of ADR reports from initiation (data collection) to delivery for consumption is long. The
388 proverb ‘no chain is stronger than its weakest link’ is clearly valid. A problem at one level may not only
389 be continued but exacerbated at a later stage.
390 Because ADR data is flowing from the reporter to the final receiver(s), the path that the data follow is
391 described throughout this document as a “pipeline”.
392 DQ checks occur at various steps of the data lifecycle and may include iterative feedback loops. Not all
393 phases are present in all data workflows and the order may differ. The “testing” part of the step
394 ‘testing and acceptance’, because it is performed intermittently, is not applicable in the ADR reporting
395 pipeline.
396
Page 15/49
第 16 页
397 Figure 4: Overview of the human ADR reporting pipeline across stakeholders along the data lifecycle
398
399
400
401 Figure 5: Overview of the veterinary AE reporting pipeline across stakeholders along the data lifecycle
402
403 4.3.3. EudraVigilance: the system underpinning human ADR reporting
404 EudraVigilance (EV) is the data-processing network and database for managing and analysing
405 information on human ADRs from the EEA and the rest of the world, concerning medicinal products
406 available in the EEA, since 2001, which operates on behalf of the EMRN. EMA is responsible for
407 operating procedures to ensure the quality and integrity of the information collected in EV in
408 collaboration with NCAs and MAHs. EV has processes in place to ensure good DQ. The data in EV
409 comes from many sources - in 2025 ADR data was submitted by more than 2,000 different senders
410 (MAHs, NCAs and Sponsors).
411 After marketing authorisation of a medicinal product, human ADR reports (both solicited and
412 unsolicited) shall be reported to the EV post-authorisation Module (EVPM) [20] by MAHs and NCAs.
Page 16/49
第 17 页
413 Reports of suspected unexpected serious adverse reactions (SUSARs) shall be reported to the clinical
414 trial module (EVCTM) of EV by Sponsors of interventional Clinical Trials [20]. In the situation where a
415 medicinal product application is under evaluation in the EU while it has already been authorised in a
416 third country, valid ICSRs from outside the EU should be submitted to EVPM [17] EV is operated by
417 EMA and contains ADR reports originating from all around the world: within the EEA both serious and
418 non-serious ADRs; outside EEA only serious ADRs. In EV, medicines are reported free-text and
419 classified against the medicinal product dictionary (xEVMPD). ADRs, medical history, test names and all
420 other medical terms are coded in accordance with the MedDRA® terminology (Medical Dictionary for
421 Regulatory Activities) [39, 40].
422 4.3.4. EudraVigilance Veterinary: the system underpinning veterinary AE
423 reporting
424 The safety and efficacy of animal medicines after they have been authorised and made available on the
425 market include assessment of unsolicited AEs, and AE reporting. EVV is the system that supports the
426 implementation of the Veterinary Medicinal Products Regulation[31, 41]. EVV uses VeDDRA for coding
427 of suspected AEs and all other medical terms in animals and humans. In EVV, medicines or active
428 substances may be reported by selecting an entry from the veterinary product dictionary (via the EVV
429 web application) or using free-text, which are then classified against the UPD.
430 4.4. Current features of ADR data
431 When it comes to the intrinsic aspects of ADR data (Section 3.1. ), the objectives of ADR reporting
432 dictate that some minimum criteria elements are needed for a valid report:
433 • One identifiable patient (human or animal) or number of treated animals (veterinary);
434 • Adverse reaction(s);
435 • Suspected medicinal product(s);
436 • Identifiable reporter.
437 Other data elements critical for scientific evaluation present in a report may include:
438 • Patient information: age, sex, height, weight, species (veterinary only), past drug therapy, and
439 medical history (including history of previous ADRs).
440 • Reaction section: description of the ADR with relevant clinical signs and symptoms, date of (or
441 time to) onset of reaction, ADR outcomes, seriousness, actions taken and duration.
442 • Drug section: suspected, interacting and concomitant5 medicinal product information: treatment
443 dates, product dosage, route of administration, indication, batch number (particularly for vaccines
444 or biological products), dechallenge, rechallenge, time to onset, causality assessment (solicited
445 reports), expectedness assessment (Human CT reports – the decision needs to be performed and
446 documented).
447 These data are captured both in a structured and unstructured way (quite often some narrative text
448 provides details on the structured information). The level of normalisation of structured data (e.g. by
449 using standard vocabularies) can also vary.
450 Additional information on suspect and concomitant medicines adds more value to the ADR report by
451 including medication name, pharmaceutical form, indication, dose, route of administration, dosing
5 In veterinary medicine there is no distinction made between suspect and concomitant medication
Page 17/49
第 18 页
452 interval, duration and batch number. Past and current medical history is also very relevant information
453 to assess whether the ADR can be possibly explained by other factors.
454 4.5. Considerations for use of ADRs in regulatory activities
455 The identification, documentation and reporting of ADRs is fundamental for both clinical trials and PV
456 activities such as determining whether a product can receive a marketing authorisation, detecting early
457 signals to ensure the benefit-risk balance of a drug remains favourable throughout its lifecycle and to
458 support regulatory decision making. PV activities are particularly essential for detecting rare ADRs
459 sometimes occurring in special populations or after prolonged exposure, and drug–drug or drug–
460 disease interactions that may not have been observed in randomised clinical trials prior to
461 authorisation. Special attention needs to be given to such reactions in order to prevent them as,
462 without spontaneous reporting, potential safety signals may remain undetected for years, thereby
463 exposing the public or the animal population to unanticipated risks. Therefore, effective and good
464 quality ADR collection, management and submission is crucial for identifying potential drug-related
465 risks, implementing risk minimisation measures, and improving patient safety.
466 Notably, for small populations, such as clinical trials, rare diseases and orphan drugs, the DQ of ADRs
467 is even more important because each case is individually representing a greater fraction of the patient
468 population and is statistically probably more impactful than a case concerning a widely used medicine.
469 4.5.1. Use of ADRs in human regulatory activities
470 On the ADRs collected via EV, EMA and NCAs perform analyses for several purposes for which
471 “question specific” determinants can be defined.
472 SUSARs (transmitted to the EV Clinical Trials Module – EVCTM) are primarily collected in EV to permit
473 Member States (Ethics Committees and NCAs) to assess whether a Clinical Trial should be permitted to
474 continue, to supervise whether the sponsor is conducting adequate CT supervision including ensuring
475 continuous positive benefit-risk of the CT and whether the protocol needs amending, and to inform on
476 which ADRs are already known at the time of marketing authorisation application. Additionally,
477 individual cases from Clinical Trials also contribute to aggregate safety assessments, including
478 Development Safety Update Reports/Annual Safety Reports, benefit-risk evaluation and emerging
479 safety issue assessments. Furthermore, Sponsors’ data safety monitoring boards will review individual
480 cases from Clinical Trials within the Sponsor’s database to perform their safety assessments.
481 The primary purpose of collecting post-marketing ADRs in the EV Post-Authorisation Module (EVPM) is
482 for safety activities, mainly focusing on signal detection and a deeper evaluation of ADRs. The EEA
483 regulatory network process from ADR submission to EV to signal assessment in the human domain is
484 described at a high level in GVP Module IX [18] and in detail in the guidance document Screening for
485 adverse reactions in EudraVigilance [42]. A simplified version is depicted in Figure 6). A pre-processed
486 cumulative output from EV is reviewed periodically by an EMA/NCA PV-assessor. In this output, some
487 calculations have been performed to guide the PV-assessor: total number of cases per MedDRA PT,
488 number of new cases, number of fatal cases, etc. Several criteria may be applied to assist with
489 determining whether an ADR requires further investigation, such as type of reaction (serious/non-
490 serious), reaction outcome (e.g. recovered, fatal, etc.), known pharmacological profile, patient
491 population type (e.g. age group, sex), product type, indication, reporting frequency, and other aspects,
492 such as inclusion of the reaction on the Important Medical Event Terms/Designated Medical Event lists
493 [43, 44]. Additionally, MAHs and NCAs perform signal detection on their own databases.
Page 18/49
第 19 页
494
495 Figure 6: Overview of use of human ADRs for routine post-marketing signal management after
496 submission to EudraVigilance.6 As described in section 4.3.1. and Figure 4, the ADRs submitted to EV
497 may be rerouted to NCAs. Therefore, the first step may be performed in EV or an NCA database.
498 Statistical methods are applied to EV data, e.g. reporting odds ratio (ROR) among others, to support
499 signal detection. Data supporting the detected signal are evaluated to verify the availability of
500 sufficient evidence demonstrating the existence of a new potentially causal association, or a new
501 aspect of a known association, therefore justifying further analysis, taking into account the strength of
502 the evidence, the clinical relevance and previous awareness of the association (see GVP module IX
503 [18]). Signals which meet this threshold and are considered to require assessment by the PRAC are
504 discussed at EU level (see Figure 6) [45]. The European pharmacovigilance issues tracking tool EPITT
505 provides the functionalities for regulatory authorities of Member States and EMA to track signals at EU
506 level.7 The list of signals assessed by PRAC is published on the EMA website [46].Other question-
507 specific activities include accommodating requests for information from stakeholders as MAHs, and
508 EMA committees and targeted analyses. It is worth noting that signals can also be identified via
509 Periodic Safety Update Reports, which can include ADR data that was not transmissible to
510 EudraVigilance.
511 4.5.2. Use of AEs in veterinary regulatory activities
512 The primary purpose of collecting AEs in EVV is for safety activities, mainly focusing on signal
513 management and a deeper evaluation of AEs. The process from AE submission to EVV to signal
514 management in the veterinary domain is depicted in Figure 7. An MAH /NCA PV-assessor may access
515 pre-defined queries in EVV DWH, such as the Signal detection query or Adverse events overview
516 query. In these reports, some calculations have been performed to guide the PV-assessor: total
517 number of cases per VeDDRA PT, number of new cases, overall number of fatal cases, etc. The PV-
518 assessor investigates AEs based on several priorities such as type of reaction, reaction outcome (fatal),
519 known pharmacological profile, patient population type (e.g. species, age group, sex), product type,
520 indication, reporting frequency, and other aspects.
521 Specific statistical calculations and algorithm are performed in EVV, e.g. reporting odds ratio (ROR) to
522 identify signals of disproportionate reporting. The MAH/NCAs PV-assessors periodically check AEs
523 submitted to EVV and may use the Medical Important VeDDRA term (MI) lists, Appendix 1 to [47], as
524 well as aggregated data analysis (ROR) and case narratives. The MAH/NCA PV-assessor submits the
525 validated signals to the signal management module of the Union Pharmacovigilance database (IRIS).
526 Signals are considered by the relevant NCAs and committees for further assessment and decision-
6 It is worth noting that signals can be triggered by sources other than ADRs, such as articles in the medical literature or
study reports. In these instances, although the starting point would differ, the subsequent processes would be the same
7 This paragraph is largely describing the process of signal detection/management for Centrally authorised Products. For
Nationally Authorised Products, Member State worksharing also needs to be taken into account.
Page 19/49
第 20 页
527 making (see Figure 7). IRIS provides the functionalities for regulatory authorities of Member States
528 and EMA to track signals at EU level, and the list of signals is also published on the IRIS public register.
529
530 Figure 7: Overview of the signal management process for veterinary medicinal products
531 4.5.3. DQ considerations for ADRs in regulatory activities
532 In these steps of signal management by EMA, there are several DQ considerations. To start with, the
533 pre-processed cumulative outputs are generated automatically from Eudravigilance by EMA. This
534 process has been robustly validated by EMA and NCAs, and the code for generating the reports for EMA
535 and NCAs includes the generation of outputs which require manual validation and control. Poor data
536 quality in the ADR reports will negatively impact the outputs generated for regulatory activities, leading
537 to weaker aggregate safety conclusions than could be generated from good quality ADR reports.
