In This Article
- Executive Summary
- Where ICH E6(R3) Stands Today
- What Changed for Computerized Systems
- Gap 1: Proportionate Validation Is Not Lighter Validation
- Gap 2: Audit Trails That Exist but Cannot Be Reviewed
- Gap 3: Access Controls and the Record of Who Could Do What
- Gap 4: Data Integrity Across the Whole Trial Lifecycle
- Gap 5: Oversight of Service Providers’ Systems
- A Sequenced Remediation Plan
- Conclusion
- For Further Reading
- References & Sources
Executive Summary
ICH E6(R3) is no longer new. The principles and Annex 1 have been in effect in the EU since 23 July 2025, FDA published them as final guidance in September 2025, Health Canada implemented them on 1 April 2026, and in the UK the principles became a legal requirement with the new clinical trials regulations on 28 April 2026. Annex 2, on decentralized elements, pragmatic designs and real-world data, reached ICH Step 4 in June 2026 and takes effect in the EU on 15 January 2027. FDA has not yet finalized it.
The guideline’s computerized system expectations are specific and testable. Sponsors need a record of important systems, validation proportionate to risk, audit trails that are interpretable and reviewed, access controls tied to delegation, and oversight that reaches into their service providers’ systems. Inspection findings show the same weak points again and again. EMA inspectors’ 2024 data names vendor qualification and oversight, and data migration and system validation, and older MHRA findings describe audit trails that could not be produced or read.
This article sets out what is in force where, what E6(R3) says about computerized systems in its own words, the five gaps we see most often, and a sequenced plan for closing them before the next inspection rather than during it.
Where ICH E6(R3) Stands Today
Before looking at what sponsors miss, it helps to be exact about what applies and where. It is tempting to describe E6(R3) as “six months in” this autumn, and that is only partly right. For the EU and the US, the guideline has been in place for more than a year. For the UK and Canada, six months is close to accurate. And one piece of the guideline, Annex 2, is not yet in effect in the EU or the US.
The Principles and Annex 1
The ICH Assembly adopted the final E6(R3) guideline (the overarching principles and Annex 1) at Step 4 on 6 January 2025.2 The European Medicines Agency states that the principles and Annex 1, as adopted by ICH and the CHMP, came into effect on 23 July 2025.1 Because the EU Clinical Trials Regulation requires trials to follow good clinical practice, that date matters: for EU trials, E6(R3) became the GCP reference for work done from July 2025 onward.
In the United States, FDA announced the final guidance “E6(R3) Good Clinical Practice” in the Federal Register on 9 September 2025.4 The notice says the guidance includes the principles document and Annex 1, and that once complete it will also include Annex 2.4 FDA’s guidance page lists it as final guidance issued in September 2025.5 Like all FDA guidance, it contains nonbinding recommendations and describes FDA’s current thinking. The binding requirements remain in the regulations, including 21 CFR Part 11 (FDA’s rule on electronic records and electronic signatures) and the investigational drug rules in 21 CFR Part 312.
In the UK, E6(R3) arrived with the largest change to the clinical trials regulations in two decades. MHRA and the Health Research Authority wrote in October 2025 that the new regulations would come into force on 28 April 2026. MHRA’s April 2026 guidance makes the legal requirement compliance with the principles, not the entirety of the guideline, and treats the annexes as guidance that sponsors must have regard to.8 The Health Research Authority confirmed on 28 April 2026 that the new regulations were in force.9
Health Canada lists E6(R3) as implemented on 1 April 2026.10
Annex 2 Is a Separate Timeline
Annex 2 adds GCP considerations for trials with decentralized elements, pragmatic elements and real-world data sources. The ICH version was adopted on 3 June 2026.7 EMA reports CHMP adoption on 25 June 2026 and an EU effective date of 15 January 2027. EMA has also published a consolidated version of the guideline with the same legal effective date.1 FDA’s page for Annex 2 still shows the December 2024 draft guidance, marked “Not for implementation.”6
| Region | Principles and Annex 1 | Annex 2 | Time in effect by early October 2026 |
|---|---|---|---|
| European Union | In effect 23 July 20251 | Adopted by CHMP 25 June 2026; in effect 15 January 20271 | About 14 months |
| United States | Final FDA guidance, September 2025 (nonbinding)45 | Draft guidance, December 20246 | About 13 months |
| United Kingdom | In force 28 April 2026 with the amended regulations89 | Not addressed in the sources cited here | About 5 months |
| Canada | Implemented 1 April 202610 | Implemented November 4, 202610 | About 6 months |
Why the dates matter for computerized systems. A global program running today is working under E6(R3) in every major region. The EU has applied it longest, so EU inspectors have had the most time to form expectations. When you prioritize remediation, assume your next EU or UK inspection will test against the text of E6(R3) and against EMA’s 2023 guideline on computerized systems in clinical trials, which predates R3 but covers the same ground in more detail.11
A Note on What the Inspection Data Can and Cannot Tell Us
The most detailed recent public data on GCP inspection findings in Europe is the GCP Inspectors Working Group annual report for 2024, adopted in December 2025.13 Those inspections were conducted before E6(R3) took effect in the EU, so they measure compliance with the earlier rules. We use them here for what they are: a record of the weak points sponsors carried into the R3 era. Nothing in the E6(R3) text makes those weak points easier to defend. In most cases the new text states the expectation more directly than R2 did.
