Regulations and compliance · Practical guide

ICH E6(R3) data governance: putting data responsibilities into practice

An EDC, central laboratory and service provider each hold study data. Who detects a transfer that omits one observation and duplicates another? ICH E6(R3) data governance connects these flows to accountable roles and controls. Trace one real data path and establish what a reviewer can verify.

Table of contents

At a glance

What does ICH E6(R3) data governance mean for the sponsor and investigator site?

ICH E6(R3) describes governance of clinical data throughout their lifecycle. The investigator and sponsor retain their respective responsibilities. Transferred activities require suitable agreements and oversight. Link collection, relevant metadata, corrections, transfers, preparation for analysis and retention according to risk. As of 6 October 2026, the Principles and Annex 1 were effective in the EU. Annex 2 follows on 15 January 2027.

  • Data responsibility cannot be solely inferred from the storage location.
  • Criticality determines the appropriate controls and the depth of those controls.
  • Same number of lines confirms neither identity nor completeness.
  • Finalization and archiving require traceable decisions.

1. Which portion of E6(R3) is currently applicable?

The EMA will implement Principles and Annex 1 from 23 July 2025 as the current EU version. Annex 2, already adopted, and the consolidated version will come into EU application on 15 January 2027. As of 6 October 2026, Section 4 of the current Annex 1 is particularly relevant for this article. A published document is not automatically applicable to all aspects solely because it has been issued.

The considered framework comprises clinical drug product investigations. GCP responsibility is not equivalent to GMP batch release or general IT security. Additional requirements, such as data privacy, must be specifically addressed within the respective protocol. This article discusses data controls, and does not constitute a comprehensive legal assessment of a study.

Sources: [1] [2]

2. Responsibility follows the activity, not just the system

Section 4 addresses investigators and sponsors. Read it in the context of their respective responsibilities. A sponsor may transfer activities to service providers but retains responsibility for its trial-related duties. Investigators likewise remain responsible for activities they delegate. Using the same cloud environment does not merge these responsibilities.

Create a concise assignment: Who generates the observation, who evaluates queries, who operates the transfer, and who has authority to approve a correction? Add the evidence supporting the execution. This makes it visible where a task has been assigned and where only a technical feasibility exists.

Sources: [2] [3]

3. Connect a data pathway with controls

The data lifecycle spans from capture through processing and review to analysis and retention. The required depth aligns with the importance for participant safety and result reliability. A frequently used field may be less critical than a rare piece of information that influences a key study decision.

The following internal matrix assigns practical questions to a data pathway. It is not a form prescribed by ICH. For a manageable process, clear statements suffice; more complex transitions require corresponding additional detail and supporting evidence.

Own Governance Matrix for a Clinical Data Pathway
StepClarifying ResponsibilityControl questionEvidence
Data capture Source and responsible role Is the observation captured with required context? Identity, Value, Unit, and Time Reference
Review/Correction Qualified technical role Is the change justified and traceable? Source, original value, and change history
Transfer Sender, recipient and subject matter expert Are identities and significances completely preserved? Transfer setup and appropriate reconciliation
Analysis preparation Responsible parties as defined in the study plan Is the selected data version finalised and traceable? Evaluated open items and documented status
Retention Responsible Party and Archival Operation Are data understandable and accessible later? Stock, access, readability, and deadline assignment

Sources: [2]

4. Consider context and blinding from the outset

A value without a unit or participant assignment may be unusable. E6(R3) addresses relevant metadata, including audit trails, and their risk-based evaluation and review. This does not imply that every technically generated information carries the same meaning, nor that a total volume of logs alone validates data quality.

Simultaneously verify authorized access. In a blinded study, an export field or unexpected linkage may reveal more than the user interface itself. Roles, data transfers, and analysis sets must take into account the intended blinding. An administrator right does not constitute a professional authorization to view or modify all study information.