538 To ensure that the reliability is high and the risk of inadequate assessments is minimised, EMA has
539 published work instructions for the design of an analysis in EV, which are followed when performing
540 assessments [48]. Regular feedback sessions as described in Section 6.5. are recommended. Also,
541 regular training of PV-assessors is recommended.
542 The aggregate safety assessments for Clinical Trials, including Development Safety Update
543 Reports/Annual Safety Reports, benefit-risk evaluation and emerging safety issue assessments, are
544 based on individual cases (both the SUSARs submitted to EV and other cases (unrelated, expected or
545 non-serious), which do not require transmission to EV) and thus the completeness and quality of
546 individual reports is also important for these aggregate safety activities.
547
Page 20/49
第 21 页
548 5. Managing challenges in ADR reporting
549 Across the ADR reporting cycle, robust and thorough processes should be in place (regulatory
550 frameworks, pharmacovigilance and quality system checks, training, record and compliance
551 management, mechanisms for timely and effective communication regarding safety concerns, SOPs,
552 work instructions and medical review sessions) for collection, management and submission of reports,
553 sometimes even at automated level [16, 31]. This includes support for case entry, coding and, where
554 practical, verification by the 4-eye principle (approval by at least 2 people). Nonetheless, some
555 challenges with DQ emerge. In this section, we provide an overview of DQ issues organised by DQ
556 dimensions (as summarised in Figure 8) with a brief discussion on how they originate from the systems
557 and processes underpinning ADR reporting and, where relevant, how they manifest in data.
558 Figure 8: Challenges in ADR reporting organised by DQ dimensions
559 5.1. Reliability
560 As described in the EMRN DQF, the DQ dimension reliability covers how closely the data reflects what
561 they are directly measuring. Are data faithfully representing what it’s meant to be? Subdimensions of
562 reliability are accuracy (the amount of discrepancy between data and reality) and precision (the degree
563 of approximation by which data represents reality). Related to reliability is also the concept of
564 traceability (are data presenting the knowledge of how data were reported, what source it originated
565 from and what processing it went through before appearing in its current form) [1].
566 5.1.1. Noise (accuracy)8
567 In the human domain the collation of ADR reports available for analysis is affected by noise: ADRs and
568 events are mixed in a way that makes it difficult to separate important signals from false signals.
569 Typical causes of noise are reporting and/or submitting incorrect data (e.g. in the human domain
570 adverse events rather than adverse reactions), duplicate reporting, duplicate submission,
571 overreporting by MAHs, and reporting of specific ADRs or of non-specific ADRs to particular drugs
572 stimulated by media coverage, which can result in reporting bias.
573 Noise due to incorrect reporting manifests in ADR reports for which relatedness cannot be assessed
574 with sufficient confidence, requiring a large number of cases to establish the reliability of a signal that
8 We note that noise arises from a collection of DQ issues not limited to reliability. For instance, over or under-
representation of ADRs also contributes to noise, as well as duplicates and lack of coherence. In this chapter, we opted to
consider noise under reliability, as per general practice and consideration of noise in signal detection.
Page 21/49
第 22 页
575 in turn may have consequences for the timeliness of (downstream) regulatory actions. The inadequate
576 assessment of causality is described in Section 5.1.2.
577 Moreover, noise is becoming an increasing issue as the volume of ADRs from particular sources has
578 substantially increased in recent years. Since GVP Module VI [17] was released in 2012, patient
579 support programmes (PSPs), market research programmes and digital platforms have specifically
580 impacted the rising number of ADR reports [49]. The effects on DQ of post-marketing reporting in the
581 context of PSPs and digital platforms is described in Section 5.6. Notably, particularly concerning PSPs
582 and digital platforms, there may be a trade-off in quality versus quantity in ADR reporting.
583 To improve accuracy, good knowledge of submission rules by Sponsors, NCAs and MAHs is crucial to
584 minimise noise (See section 6.1. – Education). Specifically, it is recommended to promote good
585 adherence to reporting rules to EV and EVV.
586 5.1.2. Inadequate assessment of causality (accuracy)
587 One of the key issues related to false positives and noise in the human domain, is the inadequate
588 assessment of causality in solicited reports that results in submission of reports that are in fact not
589 ADRs as there is no causal link between the medicine administered and the AE observed.
590 In the context of clinical trials, SUSAR reporting is particularly challenging because the assessment of
591 whether an unexpected event is suspected (i.e. considered related) can be difficult and create unstable
592 DQ. As a result, some sponsors of clinical trials may report every SAE as a SUSAR, which leads to
593 overreporting of (often) non-drug-related SAEs. On the other hand, other sponsors may not treat all
594 SUSARs as SAEs and not submit them, which leads to underreporting. This can impact adequate and
595 timely signal detection.
596 Inadequate assessment of causality in clinical trials may be caused by insufficient training of
597 Investigators. It requires thorough training to analyse the causality of an ADR, because the
598 assessment of causality is partly based on time-to-onset, the pharmacological mechanism of action,
599 the patient’s medical history and concomitant medications, and potential adverse reactions. This
600 touches upon foundational determinants, as study protocols, work instructions and SOPs may not be
601 clear enough. For the primary source (consumers, HCPs, Investigators), inadequate causality
602 assessment can also relate to intrinsic determinants, such as a lack of structure in the ADR data entry
603 form. The data entry forms for ADRs do not always make reporters think systematically about the
604 report, which might lead to over-complete case narratives with unstructured information, without a
605 clear causality assessment.
606 To improve the adequacy of the causality assessments and consequently improve accuracy, better
607 training and education of investigators is recommended (see Section 6.1. - Training). Better
608 understanding of the pharmacological profile is of importance (see Section 6.2. - Integration of clinical
609 pharmacology in ADR processes). Moreover, it is recommended that ADR reporting forms are
610 improved, particularly regarding causality. For clinical trials reporting forms, investigators should be
611 encouraged to document the reasoning supporting the causality assessment, including key elements
612 such as temporal relationship, alternative explanations, dechallenge/rechallenge information and
613 relevant medical history (see 6.3. – Improvement of standard implementation).
614 5.1.3. Precision
615 The challenge of precision is the lack of specificity in the description of an ADR. This can arise when
616 multiple patients or events are conflated, which can happen for a variety of reasons.
Page 22/49
第 23 页
617 In human ADR reports, a lack of precision can emerge when an MAH incorrectly makes a single case
618 referring to multiple patients described in a literature article or, by contrast when an MAH operating a
619 PSP makes a single case for a patient that covers many years of their life, conflating many events and
620 reactions which should be separate cases into one. Additionally, a lack of precision can emerge from
621 the selection of an overly broad MedDRA or VeDDRA term, rather than the most-specific term(s) for
622 the reported reaction(s).
623 In the human domain, a lack of precision can emerge due to a lack of format coherence (see 5.3.2. )
624 because PV databases may still be based on the ICH E2B(R2) format rather than the ICH E2B(R3)
625 format. A key difference between these formats is that in R2 seriousness is captured at case level,
626 whereas in R3 seriousness is captured at reaction level. This means that if an ADR report describes a
627 patient who experienced a headache and a fatal myocardial infarction, in R3 only the myocardial
628 infarction would be classed as fatal; whereas for the same report stored in R2 format both reactions
629 would be classed as fatal.
630 In veterinary AE reports for instance, for animals treated as a group, one report can refer to
631 approximately 1,000 animals, often identified by age/weight range (or in the case of fish, by tonnage).
632 AE reports include and conflate different events with no specificity to the individual animal. This is
633 different for pets, where AEs are usually reported on an individual basis.
634 A potential challenge related to precision based on an intrinsic determinant of DQ in veterinary AE
635 reporting, compared to human ADR reporting, is the lack of granularity in VeDDRA compared to
636 MedDRA.
637 5.1.4. Traceability
638 Data lineage (or traceability) describes the history of data, detailing its origins, the sources it came
639 from and the transformations it underwent before reaching its current state [1]. The pipeline for ADR
640 reporting can be extensive with many stakeholders involved in the same case.
641 Traceability issues manifest in missing links between ADRs, when these derive from each other (e.g.
642 mother & baby reports) or from the same primary source (e.g. same reporter or same literature
643 article).
644 Retention of the integrity of the report, which should remain as faithful as possible to the source data
645 (e.g. the original report, follow-up requests and replies), is considered a challenge in ADR reporting.
646 The primary source narrative may change along the ADR reporting pipeline, and the original narrative
647 is not always available. This is specifically important for health authorities, who receive the data from
648 MAHs and then distribute it to other MAHs. If different stakeholders have adapted versions of the same
649 ADR report this leads to reliability issues (limited traceability, data entry errors, transcription errors,
650 translation errors), and coherence issues (duplicates).
651 Ensuring traceability and maintaining data ownership9 as the ADR travels through the pipeline could
652 improve reliability and could reduce duplication. To improve traceability, it is recommended that the
653 entire process of data management (SOPs, manuals, trainings) is documented, true, verifiable, and
654 versioned. Moreover, it is recommended that the concept of “data immutability” is applied to ADR data
655 (See Section 6.6. - Maintain immutability of data at the point of transmission). Lastly, to improve
656 traceability, good feedback across the ADR pipeline would improve DQ (See Section 6.5. -
657 Implementation of feedback mechanisms to allow for data corrections).
9 In this context, data ownership refers to the right and responsibility of senders of ADR data to correct/update that data
through the submission of follow-up versions of the originally submitted reports
Page 23/49
第 24 页
658 5.2. Extensiveness
659 As described in the EMRN DQF, extensiveness is defined as the dimension capturing the amount of
660 data available. Extensiveness answers the question, “how much data do we have”? When considering
661 the “fit for purpose” definition of quality, extensiveness covers how sufficient the data are for the
662 intended purposes – clinically meaningful safety information needs to be as complete as possible, but
663 other information is less relevant, and more data of this type is not always better. Subdimensions of
664 extensiveness are coverage, and completeness [1].
665 5.2.1. Underreporting (coverage)
666 Underreporting in ADRs is the lack of reported ADRs that could have been reported. Underreporting is
667 a significant challenge in PV. It has been estimated that only 5-10% of spontaneous ADRs & AEs (both
668 human & veterinary) are reported, indicating that the majority of the cases is unknown [50, 51].
669 The level of reporting is likely dependent on seriousness and may be relatively lower for non-serious
670 events, well-established side-effects and for ADRs specific for certain medicines. This problem may
671 lead to a reporting bias. Furthermore, variations between ADR reports are often influenced by HCP
672 attitudes and beliefs, knowledge and skills, clinical expertise, and practice settings, as well as
673 familiarity with local protocols and reporting requirements and possibilities from regulatory bodies
674 [52]. This can lead to issues of extensiveness (underreporting), completeness, and coherence.
675 To improve coverage, continuing education and communication programmes for the primary source are
676 recommended (see Section 6.1. – Training).
677 5.2.2. Missing information (completeness)
678 Completeness issues in ADR reports are missing information that ideally would have been provided (or,
679 in the case of information that was provided by the reporter to the sender organisation, should have
680 been provided) because this information is both necessary and useful to assess an ADR.
681 Unsolicited reports can have a low level of completeness, with data either not reported at the start or
682 after follow-up, or lost in the journey (unintentionally, e.g. through data entry errors; or intentionally
683 to protect privacy, e.g. when implementing the General Data Protection Regulation (either correctly or
684 incorrectly) in the human domain).
685 In 2025, a total of 2,002,916 valid ADR reports were submitted to EV of which 1,181,416 (59%)
686 included both time-to-onset data and reaction outcome. The lack of information in the other 41% of
687 cases especially complicates signal validation by affecting reliability. The value of ADR reports in signal
688 detection is inherently related to the completeness of clinically relevant information [53, 54]. An
689 incomplete or low-quality report undermines the strength of the association between a medicine and
690 the reaction. Furthermore, a low level of completeness can create problems with uniqueness (risk of
691 apparent duplicates).