What Changed for Computerized Systems
E6(R2) addressed electronic systems mainly in section 5.5.3, as part of the sponsor’s data handling duties. E6(R3) gives computerized systems far more space and places them inside a broader data governance section that applies to both investigators and sponsors. The changes that matter most for sponsors fall into four groups.
Data Governance Is Now Its Own Section
Section 4 of Annex 1, “Data governance: investigator and sponsor,” lists key processes that should cover the full data lifecycle with a focus on the criticality of the data. One of those processes is managing computerized systems so that they are fit for purpose and used appropriately.3 The section then works through the data lifecycle: data capture, metadata and audit trails, review of data and metadata, data corrections, data transfer and migration, finalizing data sets before analysis, retention and access, and destruction. Section 4.3 then covers computerized systems directly: procedures, training, security, validation, system release, system failure, technical support and user management.3
The “Responsible Party” Replaces Vague Ownership
Section 4.3 is written around the “responsible party,” which means the sponsor or the investigator, depending on who deploys and relies on the system. It says the sponsor’s and investigator’s responsibilities, and the activities of other parties, for each computerized system should be clear and documented. It also says the responsible party should make sure that anyone developing systems on its behalf knows the intended purpose and the regulatory requirements that apply.3 That sentence puts a duty on the sponsor to brief its vendors. It is not enough to buy a validated product and assume the vendor understands how your protocol will use it.
The Sponsor’s Duties Are Split by Who Deploys the System
Section 3.16.1(x) sets out what the sponsor should do when computerized systems are used in a trial. It separates systems deployed by the sponsor from systems used or deployed by the investigator or institution, and then adds one duty that applies to all systems.3
Record, Requirements, Users, Defects
Keep a record of important systems (use, functionality, interfaces, validation status, owner, access controls and security measures). Make sure validation, audit trail, user management, backup, disaster recovery and IT security requirements are in place and proportionate. Keep a record of authorized users, roles and permissions. Make site permissions match investigator delegation and be visible to the investigator. Give service providers and investigators a way to report defects.
Fitness-for-Purpose Assessment
Where site systems hold source records (electronic health records, other record-keeping systems, investigator site files), assess whether they are fit for purpose, or whether known risks can be mitigated. Do this during site selection, document it, and complete it before the system is used in the trial. Consider data security, backup, user management and audit trails, proportionate to the importance of the data.
Incident Escalation
Have a process for service providers and investigators to tell the sponsor about incidents that could be a serious noncompliance with the protocol, trial procedures, regulatory requirements or GCP.
Fit for Purpose, Validated Based on Risk
The principles say computerized systems used in trials should be fit for purpose (for example, through risk-based validation, if appropriate), and that the factors critical to their quality should be addressed in their design or adaptation for trial use.
Proportionality Cuts Both Ways
E6(R3) asks for requirements that are “proportionate to the importance of the computerised system and the data or activities they are expected to process.”3 Many teams have read proportionality as permission to do less. It is permission to do less where risk is low, and it is also an obligation to do more where risk is high, and to be able to show the reasoning. A proportionate approach without a written risk assessment is not proportionate. It is just a smaller effort with no record of why.
FDA’s October 2024 final guidance on electronic systems, records and signatures in clinical investigations uses almost the same three risk considerations as E6(R3): the intended use of the system, the purpose and importance of the data or records it handles, and its potential to affect participant rights, safety and welfare or the reliability of trial results.12 That alignment is useful. One well-built risk assessment can serve both the EU and US expectations.
Gap 1: Proportionate Validation Is Not Lighter Validation
The first gap is the one sponsors are most confident they have closed. Most sponsors use vendor platforms for electronic data capture (EDC), interactive response technology (IRT) for randomization and supply, and electronic clinical outcome assessment (eCOA). Most have a vendor audit on file and a user acceptance test for each study build. The problem is usually not that validation is missing. It is that validation does not cover what E6(R3) says it should cover.
What the Text Asks For
Section 4.3.4 is specific. It says:
- “The responsible party is responsible for the validation status of the system throughout its life cycle.”3 The duty does not end at go-live.