Sources: [2]

5. Example: three lines do not constitute three valid observations

Example without real patient or study data: A verified source contains the observations 101, 102, and 103. Their values are 1.25, 2.50, and 3.75 in the same predefined unit. Each identifier shall appear exactly once in the defined transfer. The expected sum is 7.50; this sum serves as an additional control value and is not a medical endpoint.

The received file contains three lines, but the identifiers 101, 102, and 102. The second observation was duplicated, and identifier 103 is missing. As a result, only two unique identifiers are present, and the total sum is 6.25. A simple quantity reconciliation would be successful, despite the incorrect inventory.

The recipient confirms the finding and verifies the handover based on the unaltered source version. After controlled correction, the file 101, 102, 103 contains appropriate values. The erroneous delivery, justification for correction, and the new version used remain traceable and transparent. Neither a duplicate nor a missing value is silently attributed to random imperfection.

Own matching matrix: check number, identity and content separately
StatusIdentifiersRows and unique identifiersChecksumAssessment
Expected source 101, 102, 103 3 / 3 7.50 Expected stock balance
Faulty transfer 101, 102, 102 3 / 2 6.25 Dublette 102; 103 missing
Controlled and corrected 101, 102, 103 3 / 3 7.50 Additionally verify values and context

Sources: [2] [3]

6. Review and correction are planned professional activities

E6(R3) describes a planned, risk-based review of study data, audit trails, and relevant metadata. Define the rationale, scope, responsibilities, and handling of anomalies. "Audit Trail activated" describes a technical attribute but does not constitute a completed review of a relevant trail.

In the example, it must first be determined whether the error occurred solely during transfer or already at the source. Any correction must include proper assignment and valid justification; it must not render the original entry incomprehensible. If the error recurs, the responsible process must also be investigated. A general instruction to manually correct every deviation does not protect future transfers.

Sources: [2] [3]

7. Plan finalization, operation, and storage together

Before analysis, the data version used must be determined in accordance with established procedures. Open queries, corrections, and relevant deviations will be assessed for their significance. A technical database lock is a possible means; however, it alone confirms neither technical completeness nor the correctness of the selected analysis set.

Even after active data capture, responsibility remains. Data and required context must remain retrievable, readable, and protected against unauthorized modifications throughout the relevant retention period. System transitions and migrations must not implicitly terminate this capability. Required retention periods shall comply with applicable requirements; E6(R3) does not specify a single retention period for each file.

Sources: [2] [3]

Further reading:

8. Checklist for a manageable Governance Process

A small study team requires clear decisions and a realistically assignable accountable individual for each transition. Start with a data flow of significant importance. At every step, ask what a plausible error might look like and what evidence would make it visible. This makes the concept more concrete than a mere abstract role description.

Our checklist supports preparation. It does not replace study-specific planning and does not require an additional document when the information is already reviewed and readily accessible.

  • Purpose and significance of the data for the study.
  • Source, identity, unit, and relevant metadata are defined.
  • Clearly define the responsibilities of sponsors, reviewers, and service providers.
  • Consider access and blinding also in exports.
  • Compared on values and identities, not just quantities.
  • Plan review, inquiries, and justified corrections.
  • Finalize the analysis and open items with traceable documentation.
  • Assign storage, future access, and controlled withdrawal.

Sources: [2] [3]

Project experience and collaboration

Trace a study data item across handovers

With clinical data owners, we trace one observation from capture to its intended analysis. At each handover, we examine how changes, missing data and duplicate deliveries are handled and recorded. A straightforward transfer may need clearer reconciliation rules. Multiple vendors or processing stages require coordinated ownership of information and controls. We support technical evidence and clear responsibilities. The sponsor and investigator retain their respective duties.

Frequently asked questions about ICH E6(R3): Data Governance

What does data governance mean in practice according to ICH E6(R3)?

It combines data responsibility and appropriate controls over the entire life cycle. Practically this means: For a relevant observation source, processing, transmission, decision and later availability are comprehensible. A software function or an organizational chart alone does not cover this process.