692 The challenge with missing information is related to foundational determinants. In the current
693 processes, it is possible to reports AEs and ADRs to stakeholders without even sufficient information to
694 make the report valid and to submit AE and ADR reports with only the minimum criteria required to
695 validate the case (see Section 4.4. ). Furthermore, for patients or animal owners with limited medical
696 or veterinary knowledge, it may be challenging to provide precise medical/veterinary information and
697 to know what ancillary medical/veterinary information (e.g. test results) would be useful. Moreover,
698 HCPs and vets may lack an understanding of the work of PV and the information that assessors need.
699 Also, time constraints can hinder complete reporting of potential ADRs. In addition, MAHs often need to
700 seek permission to obtain follow-up and this is often denied or retracted after the first follow-up.
Page 24/49
第 25 页
701 To improve completeness, it is recommended to include assessments of completeness in the DQ
702 processes focussing on key-data elements (6.4. – Early implementation of data validation processes)
703 [53, 55]. It is recommended that, where appropriate and in addition to established pharmacovigilance
704 quality systems, sender organisations develop general metrics and also metrics that are specific for
705 certain products or reactions of particular interest, because key data elements may differ for each
706 product or ADR. For example, a key element of the reaction could be to include relevant test data, e.g.
707 an ADR of a drug-induced liver injury should ideally include liver function tests.
708 Moreover, to improve completeness it is recommended that automated checks are integrated in the
709 database and electronic reporting tools to reduce missing information upon data entry (see Section
710 6.4. - Early implementation of data validation processes). Additionally, a clear explanation and/or
711 automated notifiers will guide the reporter (patients and HCPs) to provide the critical data elements
712 resulting in higher DQ.
713 One significant problem affecting completeness is the difficulty in follow-up on an initial ADR report to
714 receive more detail, particularly for unsolicited ADR reports. Initial ADR reports may be incomplete for
715 a variety of reasons, including legitimate ones such as speedy reporting for severe cases. Although
716 follow-up information is very important for DQ of the ADRs and its use in signal detection, especially
717 because it often provides more information on the diagnosis, nevertheless, MAHs report that the
718 majority of follow-up attempts for unsolicited reports are unsuccessful [56]. By contrast, in clinical
719 trials the sponsors usually have direct access to investigators and structured mechanisms for obtaining
720 additional information. As a result, follow-up success rates and completeness are generally higher than
721 reports from other sources.
722 Furthermore, there is heterogeneity in the way follow-up is reported, leading to coherence issues. If
723 follow-up is provided, the extensiveness of the report ranges from one-word answers to the specific
724 questions, to the full case notes of the patients. Moreover, some NCAs and MAHs submit a separate
725 small follow-up summary, while others submit a fully integrated follow-up in the original narrative.
726 When the full case notes are provided, the NCA or MAH needs to decide what is relevant, which can
727 lead to incomplete or excessive information. Next to missing information (completeness issues), the
728 heterogeneity in reports can lead to issues with coherence.
729 Prior to 2025, a potential cause for this challenge was a lack of guidance on how to acquire and report
730 follow-up information. This has been somewhat alleviated in the human domain by the publication of
731 the Guideline on specific adverse reaction follow-up questionnaires [57]. A cause for the variability in
732 the ways of reporting follow-up may be a lack of structure in the data entry forms to incorporate
733 follow-up.
734 To improve completeness of follow-up information, it is recommended that customisable template
735 forms are developed to standardise information collected at the primary source (see Section 6.3. -
736 Improvement of standard implementation).
737
Page 25/49
第 26 页
738 5.3. Coherence
739 As described in the generic EMRN DQF, coherence is the dimension that expresses how different parts
740 of an overall dataset are consistent in their representation and meaning. Coherence answers the
741 questions: is the format of values the same across the dataset? Is the precision of values the same?
742 When considering the “fit for purpose” definition of quality, coherence relates to the analysability of
743 data. Subdimensions of coherence are uniqueness, format coherence, and semantic coherence [1].
744 5.3.1. Duplicates (uniqueness)
745 Duplicate ADR reports are the problem of multiple same records being present in a database without
746 using coherent identifiers and hence making it hard to detect them. This in turn creates accuracy
747 issues in that it distorts the relation between data, and what is observed, in terms of sampling biases.
748 Duplicate reports pose several problems for PV by creating false or misleading signals and therefore
749 impact the safety monitoring and potential regulatory actions. The most obvious challenge is related to
750 signal detection when the observed ADR counts are higher than the expected ADR counts. Duplicates
751 arise from various causes as depicted in Figure 9.
752 Some duplicates arise by manual errors such as incompleteness of reports, data entry errors or
753 multiple reporters. There may be multiple reports in EV or EVV about the same ADR in the same
754 patient, reported by different senders through different routes, whereby the ADR information may be
755 handled differently by the processor of the case, which makes it difficult to identify the reported cases
756 as duplicates. It could also happen that a single patient or HCP reports the ADR in multiple places.
757 Another possibility is that follow-up information may be mistakenly unlinked to the original report, thus
758 creating a new report [58].
759 Other causes appear due to computational errors such as a lack of standardised management
760 practises, multiple representation of the same data due to the merging of PV databases and challenges
761 in data integration and harmonisation in PV databases.
762 There can also be automated causes of duplicates when organisations automatically share data with
763 other organisations that already receive copies of it (e.g. MAH-MAH sharing as business partners10);
764 this type of duplication is also known as ‘case replication’ and appears to be a greater problem in the
765 human domain than veterinary. Replication includes reproducing or recreating ADR reports for various
766 purposes, for example, research, regulatory compliance or data analysis. Replication proceeds by
767 making multiple copies of reports for different stakeholders or systems without necessarily meaning
768 that the content or details of the reports are identical. Replication could result in duplication, when
769 multiple stakeholders report the same case to the same regulatory authority for compliance. This is an
770 important DQ problem as it also undermines data integrity [58, 59].
771 To improve uniqueness and coherence, it is recommended that databases are routinely screened to
772 detect and eliminate duplicate reports by collaboration of stakeholders across the ADR pipeline (see
773 Section 6.7. - Routine duplicate detection and elimination by databases).
774 Furthermore, to reduce case replication across multiple partners, it is highly recommended that data
775 interchange agreements are developed with a focus on eliminating the submission to EV of the same
776 replicated case by multiple MAHs. These agreements should be regularly reviewed particularly when
777 new partnerships are initiated.
10 A business partner is an organisation with a medicinal product licensed to another pharmaceutical company.
Page 26/49
第 27 页
778 Figure 9: Causes of duplicate reports classified by manual or computational cause
779 5.3.2. Different ways of reporting data (format coherence)
780 Format coherence is impacted by the use of different ways to represent data from different sources in
781 a collection of ADRs.
782 An important challenge in format coherence, is that data is collected using different collection methods,
783 different data collection forms and, for non-EEA-regulators, even different systems.
784 In clinical trials, structured data collection processes are in place using case report forms. For post-
785 marketing unsolicited reports, data can be gathered in multiple ways. Most ADR reports are reported
786 electronically by the primary source, however, some are reported by phone (especially for PSPs) and a
787 small proportion is reported by regular mail, e-mail or fax, and entered by the recipient in a form
788 which often follows the Council for International Organizations of Medical Sciences (CIOMS) format.
789 The CIOMS format does not collect the same data elements as defined in the ISO ICSR/ICH E2B(R3)
790 electronic format. This can lead to differences in how CIOMS information is entered into ISO ICSR
791 format compliant databases.
792 Moreover, stakeholders use their own data collection forms for the primary source, which introduces
793 variability in the data collected. Some of the differences in forms between stakeholders will be for
794 product/disease/population-specific reasons, but not all the differences are down to this and there is
795 room for a degree of harmonisation.
796 Lastly, non-EEA-regulators use PV systems based on different standards, indicating variability between
797 systems and used definitions across the globe. [60] In the human domain, even those organisations
798 using ICH E2B format databases can have important differences, as some will use ICH E2B(R2)-format
799 databases and others ICH E2B(R3). In the veterinary domain, many non-EEA regulators do not use the
800 VICH GL35 format.
801 This all introduces practical challenges for NCAs and MAHs, in terms of inconsistent input and
802 inconsistent coding (format coherence). See section 5.1.3. for an example of such a challenge.
Page 27/49
第 28 页
803 To improve format coherence, it is recommended that template data collection forms are developed to
804 standardise the collected information (see Section 6.3. - Improvement of standard implementation). It
805 is also recommended that the data collection forms support population of the ISO ICSR/ICH E2B(R3)/
806 VICH GL35 electronic formats. Additionally, to improve coherence, it is recommended that webforms
807 for data collection have drop-downlists or suggested names for medicines to guide the primary source.
808 Furthermore, it is recommended to upgrade any human PV databases based on the E2B(R2) format to
809 be based on the E2B(R3) format.
810 5.3.3. Different interpretation of regulations & grey areas (semantic
811 coherence)
812 Semantic coherence is the issue that the same values, in a collection of ADR reports from multiple
813 sources, correspond to different real observations, due to different ways of interpreting such values
814 across sources.
815 Differences in submission rules between EEA and non-EEA can make it challenging to have a consistent
816 dataset within and across stakeholders. As an example from the human post-authorisation domain, an
817 ADR report concerning “a patient” that is valid in one region, would be invalid in the EEA. Where
818 reports are included in datasets in one region but excluded in another, this may lead to variations in
819 the cases available for aggregate review, trend analysis and signal detection activities. Similarly,
820 differing reporting requirements or validation criteria may influence the composition of datasets used
821 for signal detection, potentially affecting case counts, reporting frequencies and the interpretation of
822 observed reporting patterns. Such differences may also present challenges for duplicate detection
823 activities, when reports are captured, validated, or coded differently across regions.
824 In addition, there may be differences in interpretation of guidelines on ADR collection, management
825 and submission, which means that the same values of data correspond to different situations across
826 submitters and/or geographies. Examples are:
827 • When should an ADR be considered as new report and when should it be handled as a rechallenge
828 or follow-up?
829 • When should death reports with no reaction or information about the clinical circumstances be
830 submitted?
831 • When should cases be considered as duplicates?
832 • Which events described in a literature case should be considered to be reactions?
833 • Literature cases can be erroneously reported as multi-patient cases by MAH, although the
834 guidelines are clear that ADR reports should only contain individual patient information (individual
835 case safety reports).
836 To improve semantic coherence, it is recommended to reach improved alignment between regulators
837 at international levels within the boundaries of local legislation. It is recommended that industry
838 associations outline where these differences exist between important regulatory spheres, and to
839 actively engage with all relevant authorities to clarify the differences in interpretations to develop
840 harmonised interpretations for these specific scenarios.
841 5.3.4. Inconsistent coding (semantic coherence)
842 The quality of data entry can be measured by the degree to which the recorded data is representative
843 of the original information when retrieved and presented as data output. DQ is influenced by a certain
Page 28/49
第 29 页
844 degree of judgement from free text to a standardised code. Inconsistent coding of events can lead to
845 lack of coherence in the data.
846 This problem is most prominent for coding of the product name: the product can be identified with the
847 brand name (at the country level), but also as the active substance(s), and then as only the active
848 moiety as a salt or salts. There are myriad different dictionaries and implementations of rules
849 regarding the names of medicines in MAHs’ databases, and these are free-text fields in both EV and
850 EVV. Input systems offer guidance, but they are different, not enforced and multiple routes are
851 possible.
852 Moreover, for diagnosis codes, the same type of event for the same medical product can receive a
853 different MedDRA or VeDDRA code, for example application site pain or ear pain for a topical product
854 applied in the ear. This problem can be caused by different people responsible for coding; many
855 plausible codes in the dictionary; or increasing insights (after thorough assessment something that
856 starts with a headache, can become migraine). Moreover, lack of guidance for some specific fields can
857 lead to subjective judgement and thereby inconsistent coding. This is important to bear in mind when
858 reviewing ADR data for signal detection: there can be different expressions of a single
859 pathophysiological mechanism, which may be missed when filtering on clinical diagnosis. For literature
860 cases the same source documentation can be interpreted differently by stakeholders, which can lead to
861 inconsistent coding.