- Both standard system functionality and protocol-specific configurations and customizations, including automated data entry checks and calculations, should be validated. Interfaces between systems should also be defined and validated.3
- Validation should cover critical functionality in particular, and the guideline names examples: randomization, dosing and dose titrations and reductions, and collection of endpoint data.3
- Changes to a system should be validated based on risk and should consider both previously collected and new data, in line with change control.3
- Periodic review may be appropriate to confirm systems remain in a validated state.3
- The responsible party should make sure systems developed by other parties are validated as fit for purpose for the trial, and that validation documentation is maintained and retained.3
- Unresolved issues should be justified and, where relevant, have mitigations in place before or during continued use.3
Where Sponsors Fall Short
The study build is tested, but the derivations are not. User acceptance testing often confirms that forms display and that data saves. It is less likely to test every edit check, every calculated field (body mass index, dose based on weight, derived eligibility flags), and every branch of the randomization and dosing logic. E6(R3) names automated checks, calculations and dose titration explicitly. If a calculated dose field is wrong, participants are at risk. An inspector who asks for the test evidence for one calculated field on a critical endpoint form will learn a lot about your program.
Interfaces are assumed rather than validated. Data moves between IRT and EDC, from central labs and imaging vendors, from eCOA devices and increasingly from electronic health records. Section 4.3.4(e) says interfaces should be defined and validated, and section 4.2.5 says data transfers should use validated processes or other appropriate processes such as reconciliation, with documentation that supports traceability.3 FDA’s 2024 guidance lists data transfers and interfaces among the things validation should cover.12 Many sponsors can show that each system was validated on its own, but cannot show a validated transfer specification and test for the link between them.
Vendor validation is relied on without being examined. Relying on a vendor’s validation work is allowed. EMA’s guideline on computerized systems in clinical trials says the responsible party using a vendor’s validation documentation should be thoroughly familiar with the vendor’s quality system and validation work, usually through an in-depth examination such as an audit. It adds that the examination should review a suitable number of relevant key requirements and the matching test cases, and should be documented.11 A postal questionnaire and a certificate do not meet that description.
Vendor releases outpace sponsor change control. Software-as-a-service platforms release on the vendor’s schedule. EMA’s guideline recognizes that some vendors release new versions at short notice and says that in that situation the sponsor should evaluate the vendor’s pre-release validation process and strengthen its own periodic review and change control. It also says new functions should not be used until the sponsor has validated them or reviewed and assessed the vendor’s documentation.11 FDA’s guidance says changes, including software upgrades and security patches, should be evaluated and validated across the system lifecycle depending on risk.12 In practice, few sponsors track which platform release each active study is running and what changed in it.
These are not new problems. In a 2017 post on computer system validation in GCP, an MHRA inspector described an organization that could only produce draft user requirements and functional specifications for a major release. Of four business-critical functions the inspector sampled, test scripts could be found for only one, that one had failed, and there was no documented evidence the failure had been fixed.15 The same post lists common user-related findings, including users given access with no training and users given higher access than their role needed.15
A practical test. Pick one active study. Ask for the validation evidence for (1) the randomization and dosing logic, (2) one calculated field on a primary endpoint form, (3) the IRT-to-EDC interface, and (4) the most recent vendor release that touched the study. If any of the four takes more than a day to produce, or does not exist, that is the gap an inspector will find.
What Proportionate Looks Like
A proportionate approach starts with the protocol’s critical-to-quality factors and works outward. Systems and functions that touch eligibility, randomization, dosing, safety reporting and primary endpoints get full requirement-level testing with traceable evidence. Supporting functions (report formatting, site contact lists, non-critical dashboards) can rely more on vendor evidence and lighter checks. What makes this defensible is the written link between the risk assessment and the depth of testing. E6(R3) also notes that different degrees of validation may be needed for bespoke systems, configurable systems, and systems used with no changes.3 Document which category each system belongs to and why.
Gap 2: Audit Trails That Exist but Cannot Be Reviewed
Almost every clinical system on the market records an audit trail. The E6(R3) expectation goes much further than having one.
What the Text Asks For
Section 4.2.2 says the approach to metadata, including audit trails, for data of higher criticality should include evaluating each system for the metadata it keeps. That covers logs of user account creation, changes to roles and permissions and user access; documentation of initial data entry and every later change or deletion, with the reason where appropriate; and records of workflow actions, not only direct data entry.3 It then lists four more duties:
- “Ensuring that audit trails, reports and logs are not disabled.” Audit trails should not be modified except in rare circumstances, such as removing a participant’s personal information that was entered by mistake, and only with a log and justification.3
- Ensuring that audit trails and logs are interpretable and can support review.3
- Ensuring that automatic date and time capture is unambiguous, with coordinated universal time (UTC) given as an example.3
- Deciding which metadata need review and retention.3
Section 4.2.3 then says procedures for reviewing trial-specific data, audit trails and other relevant metadata should be in place, and describes that review this way: “It should be a planned activity, and the extent and nature should be risk-based, adapted to the individual trial and adjusted based on experience during the trial.”3
EMA’s 2023 guideline adds detail. It expects procedures for risk-based, trial-specific audit trail reviews, focused on critical data, and says review should be proactive and ongoing unless justified. It lists what review can detect: missing data, signs of data manipulation, outliers, data entered at unexpected hours, incorrect processing, unauthorized access and device or system malfunction.11 FDA’s 2024 guidance says audit trails should be protected from modification and from being disabled, that periodic review may be helpful based on a risk assessment, and that audit trails should be retained in a searchable and sortable format where practical.12
Where Sponsors Fall Short
The audit trail cannot be read at the level a reviewer needs. Many systems export audit data as long, flat files with internal field codes. They are complete but not interpretable. E6(R3) asks for interpretable audit trails, which means someone has to define, per system, what a usable review output looks like and confirm the vendor can produce it.