Sources: [2]

Does the new Annex 2 already apply upon its publication?

No. The EMA designates 15 January 2027 as the EU implementation date for Annex 2. Principles and Annex 1 are already effective as of 6 October 2026. Clearly specify the version, edition, and date; a consolidated publication may contain different temporal applicability periods.

Sources: [1]

Does responsibility in a CRO extend completely?

No. According to ICH E6(R3) 3.6.6, the sponsor may delegate certain activities; however, ultimate responsibility for study-related tasks, including participant safety and data reliability, remains with the sponsor. Activities not explicitly delegated remain with the sponsor (3.6.4). The auditor's responsibilities must be considered separately (3.6.5). Therefore, specific arrangements for execution, oversight, and information flow must be clearly defined.

Sources: [2]

Must we evaluate all fields with the same intensity?

The approach is risk-based and takes into account the importance of participant safety and the reliability of results. This is not a license to ignore inconvenient data. Justify the selection and extent of controls based on the study purpose and review them in light of new knowledge.

Sources: [2]

Are metadata merely technical supplementary information?

No. Relevant metadata can only clarify to whom a value belongs, when it was generated, and how it was modified. Determine the required context and its validation and retention relevance. The indiscriminate collection of all technical logs does not substitute for this professional judgment.

Sources: [2] [3]

Does a passed edit check confirm the accuracy of an observation?

He confirms only the verified rule. A formally plausible value may still be incorrectly assigned or incorrectly entered. Therefore, source tracing, appropriate technical review, and verifiable correction remain essential; technical plausibility and factual accuracy are not equivalent.

Sources: [2] [3]

Why does the example display an error despite having the same number of entries?

The three received lines contain only two distinct identifiers. Observation 103 is missing, and observation 102 appears twice. Additionally, the checksum differs. Our own matrix indicates that the count, identity, and content provide divergent test results.

Sources: [2] [3]

Can a faulty transfer be simply overwritten?

A correction must remain traceable and justified. In the example of this article, erroneous status, source comparison, and controlled corrected delivery are linked. How this is technically implemented depends on the system; an undetected replacement would render the result incomprehensible.

Sources: [2] [3]

Is a database lock equivalent to data quality?

No. A technical status does not replace the planned activities for finalization. The responsible parties must assess the current state and the significance of open items. The lock can support this process, but it does not automatically validate a correct analysis set.

Sources: [2]

Do we need an additional data governance handbook?

Not mandatory as a separate document. Critical are controlled, clear, and accessible information regarding responsibilities and data pathways. If existing study protocols and procedures adequately cover these details, a clear linkage may be more appropriate than a new parallel description.

Sources: [2] [3]

Sources and review date

Source date: 6 October 2026. EMA Guidelines 2023, together with ICH E6(R3) Principles and Annex 1, effective in the EU from 23 July 2025. Annex 2 will become effective in the EU from 15 January 2027. Case numbers are derived from internal work aids.

  1. EMA: ICH E6 Good Clinical Practice – Status of Adoption and ApplicationRetrieval date: 06.10.2026: Principles / Annex 1 current as of 23.07.2025; Annex 2 EU implementation 15.01.2027
  2. ICH E6(R3): Principles and Annex 1 – EMA Step 5EMA/CHMP/ICH/135/1995; EU Application 23.07.2025
  3. EMA: Guideline on computerised systems and electronic data in clinical trialsEMA/INS/GCP/112288/2023; Document 09.03.2023, final acceptance 07.03.2023; Application six months after publication

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Traceable link between study processes and system validations

We support you in defining your system usage, providing appropriate evidence, and establishing clear responsibilities among functional areas, QA, IT, and external service providers.

  • Computer system validation →

    Bring your GxP systems into validated operation, from laboratory software, MES and eQMS to ERP. We define the scope, plan risk-based tests using GAMP 5 and hand the evidence to QA for release.

Send a project enquiry Customer testimonials →