862 To ensure that the same value means the same thing throughout a dataset, it is recommended that
863 one standard coding terminology for product name/active substance is applied to all systems and
864 stakeholder databases (see Section 6.3. - Improvement of standard implementation). Moreover,
865 implementing thorough training for ADR processing and submitting organisations and establishing a
866 standardised coding manual (implementation rules) is recommended to reduce coding inconsistencies
867 and avoid misinterpretation of data (see Section 6.1. - Education).
868 5.4. Timeliness
869 Timeliness is defined as the availability of data at the right time for regulatory decision making. That,
870 in turn, entails that data are collected and made available within an acceptable timeframe. Timeliness
871 answers the question: are the data reflecting the intended reality at the point of time of its use? When
872 considering the “fit for purpose” definition of quality, timeliness covers how closely the data reflect the
873 intended reality, at the time in which it is used [1].
874 The problem of timeliness in ADRs is on four levels: the time it takes for a primary source to report an
875 ADR, the time it takes for an ADR to be submitted to EV/EVV and the time it takes for a validated
876 signal to be flagged as a signal of disproportionate reporting and, more notably, the time that it takes
877 to get regulatory action from the validated signal.
878 This can be affected by the need to procure missing information, the time needed to consult other data
879 sources, the time to assess weak causal assessment or the need to accumulate more cases to
880 compensate for DQ issues.
881 Timely submission is crucial for rapid identification of potential safety issues. The timelines required for
882 submission of ADRs to EV/EVV are laid down in the legislation, which outlines the requirements
883 governing the regulatory process. Table 1 shows an overview of the legislation and submission
884 requirements. In the human domain, EMA informs all organisations submitting ICSRs to EV about their
885 compliance with the regulatory timelines every month to help them identify areas for improvement
886 [61, 62].
Page 29/49
第 30 页
887 5.5. Relevance
888 In terms of DQ assessment, relevance is defined as the extent to which a dataset presents the data
889 elements useful to answer a given research question. The relevance DQ dimension answers the
890 question: does the dataset present the values (or data elements) that are needed to address a specific
891 question, using a specific method? When considering the “fit for purpose” definition of quality, the
892 relevance DQ dimension describes how the data cover the aspects of reality that are intended to be
893 measured [1].
894 In ADR collection, specifically in some solicited reporting scenarios, the data that is provided may not
895 be describing adverse drug reactions.
896 Spontaneous reports by a patient or HCP are, by definition, relevant and classed as suspected.
897 Solicited ADR reports originating from an organised data collection system require a causality
898 assessment (see ICH E2D(R1) [10], sections 2.8 and 3.2) and should only be submitted to a regulatory
899 authority if suspected to be causally related. Unsolicited reports originating from various sources such
900 as literature, or non-medical sources (lay press or digital platforms), where the MAH has obtained the
901 information, rather than it being reported to them, require a causality assessment for the events
902 reported in the source documents unless the author of the documents is clear in attributing causality.
903 As described in Section 5.6.2. , digital platform data in particular are generally difficult to analyse due
904 to their unstructured nature and lack of credibility [63].
905 ADRs collected from interventional clinical trials have specific challenges. All serious AEs should be
906 reported by the investigator to the Sponsor (unless risk-adapted safety data collection is applicable –
907 see ICH E1911 [12] & the Guideline Risk proportionate approaches in clinical trials. [22]); whereas only
908 unblinded Suspected Unexpected Serious Adverse Reactions should be submitted to EEA Member
909 States (NCAs & Ethics Committees). This requires 3 separate assessments by different actors:
910 Suspectedness (causality – assessed by investigator and also by the sponsor), Expectedness (assessed
911 by the sponsor) and Seriousness (assessed by the investigator and the sponsor).
912 • Suspected
913 The requirement to assess causality and only submit suspected unexpected serious ADRs
914 means that each adverse event identified during a clinical trial requires assessment of whether
915 the investigator thinks that the event was caused by the (often blinded) investigational
916 medicinal product that the patient received. In a blinded trial, it is vital to maintain the blinding
917 of the investigator, so they must be required to assess the causality without knowing what the
918 patient received. To prevent oversubmission while maintaining the blind, the reporting forms
919 should ask the investigator to assess causality as though the patient had received the
920 investigational medicinal product or the blinded placebo/comparator.
921 If the investigator decides that the event was not suspected, the Sponsor may perform their
922 own assessment and may decide that it was suspected. If either party (investigator/sponsor)
923 decides that the event was suspected, then it should be treated as suspected for the purposes
924 of SUSAR submission;
925 • Unexpected
926 The assessment as to whether the reaction is expected, or unexpected, should be performed
927 by the Sponsor with reference to the valid Reference Safety Information (RSI) at the time
928 the reaction occurred. For this, it is important that the RSI is clearly labelled and version
929 controlled, with a date range for the validity of each version. It is also important that the RSI
930 has clear wording and is sufficiently granular, with precise descriptions of the expected
11 This is not to say that risk-adapted safety data collection poses a challenge to ADR DQ – merely to acknowledge that
there are situations when not all serious AEs should be reported by the investigator to the sponsor
Page 30/49
第 31 页
931 reactions and symptoms, including the severity and seriousness of the reactions. The Sponsor,
932 and any organisations operating on their behalf for the purposes of ADR recording/submission,
933 should always have access to every version of the RSI and their staff need to be adequately
934 trained to use the correct version.
935 Sponsor’s staff should be sufficiently trained and experienced to understand that the same
936 medical concept can be expressed in different ways, e.g. “increased LFTs” and “increased
937 hepatic enzymes” should be understood to be equivalent terms. Additionally, staff need to be
938 aware that the expectedness can be dependent on the seriousness or frequency of a reaction.
939 For example, if the RSI lists ‘transient increases in hepatic enzymes’, but a patient experienced
940 such a significant increase in the value of their LFTs that they are hospitalised for a week, then
941 this should be considered unexpected. Equally, if an ADR is described as likely occurring rarely,
942 yet the Sponsor identifies a large number of occurrences of this ADR, then this should be
943 considered unexpected.
944 • Serious
945 There are 6 reasons why an event should be considered serious: Fatal, Life-threatening,
946 Disabling/incapacitating, Hospitalisation, Congenital anomaly, Other medically important
947 condition. These terms are defined in ICH E2A [8] and further in GVP Module VI [17], chapter
948 VI.A.1.6.
949 Seriousness criteria should be assessed with respect to the specific event/reaction and not to
950 the overall status/outcome of the patient. This can mean that even seemingly clear and binary
951 seriousness criteria (did the patient die? Was the patient hospitalised?) are context-specific. A
952 patient in a clinical trial for people with advanced lung cancer will have a very high chance of
953 mortality. If that patient experiences a SUSAR of jaundice, for which they are hospitalised, and
954 they subsequently die due to their lung cancer before they have recovered from the jaundice,
955 then that does not mean that the jaundice was fatal.
956 Those seriousness criteria which the investigator has deemed apply to the reported event(s)
957 should only be downgraded from serious to non-serious by the investigator (while maintaining
958 the blind, as applicable).
959 The challenges relating to Suspectedness and Seriousness are not unique to interventional clinical trials
960 and also apply to other domains: both apply to human post-marketing, and seriousness applies to the
961 veterinary domain.
962 5.6. Special considerations for specific ADR data sources
963 5.6.1. Considerations on patient support programmes for ADR reporting
964 The volume of ADR reports originating from patient support programmes (PSPs) and submitted to EV is
965 high, but the quality of the reports is highly variable. The quality of the reports ranges from low quality
966 (comparable to e.g. digital platforms) and lacking any causality assessment, to being comparable to
967 spontaneous reports from healthcare professionals [49]. Some reports from PSPs which should be
968 regarded as solicited and requiring a causality assessment, are instead erroneously handled as
969 spontaneous with implied causality, resulting in oversubmission and issues of accuracy (noise).
970 A possible cause for the variable DQ was the broad definition of patient support programmes leading to
971 different designs of these programmes [49]. The requirements for data collection from patient support
972 programmes differs across regulatory authorities. Although patient support programmes are not
973 always designed to probe safety, information collected through these programmes should be
974 considered solicited reporting [17].
Page 31/49
第 32 页
975 To reduce noise in data resulting from patient support programmes, ICH E2D(R1) [10] has been
976 published with sections focussing specifically on PSPs and other related forms of Organised Data
977 Collection Systems (ODCS). Prompt implementation of E2D(R1) and further harmonisation of
978 international guidelines, to facilitate alignment and to focus on collection of the relevant cases, would
979 be an area of improvement. Furthermore, a risk-based approach (e.g. patient support programmes
980 which will likely deliver valuable safety information) may be a solution; but this would require changes
981 to existing guidance and possibly legislation in some regions.
982 Note: This section is not applicable to veterinary ADR reports.
983 5.6.2. Considerations on social media for ADR reporting
984 Social media has been identified as one of the data sources that could provide real-time, user-
985 generated information about ADRs. It could play a complementary role as an early warning system for
986 potentially new ADR signals.
987 The first issue with social media as a source for ADR reports is that it provides a vast amount of data,
988 but these data are hard to analyse due to their unstructured nature and lack of credibility and could
989 therefore possibly negatively impact signal detection systems [63, 64]. This leads to reliability issues
990 (noise). Reactions/events described on social media might be unrelated, non-specific and written in
991 nonsensical terms such as “My head feels sloppy” or “I’m off my feet”. Moreover, consumers, animal
992 owners or malicious actors could easily write fake reports and may duplicate a report via several
993 platforms. This process is very hard to verify in terms of who is the reporter.
994 A second issue where social media impacts ADR reporting, is stimulated reporting. When high media
995 interest is involved, the median reporting rate might visibly increase [65, 66, 67]. An example is heavy
996 menstrual bleeding after COVID-19 vaccination. At social media forums, women reported their
997 experiences of a menstrual cycle disturbance and a potential link with recent COVID-19 vaccination.
998 During the continuous safety review of COVID-19 vaccines, PRAC performed an analysis in EV of
999 spontaneous reports of heavy menstrual bleeding reported by consumers. In some European Union
1000 countries with high reporting rates (e.g. the Netherlands, France), the consumer reporting appeared to
1001 be influenced by communications from media and/or health authorities, with high volume of reports
1002 shortly after publications. [65]
1003 Patient/animal owner reporting directly to regulatory authorities and MAHs is of high value for
1004 regulatory safety reviews, albeit the reporting behaviours are not free of the influence of the media.
1005 Regardless, they provide a better understanding of the context of safety concerns of the public.
1006 Regulatory organisations have produced guidelines on how unsolicited reports of ADRs collected from
1007 social media should be handled for PV purposes in the human domain, and MAHs hosting social media
1008 sites/pages or other digital platforms have additional responsibilities regarding the collection of ADR
1009 reports [10, 17]. For veterinary medicinal products, guidance on handling AE reports collected from the
1010 internet or from digital media is provided in the VGVP Module: Collection and recording of suspected
1011 adverse events for veterinary medicinal products [26].
1012 Moreover, the research project WEB-RADR [68] has explored the value of social media (i.e.,
1013 information exchanged through the internet, typically via online social networks) for identifying ADRs
1014 as well as for safety signal detection [69] and found limited value for formal, automated signal
1015 detection of new or unknown drug safety issues. This is due to the high volume of noise, lack of clinical
1016 context, and difficulty in verifying individual reports. Many patients and clinicians have discussed their
1017 positive and negative experiences of medications on social media, creating a source of publicly
1018 available information that has the potential to provide insights into medicinal product safety concerns.