Review is not planned, documented, or linked to risk. The typical position is that data management “looks at audit trails as needed.” That does not meet a standard that calls for a planned, risk-based, documented review adjusted during the trial. MHRA’s GCP inspection metrics for April 2019 to March 2020 describe a contract research organization with four critical findings. The inspectors found that the eCRF audit trail was not in a format that supported system-level review of who changed what and when, and that there was no documentation showing the audit trail had been reviewed between database locks or during the trial. That mattered because CRO data management staff had edit rights in the eCRF that the investigator had not delegated to them.14
Audit trails for integrated data are missing or incomplete. When data enters one system and is pushed to another, the audit trail is often not transferred with it. In a 2021 post, an MHRA inspector described inspection findings where protocol-required clinical data was entered into IRT and integrated into the clinical database. The investigator did not keep control of that data or receive a copy of it and its metadata, and some of it was source data with no other record. The post states: “Without the full audit trail, it was not possible to verify what data was entered, when, by whom, and whether the data had changed at all.”16
Time stamps are ambiguous. Global trials run across time zones, and systems record times differently: server time, site local time, device time. Where two systems disagree about when something happened, reconstruction becomes hard. E6(R3)’s mention of UTC points to a simple fix: one stated time convention per system, recorded in the system record.
What good audit trail review looks like. A short, study-specific audit trail review plan that names the critical data and the systems holding them, defines what reviewers look for (late entries, changes after source data review, changes to endpoint data after database milestones, entries by unexpected roles, access outside normal hours), sets a frequency tied to trial risk, and records each review and its outcome in the trial master file. Technology can help with large volumes. The plan is what makes the review defensible.
Gap 3: Access Controls and the Record of Who Could Do What
Access control is where GCP and IT security meet. It is also where evidence goes missing most easily, because the facts change every week as site staff, monitors and vendor staff join and leave studies.
What the Text Asks For
Section 4.3.8 begins: “Access controls are integral to computerised systems used in clinical trials to limit system access to authorised users and to ensure attributability to an individual.”3 It then asks for procedures to assign access based on a user’s duties, blinding arrangements and organization; to revoke access when no longer needed; to review access and roles periodically where relevant; and to document authorized users and permissions, including time-stamped changes.3
For sponsor-deployed systems, section 3.16.1(x) adds two duties that sponsors often overlook: keep a record of the individual users authorized to access the system with their roles and permissions, and make sure access permissions for investigator site staff match the investigator’s delegations and are visible to the investigator.3 Section 3.16.1(r) says edit access to data acquisition tools should be restricted before data are provided for final analysis and, where applicable, before unblinding.3
EMA’s guideline goes into more detail in its annex on user management. It says access should be granted to trained site users only after the necessary approvals and documents are in place. It asks for user reviews at suitable intervals, least-privilege access, segregation of duties, and individual accounts. It states that “Sharing of accounts (group accounts) is considered unacceptable and a violation of data integrity and ICH E6 principles as data should be attributable.”11 It also says users with privileged or administrator access should be sufficiently independent from, and not involved in, the management and conduct of the trial and the generation, change and review of data.11
Where Sponsors Fall Short
- Site access does not match the delegation log. Site staff are given EDC or IRT roles by a help desk, based on a request form, with no check against the investigator’s current delegation. When staff leave, access often stays open until a periodic review, if one happens.
- The investigator cannot see who has access. E6(R3) asks that site permissions be visible to the investigator. Many platforms can produce a site user report, but no one sends it to the investigator or asks them to confirm it.
- Vendor and CRO administrator access is not reviewed. Sponsors review their own staff’s access but not the vendor’s support engineers, database administrators and study builders, who often hold the most powerful rights. EMA’s guideline is direct about the risk: privileged users can change system settings, deactivate audit trails and make back-end changes the audit trail does not capture.11
- Data management holds edit rights it should not have. The MHRA critical finding described above involved CRO data management staff with edit rights to the eCRF that the investigator had not delegated.14 E6(R3) also says the sponsor should not change data entered by the investigator or trial participants unless it is justified, agreed in advance by the investigator and documented.3
- Blinding is not built into role design. E6(R3) section 4.1 says maintaining the blind matters especially in system design, user account management and data access. It asks that roles and procedures for access to unblinded information be defined and documented.3 Role matrices often mix blinded and unblinded functions for convenience.