1019 This research produced several recommendations that can be considered:
Page 32/49
第 33 页
1020 • Social media platforms as Facebook and Twitter (currently known as X) are not recommended for
1021 broad statistical signal detection.
1022 • Social media channels may provide a useful adjunct to PV activities in specific niche areas such as
1023 exposure during pregnancy and abuse/misuse of medicines.
1024 Future enhancement of ADR recognition algorithms may broaden the scope and utility of social media
1025 over time.
1026 5.6.3. E-commerce
1027 E-commerce of medicines (the buying and selling of medicines over the internet), can lead to similar
1028 issues as digital platform reports: noise and excess reporting. Specific causes for noise in e-commerce
1029 for the MAH/NCA are that the medicine may not be authorised in the country where it is reported,
1030 making identification of the actual product particularly difficult. Furthermore, e-commerce can be a
1031 significant source of falsified or counterfeit medicines [70, 71], which can further complicate the
1032 analysis of signals.
1033 5.6.4. Blinded clinical trials
1034 Special consideration is needed for blinded clinical trials, where it is vital to maintain trial integrity
1035 while ensuring adequate safety reporting. In accordance with the Clinical Trials Regulation [23], Annex
1036 III Safety Reporting, SUSARs shall be accessible only to persons who need to be involved in the safety
1037 reporting to the Agency, to Data Safety Monitoring Boards (’DSMB’), or to persons performing ongoing
1038 safety evaluations during the clinical trial. By contrast, the blind shall be maintained for other persons
1039 responsible for the ongoing conduct of the clinical trial (such as the management, monitors,
1040 investigators) and those persons responsible for data analysis and interpretation of results at the
1041 conclusion of the clinical trial.
1042 It is standard practice of many sponsors to return a copy of the SUSAR to the investigator. This copy of
1043 the SUSAR must be blinded, so as not to risk unblinding the investigator or their personnel. Not only
1044 must the structured drug and substance information be blinded, but the case narrative and all free-text
1045 fields must also be blinded.
1046
Page 33/49
第 34 页
1047 6. Recommendations to improve DQ in ADR reporting
1048 6.1. Education
1049 Education is crucial for improving the DQ of ADR reports. All stakeholders in the ADR pipeline should
1050 ideally be educated about basic concepts such as what is the most important and relevant information
1051 to provide.
1052 6.1.1. Training of primary source
1053 The ADR reporting process requires technical skills and knowledge by HCPs, including knowledge of
1054 ADR signs and symptoms, risk factors, recognition of underlying comorbidities and concurrent
1055 conditions mimicking an ADR. Other skills are familiarity with pharmacological profiles of medications
1056 and drug interactions and the ability to construct a relevant medication timeline [39].
1057 Ideally, HCPs would receive training on ADR recognition, reporting procedures, and the significance of
1058 reporting. This should preferably start at student level but more importantly should be continued at
1059 hospitals, veterinary surgeries and all care centres. This would improve coverage and completeness
1060 and the overall quality of ADR reports.
1061 The ability of HCPs to understand and provide sufficient details in the reported verbatim can have a
1062 significant impact on ADR reporting. At database level, the narrative is coded to medical terms with the
1063 same medical vocabulary as they need to be mapped to the back-end system of the (local) regulator.
1064 However, HCPs from different countries may use different medical terminology in ADR reporting [59,
1065 32, 72]. For instance, a skin rash on a leg, specifies the part of the body but could not be specific
1066 enough for a regulator. The lack of harmonized medical terms and definitions across regions, combined
1067 with translation issues, may lead to variations in the standardized terms entered into safety databases,
1068 resulting in semantic imprecision and potentially differing clinical assessments and regulatory
1069 decisions.
1070 In spontaneous reporting, patients, animal owners and HCPs can be guided on what information should
1071 be provided to report an ADR. This can be supported through the use of standardized electronic
1072 reporting forms with integrated local dictionaries helping for example on the selection of medicinal
1073 products and active substances. Paper forms and the operators of medical information lines can also
1074 help to guide reporting of ADRs. Public campaigns about awareness when to report an ADR and the
1075 appropriate reporting routes can be useful as well, although this is highly dependent on healthcare
1076 systems and local practises.
1077 In clinical trials and some other solicited forms of reporting, investigators and other HCPs involved in
1078 the trial should receive thorough training on causality assessments and on the need to provide all
1079 available information relevant to the ADR, such as relevant test data and medical history, when first
1080 reporting the AE to the sponsor.
1081 6.1.2. Training of data processors, managers and submitters to EV or EVV
1082 (MAH, NCA, sponsors)
1083 The entire process of data generation and coding (SOPs, manuals, trainings) should be documented,
1084 true, verifiable, and versioned. These aspects should therefore be included in training for submitters to
1085 EV, which improves DQ issues in determinants for reliability, extensiveness, coherence and timeliness
1086 [1]. After processing, ADRs are thoroughly reviewed by relevant organisations (MAH, NCA, Sponsor)
1087 which received the original information from the primary source, based on expectedness, causality,
1088 inconsistencies, and corrected as soon as possible. To make the best use of the received information,
Page 34/49
第 35 页
1089 these employees need medical/veterinary knowledge, experience and, for collecting additional relevant
1090 information, good communication skills. The information initially available is often insufficient, and
1091 efforts to collect additional follow-up information for spontaneous or literature reports mostly fail. An
1092 important point of improvement for which thorough training is recommended, is inconsistent coding
1093 and to ensure that the same recorded value has the same meaning throughout a dataset. Establishing
1094 a standardised coding manual is recommended to reduce inconsistencies in the coding of
1095 medical/veterinary terms, thereby avoiding misinterpretation of data. Specifically, data
1096 entry/management/submission staff should attend training sessions and regularly refer to the coding
1097 manual and should have access to the training materials outside of the training. As part of the regular
1098 audits as described in GVP Module IV [68] analysts are encouraged to conduct periodic audits of coded
1099 data to identify and rectify inconsistencies.
1100 Furthermore, education on submission rules is pivotal to improve accuracy and to minimise noise. It is
1101 recommended to promote good use of submission rules throughout the network.
1102 6.1.3. Training of data users and PV-assessors
1103 As a last step, data users (e.g. regulators, signal detection staff) should be trained to report any
1104 identified incorrect submission either to an internal quality control office or back to the sender
1105 organisation for continuous improvement of DQ.
1106 6.1.4. Principles of good ADR submission
1107 What makes a good ADR report? During signal validation, in the human domain ADRs will be evaluated
1108 against the criteria detailed in GVP Module IX [18] chapter IX.B.3. In the veterinary domain AEs will be
1109 evaluated against the criteria detailed in the Guideline on veterinary good pharmacovigilance practices
1110 (VGVP) Module: Signal Management [30]. Consequently, staff processing ADR reports should receive
1111 adequate training to develop an understanding of the information required by pharmacovigilance
1112 assessors to perform signal evaluation based on the data recorded in the database; in this regard the
1113 following principles [73] should be considered:
1114 • Submit all valid ADRs to EV/EVV. Record all ADRs and all reports of use in special situations (e.g.
1115 pregnancy or breastfeeding), and ensure that only those ADRs which should be reported are
1116 reported;
1117 • Provide accurate non-mandatory12 details: include patient demographics and relevant medical
1118 history, concomitant medication, ADR description including time-to-onset and what happened after
1119 drug withdrawal and re-administration (dechallenge, rechallenge);
1120 • Submit promptly: timely submission enables quick identification of potential safety issues (see
1121 reporting timelines in Table 1);
1122 • Causality assessment for solicited reports: causality is related to the concept of change:
1123 “something new or something different”, which typically excludes symptoms that are part of a
1124 chronic illness and pays particular attention to the time-to-onset. Moreover, Causality assessment
1125 can be performed using several methods, such as Bradford-Hill principles, Agence nationale de
1126 sécurité du médicament et des produits de santé (ANSM) imputability, Naranjo algorithm, the
1127 WHO-UMC system or CIOMS criteria or, in the veterinary domain, the ABON system. In general, an
1128 important criterium of causality is: “Time is key”. Given the large number of causality assessment
12 Although such information is not mandatory for an ADR report to be valid, if the MAH/NCA/Sponsor has received it from
the primary source, then it should always be correctly structured in the ADR report
Page 35/49
第 36 页
1129 methods, it is recommended that global regulators attempt to harmonise assessment
1130 methodologies for each domain (post-authorisation, clinical trials, veterinary).
1131 For clinical trials, the investigator has more training and obligations than reporters of spontaneous
1132 reports and the data is generated in a controlled environment. Therefore, more information than the
1133 bare minimum should be available and reported by the investigator. In particular, investigators should
1134 report (and sponsors should submit) clinically meaningful data for each SUSAR, including time-to-
1135 onset, action taken with the suspect drug, the outcome of the reaction (at the time of reporting and
1136 follow-up) and relevant medical history and test data. If this information is not immediately available,
1137 follow-up should be expedited to provide it and to update the ADR report as necessary (which is true
1138 for all ADR reports across all domains).
1139 6.2. Integration of clinical pharmacology in ADR processes
1140 Identification, reporting, analysis, management, submission, and public awareness of ADRs should
1141 become a more integrated system from clinical trials to post-authorisation. To integrate the system,
1142 clinical pharmacology should be at the base of ADR reporting.
1143 It starts with profiling of all potentially relevant pharmacological activity, starting in the early phases of
1144 drug development. Most ADRs can be predicted from the pharmacological and preclinical profile: one
1145 can predict in which dose ranges they might occur, what risk factors increase their occurrence,
1146 interactions or organ failure. Therefore, the pharmacology of the medicine, potential ADRs, and
1147 characteristics of a patient population should be acknowledged when writing the common technical
1148 document and the development risk management plan, that later on inform the risk management plan.
1149 This could improve identification of (relevant) ADRs, and the quality of the reported ADRs in terms of
1150 reliability, causality assessment and coherence. This is of high importance, because when the safety
1151 analysis is misguided, e.g. by incorrect specification of the patient population in a clinical trial, this may
1152 lead to unnecessary hazard to patients and risks that a good medicine (or even a drug class) is
1153 discontinued in the drug development process or post-marketing.
1154 Furthermore, clinical pharmacology, together with pharmaco-epidemiology and clinical judgement,
1155 plays an important role in signal detection and signal assessment. Statistical and quantitative methods
1156 provide valuable tools for identifying potential safety signals; however, interpretation of these findings
1157 is strengthened by an understanding of the pharmacology of the medicine, biological plausibility, and
1158 the overall clinical context. The integration of these complementary approaches can support the timely
1159 identification and evaluation of emerging safety concerns. A challenge in signal detection is to
1160 recognise causality. This decision can be hampered by lack of knowledge on the pharmacology of the
1161 medicine. If a signal is only raised after a number of events, mainly based on pharmaco-epidemiology
1162 this may lead to a delay in recognition and consequent risk of casualties.
1163 Ultimately, integration of all ADR processes as described above could lead to a more Bayesian
1164 approach to PV by predicting ADRs and mitigating risks, informing (development) risk management
1165 plans and product safety information.
1166 6.3. Improvement of standard implementation
1167 Standardisation is key to improve DQ of ADR reports. This includes standardisation of the process, but
1168 also of the data itself (e.g. data fields and terminology).