For a closer look at how single sign-on interacts with Part 11 signature expectations, see our earlier piece on that subject in the reading list below. The short version is that federated login does not remove the need to show who held which role on which date in each system.
| Access control expectation | Common gap | Evidence to have ready |
|---|---|---|
| Access limited to authorized users, attributable to one individual (4.3.8) | Shared study build or support accounts at vendors | Named account list for every role, including vendor staff |
| Site permissions match delegation and are visible to the investigator (3.16.1(x)(iv)) | Help desk grants access from request forms | Reconciliation of system roles against delegation logs, signed or acknowledged by the investigator |
| Access revoked when no longer needed; periodic review (4.3.8(b)) | Leavers remain active until annual review | Timely revocation records and dated periodic review outputs |
| Edit access restricted before final analysis and unblinding (3.16.1(r)) | Lock steps documented but role changes not evidenced | Role change logs at database lock and unblinding milestones |
| Privileged users independent of trial conduct (EMA guideline, Annex 3)11 | Vendor administrators also build and support the study | Privileged user list, justification, and review of privileged activity |
Gap 4: Data Integrity Across the Whole Trial Lifecycle
Data integrity in clinical trials is usually discussed at the point of entry: is the eCRF value attributable, legible, contemporaneous, original and accurate (the core ALCOA principles)? E6(R3) asks sponsors to think across the full lifecycle, from before the first participant is enrolled until well after the last system is switched off. The weak points tend to be at the two ends: the setup documentation that should exist before first use, and the retention and decommissioning work that should continue after the trial.
Before the Trial: The System Record and the Data Flow
Section 3.16.1(x)(i) asks sponsors to keep a record of the important computerized systems used in a trial. The record should describe each system’s use, functionality, interfaces and validation status, who is responsible for managing it, and the access controls and internal and external security measures in place.3 Section 3.16.1(c) says data to be collected and the method of collection should be pre-specified in the protocol and, where necessary, in a protocol-related document such as a data management plan, including a data flow diagram.3 FDA’s 2024 guidance similarly says sponsors should document, for each clinical investigation, the electronic systems used and their requirements, including a diagram that shows the flow of data from creation to final storage.12
Most sponsors keep an enterprise list of validated systems. Fewer keep a trial-level record that shows which systems a specific study uses, which vendor hosts each one, which version is live, how data moves between them, and who owns each link. That trial-level view is what an inspector asks for first. We wrote about building a defensible computerized system inventory in September; the same discipline applies here at study level.
Section 4.3.5 adds a release control: trial-specific systems, including updates for protocol amendments, should only be released or activated for a site after all necessary approvals for that site are in place.3 Where a single global build goes live for all sites at once, sponsors need a way to show that no site could enter data before its own approvals were complete.
During the Trial: Exclusive Control and Incident Reporting
E6(R3) keeps a long-standing principle and states it directly: “The sponsor should not have exclusive control of data captured in data acquisition tools in order to prevent undetectable changes.”3 EMA’s guideline explains that this requirement is not met when data are stored on a central server under the sole control of the sponsor, or of a service provider that is not independent from the sponsor. It also says that before an investigator’s read-only access is removed, the investigator should receive a complete copy of their data including the audit trail, and that where a service provider hosts the data, that copy should not pass through the sponsor.11 Sponsors who move to direct data capture, eCOA and integrated site systems need to recheck this principle for each new data path, because each one can move control of source data away from the investigator.
Section 3.16.1(w) asks for processes to report incidents, including security breaches, that have a significant impact on trial data to relevant parties, including regulators.3 In the EU, EMA’s guideline notes that the Clinical Trials Regulation requires sponsors to report serious breaches, including important data and security breaches, within seven days, and says agreements should state which information vendors must escalate immediately so the sponsor can meet that deadline.11 A vendor that notices a problem and takes a week to tell the sponsor has used up the sponsor’s entire reporting period.
Section 4.3.6 asks for contingency procedures to prevent loss of, or lack of access to, data needed for participant safety, trial decisions or trial outcomes. Section 4.3.3 asks for backup, disaster recovery and security measures that are tested periodically, as appropriate, and lists specific controls to consider: user authentication and password management, firewall settings, antivirus software, security patching, system monitoring and penetration testing.3 For most sponsors, these controls live at the vendor. That makes them a service provider oversight question as much as a data integrity one, which is the subject of Gap 5.