1169 Data collection forms
1170 It is recommended that standardised templates for all types of data collection form (webforms, paper
1171 forms and forms for medical information phone lines) is developed for data collection at the primary
Page 36/49
第 37 页
1172 source, which should be used by all stakeholders (MAHs, NCAs, and sponsors of clinical trials). It is
1173 recommended that, as much as possible, these data collection forms support the ISO ICSR/ICH
1174 E2B(R3)/VICH GL35 electronic format. These standardised forms should be simple and should guide
1175 the primary source to report the most important and useful information for a high-quality ADR report,
1176 in particular clinical data in order to help the scientific evaluation. Overcomplete reports and extremely
1177 lengthy narrative are certainly not necessary to provide a good ADR report. In the human domain, the
1178 Strengthening Collaboration for Operating Pharmacovigilance in Europe (SCOPE) Work Package 4
1179 Paper ADR Reporting Forms template [74] provides a basis for such forms.
1180 Data collection tools should also be developed to allow/encourage the direct electronic reporting of
1181 ADRs from hospital, veterinary surgery and other medical/veterinary health records.
1182 For reports from organised data collection systems, the template should include questions that require
1183 the reporter to provide a causality assessment. For clinical trials reporting forms, investigators should
1184 be encouraged to document the reasoning supporting the causality assessment, including key elements
1185 such as temporal relationship, alternative explanations, dechallenge/rechallenge information and
1186 relevant medical history
1187 Moreover, webforms should have drop-down lists or suggested names for medicines, medical terms
1188 and other data elements that have controlled vocabularies when transmitting the report as an ICSR,
1189 while still allowing reporters to enter terms not on the suggested lists. This would increase coherence
1190 and decrease incorrect coding. It would also improve precision at the time of reporting. A clear
1191 explanation and/or automated notifiers will guide the primary source (patients, animal owners and
1192 HCPs) to provide the critical data elements resulting in higher DQ.
1193 Product/substance coding terminologies
1194 There are many different product/substance coding terminologies and implementations of rules
1195 regarding product names in stakeholder databases, and these are free-text fields in EV & EVV. To
1196 improve semantic coherence, it is recommended that one standard coding terminology for each
1197 domain (human/veterinary) for product and substance names is applied to all systems and stakeholder
1198 databases. Moreover, training for data entry/management/submission staff to EV & EVV, and one
1199 standardised coding manual for each domain (implementation rules), is recommended to reduce
1200 coding inconsistencies and improve coherence.
1201 Medical and veterinary terminology coding
1202 Correct, accurate, consistent coding of medical and veterinary terms is essential to allow for consistent
1203 analysis of potential signals. MAHs, Sponsors and NCAs should all ensure that MedDRA and VEDDRA
1204 coding principles [75, 76] are implemented in their systems and are followed by data entry and
1205 medical/veterinary review staff.
1206 Harmonised processes
1207 Lastly, it is recommended that global regulators adopt harmonised approaches to sharing data with
1208 MAHs to try and maintain the coherence as much as possible.
1209 6.4. Early implementation of data validation processes
1210 Automated checks to improve DQ should be integrated in PV databases, e.g. age should not be <0
1211 years, and a check for the number of completed fields. Ideally, automated data validation should be
1212 performed as early as possible, upon data collection. These could be implemented through e.g.
1213 automated notifications upon data entry in the webform for data collection, to reduce errors, stimulate
1214 completeness of the minimum required fields (Section 4.4. ) and encourage collection of other data
Page 37/49
第 38 页
1215 elements for scientific evaluation. Another example would be the use of code lists for dates, which
1216 could improve format coherence.
1217 During the data management and processing step of the data life cycle, it is recommended that data
1218 validation is performed for structured fields. This could include metrics of completeness. It is
1219 recommended that MAHs develop a metric that is specific for a certain product or reaction, because
1220 key-data elements may differ for each product.
1221 6.5. Implementation of feedback mechanisms to allow for data corrections.
1222 Data lineage (or traceability) describes the history of data, detailing its origins, the sources it came
1223 from and the transformations it underwent before reaching its current state [1]. The pipeline for safety
1224 data collection and ADR reporting can be extensive with many stakeholders involved in the same case.
1225 Traceability and data ownership as the ADR travels through the pipeline could improve reliability and
1226 could reduce duplication.
1227 Good feedback across the ADR pipeline in all steps of the data lifecycle will improve DQ. These
1228 activities are recurring and should be monitored on a continuous basis. Therefore, information on the
1229 organisation that submitted each ADR report is mandatory in EV and EVV to provide the possibility for
1230 contacting sender organisations regarding the quality of the data in an ICSR. Furthermore, information
1231 on the primary source is stored in the database of the organisation contacted by them so that they can
1232 be contacted to request follow-up information. It is important that the original receiver of an ADR
1233 report (MAHs/sponsors/regulators) contacts the primary source when they find a problem with the
1234 data in an ADR report. Therefore, it is important to implement oversight activities such as defined and
1235 quantifiable metrics to enable MAHs, Sponsors and Regulators to objectively assess if data is fit for
1236 purpose. In addition, regular medical review discussions or regular compliance monitoring meetings
1237 should be held in each organisation across the ADR pipeline. This is especially important when certain
1238 data management or data submission activities are outsourced to third parties, or when collaborating
1239 with business partners. With these mechanisms in place, it is easier to reach out to other stakeholders
1240 and come to an agreement to make data corrections.
1241 Another mechanism for data correction could potentially be the use of Artificial Intelligence (AI) to
1242 aid human data management activities. Technology and automation become more and more common
1243 as seen by the rise of AI for PV activities such as data entry and scoring of disease severity [77]. These
1244 PV applications should allow a flexible approach to AI modelling and deployment as these increase in
1245 terms of precision over time due to incremental learning and enhanced models for classification and
1246 severity scoring. These AI tools may be helpful for initial validation by automated data corrections
1247 regarding data entry, adverse reaction coding, diagnosis checking and scoring of disease
1248 severity/seriousness; but these should always be suggestions and a humans should always be taking
1249 the final decisions. It should be noted that the MAH/NCA/Sponsor remains responsible to validate,
1250 monitor and document model performance and include regular reviews of AI operations in the PV
1251 system to mitigate risks related to algorithms and models used. The Report of CIOMS Working Group
1252 XIV “Artificial Intelligence in Pharmacovigilance” contains many recommendations regarding the use of
1253 AI in human PV [78], and these should be taken into account before using AI in PV (and these are also
1254 broadly applicable to the use of AI in veterinary PV). Another important consideration before using AI
1255 in PV, or any health data, is ensuring compliance with the AI Act (Regulation (EU) 2024/1689).
1256 6.6. Maintain immutability of data at the point of transmission
1257 One of the issues with ADR reporting can be the apparent lack of immutability. Data immutability
1258 refers to the principle that data is never deleted or modified. Once data is recorded (e.g. once entered
1259 into a database), it can only be supplemented with additional information that may invalidate or
Page 38/49
第 39 页
1260 replace the previous data, but it cannot be removed. In other words, once data is entered into a
1261 system (and potentially relied upon by other data or actions), it cannot be altered without explicitly
1262 noting a new state of the information while preserving the knowledge of the previous state. Data used
1263 for regulatory purposes should be immutable. This does not mean that evidence cannot be updated;
1264 but any update should be treated as a separate, though related dataset. This principle is fundamental
1265 to many frameworks, such as ALCOA++ [79] and FAIR [80] and underpins the EMA guideline on
1266 computerised systems and electronic data in clinical trials [79]. A general rule is that aspects of DQ
1267 should be addressed as early as possible in the data collection and generation process. The further
1268 data travels from the original context, the harder it becomes to correct issues [1]. In ADR collection,
1269 management and reporting, this should be through the creation of different versions of a case, with a
1270 full audit trail for each version and the case as a whole. If any versions require transmission to another
1271 party, they should be transmitted, stored and managed as separate ICSRs under the same case.
1272 Applying the concept of immutability to ADR data could improve reliability. It is critical that in all
1273 versions of a case, the case number (ICH E2B(R3) C.1.8.1/VICH GL35 A.4.1) remains the same.
1274 Ideally in the case narrative, there should be an amendment version date to match the last received
1275 date recorded in the case (or the date the information was amended if a correction is made with no
1276 new information). Follow-up information should be clearly identifiable, and from the perspective of
1277 signal detection it would be ideal if it were integrated in the narrative by correcting the information
1278 entered in the first version of the narrative that is not applicable anymore. All the changes should also,
1279 where applicable, be reflected in the appropriate structured E2B(R3) data elements. At the bottom of
1280 the narrative, there would ideally be a note highlighting the changes, so assessors can easily see the
1281 changes. However, there is no clear recommendation on how to include follow-up information in case
1282 narratives in the existing guidance. It is recommended to include such recommendations in future
1283 guideline updates. Because of the challenges with obtaining follow-up information, it is of importance
1284 to collect as much relevant information as possible upon first receipt of the ADR. For stakeholders, it is
1285 critical to maintain an audit trail to be able to see the different versions of the case. For the data
1286 procurement step, it is also important to pay attention to data immutability as removal of identifiable
1287 information on patient/reporter might impact the information in the case narrative.
1288 6.7. Routine duplicate detection and elimination by databases
1289 The detection of duplicates remains a challenge (Section 5.3.1. ) despite tremendous efforts across the
1290 ADR pipeline [81]. It is clear that duplicate reports should be removed from PV databases before
1291 statistical analyses are conducted; however, the applicable submission rules cannot completely avoid
1292 duplicate submission. For example, in the human domain GVP Module VI states that when one party is
1293 made aware that the primary source(s) may also have reported the ADR to another concerned party,
1294 the report should still be considered as a valid ADR report [17].
1295 Multiple approaches are in place for duplicate detection such as probabilities, comparing the likelihood
1296 of specific combinations of country, medicines, ADRs, sexes, ages, species etc. occurring on a given
1297 pair of reports based on background reporting rates. Databases should be routinely screened to detect
1298 and eliminate duplicate reports by collaboration of stakeholders across the ADR pipeline (NCA, MAH,
1299 sponsors, EMA and WHO-UMC). The detection and handling of duplicates by these stakeholders form
1300 an important element of good case management and in the human domain collaboration between
1301 MAHs, NCAs in Member States and EMA in the detection of duplicates in EV is mandated by Directive
1302 2001/83/EC [33]. In the veterinary domain, Commission Implementing Regulation (EU) 2021/1281
1303 [32], Article 4(f) mandates that MAHs are responsible for the management of duplicates in EVV.
1304 The EV system uses a combination of worldwide unique (case) identification number and sender
1305 identification for each ADR report and has a built-in deduplication mechanism to identify and eliminate
Page 39/49
第 40 页
1306 duplicate reports by creating master cases to merge the information from duplicate cases [81]. The
1307 EVV system has similar deduplication mechanisms in place to ensure that duplicate cases are removed
1308 from the pharmacovigilance analyses.
1309 6.8. Effect of recommendations on DQ per DQ dimension
1310 These seven recommendations to improve DQ in ADR collection and reporting have been categorised
1311 per step in the data lifecycle in Table 2. The stakeholders involved in each data lifecycle step are
1312 indicated in each column heading. The recommendation numbers correspond with the numbers in
1313 Section 6.
Page 40/49
第 41 页
1314 Table 2: Implementation of recommendations organised along the ADR data lifecycle.
1315
Definition of data Data collection or generation Data management and Data submission (EMA) Data procurement and Delivery for
requirements (NCA, (primary source: patients, animal processing (NCA, MAH, aggregation (EMA, consumption (EMA,
MAH, sponsor of clinical owners, HCPs, lawyers, sponsor of clinical trials) NCA, MAH, WHO-UMC) WHO-UMC, MAH, NCA)
trials) investigators)
Education13 (section Training of primary source Training of data managers Training of PV-assessors
6.1) (investigators)
Context-specific explanations in data
collection forms
Integrating clinical Knowledge of pharmacology for Knowledge of pharmacology Knowledge of
pharmacology (section improved causality assessment for MedDRA & VedDRA coding pharmacology for
6.2) (investigators and sponsors) and causality assessment improved signal detection
Improved standard Design standardised data Standard coding terminology
implementation collection forms for drug names
(section 6.3)
Data validation Design validation checks Checks during data entry Data validation after collection
processes (section e.g. age ≤150y by metrics of completeness
6.4)
Implementation of Gather high DQ from follow-up Feedback among stakeholders Feedback among Return cases to the Feedback among
feedback mechanisms information at the primary source for data corrections stakeholders for data sender for further follow- stakeholders e.g. MAH-
(section 6.5) corrections and fitness for up MAH
purpose of data
Maintain immutability Allow immutability for all Investigators, immutability when Apply immutability and version QC for removal of
(section 6.6) data elements in a version providing follow-up information management identifiable information
and versioning within a
case
Update guidelines
regarding the
incorporation of follow-up
information into case
narratives
Routine duplicate Perform routine duplicate Perform routine duplicate Perform duplicate detection
detection (section 6.7) detection in own databases detection in EV after downloading cases
13 Education impacts all activities in the ADR pipeline, but most impact is expected in the highlighted steps in the data lifecycle.