After the Trial: Retention and Decommissioning
Section 4.2.7 says trial data and relevant metadata should be archived so they can be retrieved and read, and protected from unauthorized access and changes throughout the retention period.3 Relevant metadata includes the audit trail. EMA’s guideline says contracts should make sure validation documentation remains available for the full retention period, even if the sponsor stops using the system or the vendor stops supporting it or goes out of business.11
This is where the real losses happen. In the MHRA CRO case mentioned earlier, the IRT audit trail for the trial could not be provided during the inspection because the vendor had decommissioned the system. As a result, the inspectors could not verify the IRT data that had been integrated into the eCRF or the activities performed in IRT.14 The trial’s randomization and supply activity could not be verified by the inspectors. Good practice during the trial cannot repair that later.
A decommissioning question worth asking now. For every system in every study that has closed in the last five years, can you produce the data, the audit trail and the validation documentation today, in a readable form, without the vendor’s live platform? If the answer depends on a vendor contract that has lapsed, or on a system that has been retired, you have found your highest-risk data integrity gap.
Annex 2 Adds New Data Paths
Annex 2 does not change the Annex 1 expectations, but it widens the range of systems they apply to. It says remote data collection (remote visits, digital health technologies, extraction from electronic health records) needs special attention to data security, including cybersecurity and data privacy.7 For real-world data, it asks sponsors to consider the fitness for purpose of the systems and tools used to collect and acquire the data, including their validation status as appropriate.7 Sponsors planning decentralized or pragmatic designs for 2027 should add these systems to their trial-level record now, while there is still time to assess them before first use.
Gap 5: Oversight of Service Providers’ Systems
Most computerized systems in a modern trial are not run by the sponsor. They are run by an EDC vendor, an IRT vendor, an eCOA vendor, a central lab, an imaging core, a safety database provider, and a CRO that may itself subcontract several of those. E6(R3) is clear that none of this moves the responsibility.
What the Text Asks For
The principles state that where activities have been transferred to service providers, responsibility for the conduct of the trial, including the quality and integrity of trial data, stays with the sponsor or investigator.3 Section 3.6 then sets out the sponsor’s duties:
- Document the activities transferred in an agreement. Activities not specifically transferred stay with the sponsor.3
- Assess the suitability of each service provider before selecting it.3
- Have access to relevant information, such as SOPs and performance metrics, for selection and oversight.3
- Oversee important transferred activities, including activities the service provider subcontracts further.3
- Expect service providers to implement appropriate quality management and report incidents that might affect participant safety or trial results.3
Section 3.6.10 adds a helpful point: service provider activities should follow relevant GCP requirements, and this can be met through the provider’s existing quality management processes, even if those were not designed specifically for GCP, as long as they are fit for purpose in the context of the trial.3 That matters for cloud and technology providers whose quality systems follow general software or security standards. The sponsor’s job is to judge fitness for purpose and record that judgment, not to demand a GCP-branded quality manual.
The EU Expectation on Agreements Is Detailed
EMA’s guideline has a full annex on agreements. Among its points:
- Agreements should be signed before services or systems are provided, and should cover subcontracting, the conditions for it, and how the sponsor will oversee subcontracted work.11
- Sponsors should oversee delegated duties, for example by reviewing defined key performance indicators or reconciliations.11
- If a provider will not allow access to important documentation (such as system requirements specifications) or will not support pre-qualification audits or regulatory inspections, “systems from such a service provider should not be used in clinical trials.”11
- System documentation (validation records, vendor SOPs, training records, issue logs) and trial-specific records such as relevant help desk tickets should be retained for the full retention period, with the agreement stating who holds what and how it will be made available.11
- For cloud solutions, the provider should be qualified, and contracts should be detailed and explicit about all relevant ICH E6 topics and legal requirements.11
The US and EU Differ on Audit Reports
One practical difference is worth knowing. FDA’s 2024 guidance says FDA will generally not review the reports of audits that sponsors perform on IT service providers’ systems, consistent with its long-standing policy on access to quality assurance audit results.12 EMA’s guideline takes a different position where the sponsor relies on the vendor’s validation work: inspectors should be given access to the full documentation and to the sponsor’s examination of the vendor, which may mean giving access to the audit report.11 A global sponsor should write its vendor audit reports on the assumption that an EU inspector may read them. Vague conclusions such as “validation adequate” will not satisfy an inspector. Specific records of which requirements and test cases were sampled, and what was found, will.
Where Sponsors Fall Short
The EMA inspectors’ 2024 report lists common critical and major findings by category. Under contracts and agreements, it names a lack of explicit GCP compliance clauses and “Inadequate documentation of third party/vendor qualification and oversight.” Under audit, it lists cases where no audits were conducted of partners or vendors working on behalf of the sponsor. Under data management, it lists “Inadequate data migration and system validation,” the use of spreadsheets for sensitive data collection, and pseudonymized and unblinded data sent without password protection or encryption.13
Two cautions about those numbers. First, the “Computer System” category does not capture every system problem. Many system-related problems are recorded under other categories: data management (27 critical and major findings across all 2024 inspections), essential documents, and contracts and agreements.13 Second, as noted earlier, these inspections were against the pre-R3 rules. They describe where sponsors started, not how they are doing under R3.
From our own work, the recurring oversight gaps look like this:
- Qualification happens once. The vendor is audited at selection and not again, even as its platform, hosting and subcontractors change.
- Subcontractors are invisible. The CRO subcontracts IRT or eCOA to a third party, and the sponsor has no agreement with that third party and no evidence of the CRO’s oversight of it. E6(R3) section 3.6.9 names subcontracted activities explicitly.3
- Oversight has no measures. Governance meetings happen, but there are no agreed indicators for system performance, defect handling or data transfer reconciliation, so there is nothing to show an inspector.
- Defect and help desk information does not flow back. E6(R3) asks for a way for service providers and investigators to report system defects to the sponsor, and for periodic review of cumulative issues to find repeated or systemic ones.3 Help desk data usually stays with the vendor.
- Site systems are not assessed. When electronic health records or other site systems hold source data, E6(R3) asks the sponsor to assess them during site selection, before use.3 The EMA report lists, among essential document findings, a lack of access at the trial site to trial-relevant electronic systems for monitors and inspectors, and source data review based only on uncertified copies of electronic medical records.13
A Sequenced Remediation Plan
None of these gaps needs a large program to fix. They need a clear order of work, owners who span clinical operations, data management, quality and IT, and evidence that is filed in the trial master file as it is produced. The sequence below starts with what an inspector will ask for first and moves toward the work that takes longer.
Build the Trial-Level System Record
For each active study, list the important systems, their owners, hosting, versions, validation status, interfaces, access controls and security measures, as section 3.16.1(x)(i) describes. Add a data flow diagram that shows every transfer from capture to archive. Every later step depends on this record.
Map Critical Data to Systems and Functions
Take the protocol’s critical-to-quality factors and trace them to specific system functions: randomization logic, dosing calculations, endpoint forms, safety reporting paths. This produces the written risk basis for proportionate validation and focused audit trail review.
Test the Evidence for Critical Functions and Interfaces
For each critical function and each interface, confirm there is validation evidence that an inspector could follow, including evidence covering the current vendor release. Fill gaps with targeted testing or documented vendor examination, not a full revalidation.
Put Audit Trail Review and Access Review on a Schedule
Write a study-specific audit trail review plan for critical data. Reconcile system roles against delegation logs, include vendor and privileged accounts, share site user lists with investigators, and document each review in the trial master file.
Rework Service Provider Agreements and Oversight Measures
Check each agreement for subcontracting terms, audit and inspection access, documentation retention through the full retention period, incident escalation timelines that fit the EU seven-day serious breach window, and investigator data access at decommissioning. Agree a small set of oversight measures per provider and review them on a fixed cadence.
Close the Loop on Retention and Decommissioning
For closed studies, confirm that data, audit trails and validation documentation can be produced in readable form without the live platform. For active studies, write the decommissioning steps now, before the vendor contract ends.
Who should own this. These gaps fall between functions. Clinical operations owns vendor relationships, data management owns the study build and audit trail review, quality owns vendor qualification and audits, and IT owns security and access. E6(R3) assigns responsibility to the sponsor as a whole. A single named owner for computerized system compliance at trial level, with authority across those four functions, prevents the most common outcome: each function assuming another has the evidence.
What Not to Do
Two responses to E6(R3) create more work without closing the gaps. The first is rewriting every SOP before looking at a single study. SOP updates matter, but inspectors test evidence in specific trials, and the gaps described here are visible only at study level. The second is revalidating every system in full to be safe. That approach runs against the proportionality E6(R3) asks for, uses effort on low-risk functions, and often still misses the interfaces and protocol-specific configurations where most of the risk is. Start with the trial-level record and the critical data. Everything else follows from those two.
Conclusion
E6(R3) has now been in effect in the EU for more than a year and published as final FDA guidance for about as long, and this spring its principles became a legal requirement in the UK and Canada implemented it. Its computerized system expectations are not vague. They ask for a record of important systems, validation that covers configurations, calculations and interfaces in proportion to risk, audit trails that can be read and are reviewed on a plan, access that matches delegation and is visible to investigators, data that stays retrievable after systems are retired, and oversight that reaches the service providers who run most of the technology. The inspection record from before R3 shows sponsors were already struggling with several of these. The new text makes each of them easier for an inspector to cite.
The sponsors who handle this well treat it as a study-level evidence problem, not a policy-writing exercise. They know which systems each trial depends on, where the critical data flows, and what they would hand an inspector for each link. Sakara Digital works with pharma and biotech organizations on exactly this kind of clinical systems governance, from trial-level system records to vendor oversight models that inspectors can follow. If you are reviewing your E6(R3) readiness for computerized systems and want an independent view of where to start, we are happy to have that conversation.