第 42 页
1316 7. Concluding remarks
1317 This chapter is an extension of the EMRN DQF for medicines regulation, specifically on human ADR and
1318 veterinary AE reporting2. The DQF application to ADR data provides a background on the impact of DQ
1319 for ADR reporting. It provides insights into the larger ADR ecosystem, the stakeholders and their role
1320 in ADR collection, management and submission. It further outlines the main challenges for ADR
1321 collection, management and submission organised by the five DQ dimensions and provides
1322 recommendations for the characterisation of systems and processes underpinning data and for all
1323 stakeholders in the ADR ecosystems. These parts provide actionable and focused recommendations for
1324 assessing DQ of ADR reporting with the goal of improving the usefulness of ADR reporting to support
1325 regulatory decision making.
1326
42
第 43 页
1327 8. References
1. Data Quality Framework for EU medicines regulation. Available from:
https://www.ema.europa.eu/en/documents/regulatory-procedural-guideline/data-quality-
framework-eu-medicines-regulation_en.pdf. 2022.
2. European Medicines Agency, Multi-stakeholder workshop on Data Quality Framework for Adverse
Drug Reaction reporting. 2024.
3. VICH GL30: Pharmacovigilance of Veterinary Medicinal Products: Controlled list of terms
4. ICH E2B Implementation Working Group Implementation Guide for the Electronic Transmission
of ICSRs: E2B(R3) Data Elements and Message Specification, Current Step 4 version 5.03. 2025.
5. ICH Harmonised Tripartite Guideline: periodic benefit-risk evaluation report (PBRER) E2C(R2).
Available from https://database.ich.org/sites/default/files/E2C_R2_Guideline.pdf. Accessed 2
July 2026.
6. Guideline on veterinary good pharmacovigilance practices (VGVP) - Module: Signal Management
(EMA/522332/2021). Available from https://www.ema.europa.eu/en/documents/regulatory-
procedural-guideline/guideline-veterinary-good-pharmacovigilance-practices-vgvp-module-
signal-management_en.pdf. Accessed 30 June 2026.
7. European Medicines Agency policy on access to EudraVigilance data for medicinal products for
human use (EudraVigilance Access Policy), EMA/759287/2009 Revision 5. 2025.
8. ICH Harmonised Tripartite Guideline, Clinical Safety Data Management: Definitions and
Standards for Expedited Reporting E2A, Current Step 4 version, dated 27 October 1994.
9. EudraLex Volume 10: guidelines and Q&As related to Regulation 536/2014:
https://health.ec.europa.eu/medicinal-products/eudralex/eudralex-volume-10_en. Accessed 30
June 2026.
10. ICH E2D(R1) Guideline on post-approval safety data: definitions and standards for management
and reporting of individual case safety reports. 2026.
11. Guideline on veterinary good pharmacovigilance practices (VGVP) Module: Collection and
recording of suspected adverse events for veterinary medicinal products
12. The rules governing medicinal products in the European Union VOLUME 10 - Guidance
documents applying to clinical trials: Clinical Trials Regulation (EU) NO 536/2014 questions &
answers - amended and endorsed by the Clinical Trials Coordination and Advisory Group. Version
7.1. Available from: https://health.ec.europa.eu/document/download/bd165522-8acf-433a-
9ab1-d7dceae58112_en?filename=regulation5362014_qa_en.pdf. 2025.
13. Davies, E., et al., Recording of suspected adverse drug reaction reporting in veterinary free-text
clinical narratives. J Small Anim Pract 65: p. 361-7. https://doi.org/10.1111/jsap.13721
14. Designated Medical Event (DME) List. Available from
https://www.ema.europa.eu/en/documents/other/designated-medical-event-dme-list_en.xlsx.
Accessed 30 June 2026.
15. ICH Harmonised Tripartite guideline: Development Safety Update Report E2F. Available from
https://database.ich.org/sites/default/files/E2F_Guideline.pdf. Accessed 2 July 2026.
16. Detailed guide regarding the EudraVigilance data management activities by the European
Medicines Agency. 2020.
17. EMA - Guideline on good pharmacovigilance practices (GVP) - Module VI – Collection,
management and submission of reports of suspected adverse reactions to medicinal products
(Rev 2). 2017.
18. Guideline on good pharmacovigilance practices (GVP) Module IX – Signal management,
EMA/827661/2011 Rev 1. 2017.
19. European Medicines Agency, Good practice guide on recording, coding, reporting and assessment
of medication errors. 2015.
20. EU Individual Case Safety Report (ICSR) Implementation Guide, EMA/51938/2013 Rev 2. 2021.
Available from https://www.ema.europa.eu/en/documents/regulatory-procedural-
guideline/european-union-individual-case-safety-report-icsr-implementation-guide_en.pdf. Last
accessed 16 March 2026.
21. ICH E6 (R3) Guideline for good clinical practice (GCP). Accessed 27 August 2026.
22. Veterinary Union Pharmacovigilance Database – Best Practice Guide
23. Regulation (EU) No 536/2014 of the European Parliament and of the Council of 16 April 2014 on
clinical trials on medicinal products for human use, and repealing Directive 2001/20/EC Text with
EEA relevance. 2014.
24. VICH GL42 (R1): Pharmacovigilance of Veterinary Medicinal Products: Data elements for
submission of adverse event reports (AERs).
Page 43/49
第 44 页
25. EudraVigilance Access Policy for Medicines for Veterinary Use, EMA/113700/2008-Rev.2.1.
Revision 2.1 2022. Accessed 13 March 2026.
26. VICH GL24: Pharmacovigilance of Veterinary Medicinal Products: Management of adverse event
reports (AERs).
27. VICH GL30: Pharmacovigilance of Veterinary Medicinal Products: Controlled list of terms.
28. Guidance notes on the use of VeDDRA terminology for reporting suspected adverse events in
animals and humans. 2025. https://www.ema.europa.eu/en/documents/regulatory-procedural-
guideline/guidance-notes-use-veterinary-dictionary-drug-regulatory-activities-veddra-
terminology-reporting-suspected-adverse-events-animals-humans-rev17_en.pdf
29. EU VICH adverse event report implementation guide. 2021.
https://www.ema.europa.eu/en/documents/scientific-guideline/eu-vich-adverse-event-report-
implementation-guide_en.pdf. Accessed 16 March 2026
30. Screening for adverse reactions in EudraVigilance. EMA/849944/2016. Available from:
https://www.ema.europa.eu/en/documents/other/screening-adverse-reactions-
eudravigilance_en.pdf. Accessed 13 March 2026.
31. Regulation (EU) 2019/6 of the European Parliament and of the Council as regards good
pharmacovigilance practice and on the format, content and summary of the pharmacovigilance
system master file for veterinary medicinal products. Available from: https://eur-
lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:02019R0006-20220128. 2022.
32. Commission Implementing Regulation (EU) 2021/1281 of 2 August 2021. Available from:
https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32021R1281.
33. VICH Guidelines for veterinary medicinal products GLS 24,29,35,42
https://vichsec.org/en/guidelines/pharmacovigilance/vich-gls-24-29-35-42.html.
34. VICH GL35 (R1): Pharmacovigilance of Veterinary Medicinal Products: Electronic standards for
transfer of data.
35. Directive 2001/83/EC of the European Parliament and of the Council of 6 November 2001 on the
Community code relating to medicinal products for human use. 2001.
36. Regulation (EC) No 726/2004 of the European Parliament and of the Council of 31 March 2004
laying down Community procedures for the authorisation and supervision of medicinal products
for human and veterinary use and establishing a European Medicines Agency. 2004.
37. Commission Implementing Regulation (EU) No 520/2012 of 19 June 2012 on the performance of
pharmacovigilance activities provided for in Regulation (EC) No 726/2004 of the European
Parliament and of the Council and Directive 2001/83/EC of the European Parliament and of the
Council. 2012.
38. Risk proportionate approaches in clinical trials. Recommendations of the expert group on clinical
trials for the implementation of Regulation (EU) No 536/2014 on clinical trials on medicinal
products for human use. 25 April 2017. Available from:
https://health.ec.europa.eu/document/download/be80ee91-f7b8-4100-be72-
42efd8362d71_en?filename=2017_04_25_risk_proportionate_approaches_in_ct.pdf. Accessed 1
July 2026.
39. European Medicines Agency. EudraVigilance - European database of suspected adverse drug
reaction reports. Last accessed August 2024. Available from:
https://www.adrreports.eu/en/data_source.html.
40. MedDRA. Medical dictionary for regulatory activities. www.meddra.org/how-to-use/support-
documentation. Last accessed August 2024.
41. European Medicines Agency, EudraVigilance Veterinary,
https://www.ema.europa.eu/en/veterinary-regulatory-overview/post-authorisation-veterinary-
medicines, last accessed September 2024.
42. Christ, P., et al, Vigi4Eudra-score: Evaluation of the completeness of spontaneous adverse drug
reaction reports in EudraVigilance. PLoS One 21(2): e0343694.
https://doi.org/10.1371/journal.pone.0343694
43. Important medical event terms list (MedDRA version 27.0).
https://view.officeapps.live.com/op/view.aspx?src=https%3A%2F%2Fwww.ema.europa.eu%2Fe
n%2Fdocuments%2Fother%2F27-0_ime_list.xlsx.
44. Guideline on veterinary good pharmacovigilance practices (VGVP) Module: Signal Management.
Available from https://www.ema.europa.eu/en/documents/regulatory-procedural-
guideline/guideline-veterinary-good-pharmacovigilance-practices-vgvp-module-signal-
management_en.pdf. Accessed 17 September 2026.
45. EMA pharmacovigilance system manual version 1.3. EMA/623550/2013.
https://www.ema.europa.eu/en/documents/other/european-medicines-agency-
pharmacovigilance-system-manual_en.pdf.
Page 44/49
第 45 页
46. EMA: List of signals discussed at PRAC since September 2012.
https://www.ema.europa.eu/en/human-regulatory-overview/post-
authorisation/pharmacovigilance-post-authorisation/signal-management/prac-recommendations-
safety-signals#list-of-safety-signals-discussed-since-september-2012-7892.
47. WEB-RADR. Available from https://web-radr.eu/. Accessed 14/06/2026.
48. European Medicines Agency, Design of an analysis of EudraVigilance data. 2018.
49. Jokinen, J., et al., Industry Assessment of the Contribution of Patient Support Programs, Market
Research Programs, and Social Media to Patient Safety. Therapeutic Innovation & Regulatory
Science, 2019. 53(6): p. 736-745.
50. Hazell, L. and S.A.W. Shakir, Under-Reporting of Adverse Drug Reactions. Drug Safety, 2006.
29(5): p. 385-396.
51. Misuse of E-Commerce for Trade in Counterfeits. OECD. 25 October 2021. Available from
https://www.oecd.org/en/publications/misuse-of-e-commerce-for-trade-in-
counterfeits_1c04a64e-en/full-report.html. Accessed 6 September 2026.
52. Aung, A.K., et al., Adverse drug reaction management in hospital settings: review on practice
variations, quality indicators and education focus. Eur J Clin Pharmacol, 2022. 78(5): p. 781-
791.