For Further Reading
For Further Reading
- Computerized System Inventory: Building One You Can Defend
- Data Integrity Findings in Contract Labs: What Sponsors Miss in Oversight
- Data Quality Ownership When Three Vendors Touch One Record
- Single Sign-On and Part 11 Signature Requirements: Getting Both
- Direct Data Capture vs EDC: What Changes for Clinical Data Management
- Clinical Trial Data Reconciliation Between EDC, Safety, and IRT Systems
- CTMS Selection for a Biotech Running Its First Phase 2 Trial
References & Sources
- European Medicines Agency. “ICH E6 Good Clinical Practice: Scientific Guideline.” EMA, page listing E6(R3) effective dates and documents, accessed September 2026. https://www.ema.europa.eu/en/ich-e6-good-clinical-practice-scientific-guideline
- International Council for Harmonisation. “Guideline for Good Clinical Practice E6(R3).” Final version, adopted 6 January 2025. https://database.ich.org/sites/default/files/ICH_E6%28R3%29_Step4_FinalGuideline_2025_0106.pdf
- European Medicines Agency. “ICH E6(R3) Guideline for Good Clinical Practice (GCP), Step 5” (principles and Annex 1), EMA/CHMP/ICH/135/1995. Legal effective date 23 July 2025. https://www.ema.europa.eu/en/documents/scientific-guideline/ich-e6-r3-guideline-good-clinical-practice-gcp-step-5_en.pdf
- U.S. Food and Drug Administration. “E6(R3) Good Clinical Practice; International Council for Harmonisation; Guidance for Industry; Availability.” Federal Register, 9 September 2025 (FR Doc. 2025-17311). https://www.govinfo.gov/content/pkg/FR-2025-09-09/html/2025-17311.htm
- U.S. Food and Drug Administration. “E6(R3) Good Clinical Practice (GCP).” Final guidance, September 2025, Docket FDA-2023-D-1955. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e6r3-good-clinical-practice-gcp
- U.S. Food and Drug Administration. “E6(R3) Good Clinical Practice: Annex 2.” Draft guidance, December 2024, Docket FDA-2024-D-5601. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e6r3-good-clinical-practice-annex-2
- International Council for Harmonisation. “Guideline for Good Clinical Practice E6(R3) Annex 2.” Final version, adopted 3 June 2026. https://database.ich.org/sites/default/files/ICH_E6(R3)_Annex%202_Guideline_Step%204_2026_0603_0.pdf
- Pound, J. and Messer, J. “Clinical Trials Regulations: Six-Month Countdown Begins.” MedRegs blog (MHRA), 28 October 2025. https://medregs.blog.gov.uk/2025/10/28/clinical-trials-regulations-six-month-countdown-begins/
- Health Research Authority. “New UK Clinical Trials Regulations Now in Force.” 28 April 2026. https://www.hra.nhs.uk/about-us/news-updates/new-uk-clinical-trials-regulations-now-in-force/
- Health Canada. “International Council for Harmonisation (ICH): Guidelines” (E6(R3) implementation date 2026/04/01). https://www.canada.ca/en/health-canada/services/drugs-health-products/drug-products/applications-submissions/guidance-documents/international-council-harmonisation/guidelines.html
- European Medicines Agency, GCP Inspectors Working Group. “Guideline on Computerised Systems and Electronic Data in Clinical Trials.” EMA/INS/GCP/112288/2023, 9 March 2023. https://www.ema.europa.eu/en/documents/regulatory-procedural-guideline/guideline-computerised-systems-and-electronic-data-clinical-trials_en.pdf
- U.S. Food and Drug Administration. “Electronic Systems, Electronic Records, and Electronic Signatures in Clinical Investigations: Questions and Answers.” Final guidance, October 2024. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/electronic-systems-electronic-records-and-electronic-signatures-clinical-investigations-questions
- European Medicines Agency. “Annual Report of the Good Clinical Practice (GCP) Inspectors’ Working Group (IWG) 2024.” EMA/INS/GCP/37608/2025, adopted 15 December 2025. https://www.ema.europa.eu/en/documents/report/annual-report-good-clinical-practice-inspectors-working-group-2024_en.pdf
- MHRA GCP Inspectorate. “GCP Inspections Metrics Report: Metrics Period 1st April 2019 to 31st March 2020.” Report dated 29 March 2023 (reissued 11 April 2023). https://assets.publishing.service.gov.uk/media/64357bfe89f19f00133cfb40/GCP_inspection_metrics_2019-2020.pdf
- Naeem, B. “Computer System Validation: GCP.” MHRA Inspectorate blog, 20 April 2017. https://mhrainspectorate.blog.gov.uk/2017/04/20/computer-system-validation-gcp/
- Budwal-Jagait, M. “Is Your eSystem Actually an eCRF (Electronic Case Report Form)?” MHRA Inspectorate blog, 11 May 2021. https://mhrainspectorate.blog.gov.uk/2021/05/11/is-your-esystem-actually-an-ecrf-electronic-case-report-form/








Your perspective matters—join the conversation.