53. Bergvall, T., G.N. Norén, and M. Lindquist, vigiGrade: a tool to identify well-documented
individual case reports and highlight systematic data quality issues. Drug Saf, 2014. 37(1): p.
65-77.
54. Edwards, I.R., et al., Quality criteria for early signals of possible adverse drug reactions. Lancet,
1990. 336(8708): p. 156-8.
55. Substandard and falsified medical products. WHO. Available from: https://www.who.int/health-
topics/substandard-and-falsified-medical-products#tab=tab_1. Accessed 6 September 2026.
56. Spontaneous adverse event notifications by patients subsequent to the marketing of a new
formulation of Levothyrox® amidst a drug media crisis: atypical profile as compared with other
drugs. Viard D et al. 2019 Aug;33(4):463-470. doi: 10.1111/fcp.12446. Epub 2019 Feb 7.
57. Guideline on computerised systems and electronic data in clinical trials.
EMA/INS/GCP/112288/2023. Available from:
https://www.ema.europa.eu/en/documents/regulatory-procedural-guideline/guideline-
computerised-systems-and-electronic-data-clinical-trials_en.pdf. Accessed 16 March 2026.
58. Kiguba, R., et al., Navigating duplication in pharmacovigilance databases: a scoping review. BMJ
Open, 2024. 14(4): p. e081990.
59. van Stekelenborg, J., et al., Individual Case Safety Report Replication: An Analysis of Case
Reporting Transmission Networks. Drug Saf, 2023. 46(1): p. 39-52.
60. Bailey, C., et al., Adverse drug event reporting systems: a systematic review. Br J Clin
Pharmacol, 2016. 82(1): p. 17-29.
61. Report of the CIOMS Working Group XIV – Artificial Intelligence in Pharmacovigilance. Available
from : https://cioms.ch/artificial-intelligence-inpv/. Accessed 21 June 2026.
62. MedDRA Term Selection: Points to Consider. Latest version available from
https://www.meddra.org/how-to-use/support-documentation/english/welcome. Accessed 22
June 2026.
63. Kakalou, C., et al., Exploiting Social Media for Active Pharmacovigilance: The PVClinical Social
Media Workspace. Stud Health Technol Inform, 2022. 290: p. 739-743.
64. Brosch, S., et al., Establishing a Framework for the Use of Social Media in Pharmacovigilance in
Europe. Drug Saf, 2019. 42(8): p. 921-930.
65. Gordillo-Maranon, M., et al., COVID-19 Vaccines and Heavy Menstrual Bleeding: The Impact of
Media Attention on Reporting to EudraVigilance. Drug Saf, 2024. 47(8): p. 783-798.
66. Recommendations on the use of Auxiliary Medicinal Products in Clinical Trials written and
endorsed by the Clinical Trials Coordination and Advisory Group (CTAG). Available from:
https://health.ec.europa.eu/document/download/1bf35e45-134c-4f2e-9c68-0e8537eef867_en.
2025.
67. Guideline on specific adverse reaction follow-up questionnaires (Specific AR FUQ).
EMA/PRAC/490455/2023. Available from: https://www.ema.europa.eu/en/documents/scientific-
guideline/guideline-specific-adverse-reaction-follow-questionnaires-specific-ar-fuq_en.pdf-0.
2024.
68. SCOPE Work Package 4 ADR Collection: Paper ADR Reporting Forms. Available from: scope-
training-paper-adr-reporting-forms_en.pdf. Accessed 02 March 2026.
69. van Stekelenborg, J., et al., Recommendations for the Use of Social Media in Pharmacovigilance:
Lessons from IMI WEB-RADR. Drug Saf, 2019. 42(12): p. 1393-1407.
Page 45/49
第 46 页
70. EudraVigilance: electronic reporting - ICSR compliance and quality. Available from:
https://www.ema.europa.eu/en/human-regulatory-overview/research-
development/pharmacovigilance-research-development/eudravigilance/eudravigilance-
electronic-reporting#icsr-compliance-and-quality-83714. Accessed 02 March 2026.
71. Compliance notifications explanation and Q&A. EMA/362031/2025. Available from:
https://www.ema.europa.eu/en/documents/other/compliance-notifications-explanation-
qa_en.pdf. 2025.
72. Introductory Guide for Standardised MedDRA Queries (SMQs) Version 23.1, 000491
https://admin.meddra.org/sites/default/files/guidance/file/SMQ_intguide_23_1_English.pdf.
2020.
73. ICH Topic E 2 A - Clinical Safety Data Management: Definitions and Standards for Expedited
Reporting 1995.
74. Kumar Vermula, G., et al., A Risk-Based Approach for Safety Case Follow-up of Adverse Event
Reports in Pharmacovigilance. Adv Ther, 2023. 3(1): p. 82-91.
75. Spontaneous reporting of adverse drug reactions as an outlet for patient dismay? The case of
Levothyrox® change of excipients. Romani S et al.Fundam Clin Pharmacol. 2022.
Jun;36(3):553-562. doi: 10.1111/fcp.12755. Epub 2022 Jan 18.
76. MedDRA® Best Practices - Recommendations for Implementation and Use of MedDRA. 2018.
77. EMA - Reflection paper on the use of Artificial Intelligence (AI) in the medicinal product lifecycle,
EMA/CHMP/CVMP/83833/2023.
78. ICH guideline E19 on a selective approach to safety data collection in specific late-stage pre-
approval or postapproval clinical trials. Accessed 27 August 2026.
79. Guideline on good pharmacovigilance practices (GVP) Module XVI Addendum II – Methods for
evaluating effectiveness of risk minimisation measures. 6 August 2024.
80. FAIR. Available from: https://www.go-fair.org/fair-principles/.
81. EMA Guideline on good pharmacovigilance practices (GVP), Module VI Addendum I – Duplicate
management of suspected adverse reaction reports, EMA/405655/2016. 2017.
82. WHO - Safety of Medicines. A guide to detecting and reporting adverse drug reactions. 2022.
1328
Page 46/49
第 47 页
1329 9. Glossary
1330 This glossary addresses the main terms and definitions that have been used in this DQF application to
1331 ADR reporting and submission for regulatory decision making.
Definitions Explanation
ADR pipeline The sequence of steps of an ADR through the data lifecycle from data
collection to delivery for consumption.
An ADR is characterised by at least a reasonable possibility of a causal
Adverse Drug
relationship between a medicinal product and an adverse event
Reaction
Any untoward medical occurrence that may present during treatment
Adverse Event
with a medicinal product, but which does not necessarily have a causal
relationship with this treatment.
Healthcare In this document the term Healthcare Professional, abbreviated to HCP,
Professional (HCP) refers to professional providers of healthcare in both the human and
veterinary domains. Examples of such persons include doctors, nurses,
pharmacists and veterinarians.
Immutability Data immutability refers to the principle that data is never deleted or
modified. Once data is entered into a system, it can only be altered with
new information while preserving the knowledge of the previous state.
Individual Case Individual Case Safety Report (ICSR) refers to the format and content for
Safety Report the submission of an individual report of adverse reactions in relation to a
medicinal product that occur in a single patient at a specific point of time
[17]
Pharmacovigilance The process of detecting, evaluating, understanding, and preventing
adverse effects of medicines is defined as pharmacovigilance.
Reporting The process of a primary source informing a Sponsor, MAH or NCA about
an ADR.
Throughout this document “ADR reporting” is used to refer to the entire
process of ADR collection, management and submission
Sender organisation An organisation which collects ADR data from primary sources and
submits ICSRs to EudraVigilance
Submission The process of a Sponsor, MAH or NCA submitting an ICSR to EV or EVV
(or to another receiver’s pharmacovigilance database)
Suspected A serious adverse reaction which occurred in a clinical trial, of which the
Unexpected Serious nature, severity or outcome is not consistent with the reference safety
Adverse Reaction information.
Unexpected events Events that affect the benefit-risk balance of a clinical trial that were
unforeseen, e.g. an unexpected increase in the incidence of
expected serious adverse reactions or new non-clinical data that may
be clinically important. Unexpected events do not include SUSARs. These
should be reported via the Clinical Trials Information System (CTIS)
1332
Page 47/49
第 48 页
1333 Appendix I – Overview of guidelines
1334 Overview of guidelines and frameworks for ADR reporting
Human ADR reporting
EMA Access to EudraVigilance data webpage [39] ADR reporting platform
EudraVigilance
EMA Detailed guide regarding the EudraVigilance data Data management and
management activities by the European Medicines quality assurance activities
Agency, 2020 [16] performed by EMA for ADRs
submitted to EudraVigilance.
EMA EU Individual Case Safety Report (ICSR) Implementation Implementation guide
Guide, 2021 [20]
EMA Good practice guide on recording, coding, reporting and Good practice guide
assessment of medication errors [19]
EMA Guideline on computerised systems and electronic data in CT Guideline
clinical trials [79]
GVP Guideline on good pharmacovigilance practices (GVP) Good Vigilance Practices
Module IV – Pharmacovigilance audits (Rev 1), 2015 [68]
GVP Guideline on good pharmacovigilance practices (GVP) - Good Vigilance Practices
Module VI – Collection, management and submission of
reports of suspected adverse reactions to medicinal
products (Rev 2) [17]
GVP GVP Module IX on signal management [18] Good Vigilance Practices
GVP Guideline on good pharmacovigilance practices (GVP) Good Vigilance Practices
Product- or Population-Specific Considerations III:
Pregnant and breastfeeding women and their children
exposed in utero or via breastmilk
ICH ICH Harmonised Tripartite Guideline: Clinical Safety Data Scientific guideline
Management: Definitions and Standards for Expedited
Reporting E2A [8]
ICH ICH E2B Implementation Working Group Implementation Scientific guideline
Guide for Electronic Transmission of Individual Case
Safety Reports (ICSRs) - E2B(R3) Data Elements and
Message Specification [4]
ICH ICH Harmonised Tripartite Guideline: periodic benefit-risk Scientific guideline
evaluation report (PBRER) E2C(R2) [9]
ICH ICH Harmonised Guideline: Post-approval safety Scientific guideline
management: definitions and standards for management
and reporting of individual case safety reports E2D(R1)
[10]
ICH ICH Harmonised Tripartite guideline: Development Safety Scientific guideline
Update Report E2F [11]
MedDRA Best practices for MedDRA – Recommendations for Best practices MedDRA
Implementation and use of MedDRa [76]
MedDRA Introductory Guide for Standardised MedDRA Queries MedDRA guide
(SMQs) Version 23.1 [72]
WHO Safety of Medicines - A guide to detecting and reporting High-level guide
adverse drug reactions [82]
Page 48/49
第 49 页
1335
Veterinary AE reporting
EMA Combined Veterinary Dictionary for Drug Regulatory VeDDRA guideline
Activities (VeDDRA) list of clinical terms for reporting
suspected adverse events in animals and humans to
veterinary medicinal products
(EMA/CVMP/PhVWP/10418/2009 Rev.16)
EMA Guidance notes on the use of VeDDRA terminology for Best practices VedDRA
reporting suspected adverse events in animals and
humans [28]
EMA EU VICH adverse event report implementation guide [29] Implementation guide
EMA Veterinary Union pharmacovigilance database best Best practice guide
practice guide (BPG) (EMA/371192/2021) [27]
VGVP Guideline on veterinary good pharmacovigilance practices VGVP practice
(VGVP) - Module: Collection and recording of suspected
adverse events for veterinary medicinal products
(EMA/306663/2021)
VGVP Guideline on veterinary good pharmacovigilance practices VGVP practice
(VGVP) - Module: Signal Management
(EMA/522332/2021)
VICH VICH Pharmacovigilance of veterinary medicinal products: PV guidelines for veterinary
management of adverse event reports (AERs) [33] products
1336
Page 49/49
来源：EMA Regulatory and Procedural Guidelines · ema.europa.eu