Why DCT Technology Changes the Validation Problem

In a traditional site-based trial, most of the computerized systems that touch trial data are run by people the sponsor can name: the EDC vendor, the IRT vendor, the central lab, the CRO, the site. Validation follows a familiar pattern. Requirements are written, the system is configured for the protocol, testing is executed, and change control governs releases. The data originator is usually a person at a keyboard.

A trial with decentralized elements breaks several of those assumptions at once. The data originator may be an accelerometer on a participant’s wrist. The firmware on that device belongs to the manufacturer. The phone that relays the data runs an operating system the sponsor cannot control, and in a bring-your-own-device design the sponsor did not even choose the phone. The data then travels over a home network to a vendor cloud, where an algorithm may turn raw signals into steps, sleep stages, or heart rate. Only after all of that does a summary value reach the EDC system or the analysis datasets.

FDA says this plainly in its guidance on digital health technologies (DHTs) for remote data acquisition: “While a sponsor may not control the life cycle processes of the DHT, they still should ensure that the DHT remains fit-for-purpose throughout its use in a clinical investigation.”1 That one sentence is the validation problem in DCTs. Responsibility stays with the sponsor while control is spread across several other companies.

The Source Has Moved

The second change is where the source record lives. For data collected directly from participants through DHTs, FDA considers the electronic data in the first durable electronic data repository to which the data are transferred to be the source data, and it does not intend to inspect individual devices for source data when the data and all associated metadata are securely transferred and retained according to the sponsor’s prespecified plan.1 FDA’s 2024 electronic systems Q&A gives examples of such repositories: an EDC system, a clinical investigation site database, or a cloud-based digital platform.2

That has a direct consequence for validation scope. If the source is a vendor’s cloud database, the transfer into that database and the controls inside it are part of the sponsor’s validated data flow, whoever built them. ICH E6(R3) makes the same point from the other direction. Its glossary defines a data acquisition tool as a tool that collects data and metadata from a data originator, and it names machines such as wearables and sensors as possible originators.3

Where the Regulatory Text Stands Now

The rules for DCT technology are spread across several documents, and their status varies by region. As of late October 2026, this is the picture.

DocumentStatusWhat It Covers for DCT Technology
FDA, Digital Health Technologies for Remote Data Acquisition in Clinical InvestigationsFinal guidance, December 20231Fit-for-purpose selection, verification, validation, usability, updates during the trial, source data location
FDA, Conducting Clinical Trials With Decentralized ElementsFinal guidance, September 20244Data management plan with a data flow diagram, local health care providers, Part 11 for electronic systems
FDA, Electronic Systems, Electronic Records, and Electronic Signatures in Clinical Investigations: Q&AFinal guidance, October 20242Risk-based validation, IT service provider agreements, DHT data originators, data transfer, signatures
FDA and OHRP, Use of Electronic Informed Consent: Q&AFinal guidance, December 20165eConsent signatures under Part 11, identity verification, version archiving, inspection materials
ICH E6(R3) Principles and Annex 1Step 4 January 2025; in effect in the EU since 23 July 20256; FDA final guidance September 20257Computerized system validation, data governance, metadata, audit trails, service provider agreements
ICH E6(R3) Annex 2Step 4 on 3 June 20268; EU effective 15 January 20276; FDA’s guidance page still lists the December 2024 draft9Decentralized and pragmatic elements, DHTs, remote consent, sponsor oversight of many service providers
EMA, Guideline on Computerised Systems and Electronic Data in Clinical TrialsPublished March 202310Device metadata, ePRO and BYOD controls, eConsent requirements, vendor agreements
EU Recommendation Paper on Decentralised Elements in Clinical TrialsVersion 02, published 29 October 2025; not legally binding11Roles of service providers, oversight of incoming data, remote consent signatures, data flow
3 June 2026ICH E6(R3) Annex 2, covering decentralized elements and DHTs, reached Step 48
15 Jan 2027Date Annex 2 comes into effect in the EU, per EMA6
75%Higher vertical counts from one activity monitor firmware version than another during slow treadmill walking12

One practical note on the US side. FDA’s DHT and DCT guidances were written before E6(R3) was final, and some of their footnotes still point to E6(R2) or to the draft version of the electronic systems Q&A. The substance lines up well with E6(R3), and US sponsors running global programs will usually find it simplest to design to the stricter of the two texts on any given point.

Five Validation Points From Sensor to Submission

Most validation plans for DCT programs are organized by system: one plan for the eConsent platform, one for the eCOA app, one for the device platform, one for EDC. That structure matches how vendors sell, but it hides the places where data integrity fails, which are usually between systems. A more useful structure follows a single data point from the moment it is created to the moment it appears in an analysis dataset.

Across the FDA, ICH, and EMA texts, five points come up again and again. Each one needs an owner and a piece of evidence.

1

The Measurement

Does the device or app measure the clinical event or characteristic the protocol needs, accurately and consistently, in this trial population and in a home setting? This is fit-for-purpose evaluation: verification, validation, and usability.

2

The Participant Interface

Is every consent, signature, and manual entry attributable to the right person? This covers eConsent identity checks, Part 11 signatures, app access controls, and training.

3

Transfer to the First Durable Repository

Does every record, with its metadata, reach the source repository completely, on time, and with an unambiguous timestamp? This is where data loss and duplication happen.

4

Integration Into EDC and Analysis Data

Do interfaces, mappings, and derivations carry the right values to the right participant and visit, and can every value be traced back to its source?

5

Change Over Time

Have firmware, operating system, app, algorithm, or platform changes during the trial altered the measurement or the data flow, and is each change recorded against each device?

A vendor’s validation package, however good, mostly addresses the internal behavior of its own platform. It rarely addresses point one for your population, point four for your EDC build, or point five for your study timeline. Those are trial-specific, and E6(R3) is explicit that both standard functionality and protocol-specific configurations, including interfaces between systems, should be validated.3

The question an inspector can ask about any endpoint value: Which participant does it belong to, which device and firmware version recorded it, when was it recorded, when was it transferred, which algorithm version produced it, and what has changed since? FDA’s Q&A expects each DHT data element to carry an identifier with its origin, date and time, and participant,2 and the EMA guideline lists device version, device identifiers, firmware version, last calibration, data originator, and event timestamps as metadata that should accompany data from wearables.10 If your data flow cannot answer that question, the gap is in validation, not in monitoring.

Wearables and DHTs: Fit-for-Purpose Evaluation

FDA’s DHT guidance uses “fit-for-purpose” to mean that the level of validation behind a DHT is sufficient to support its use in the trial, including how its data will be interpreted.1 The guidance breaks this into three activities, and it is worth being precise about them because the words overlap with computerized system validation (CSV) in confusing ways.

Verification, Validation, and Usability Mean Specific Things Here

In the FDA guidance, verification confirms that the parameter the DHT measures (acceleration, temperature, pressure) is measured accurately and precisely. Validation confirms that the selected DHT appropriately assesses the clinical event or characteristic, such as step count or heart rate, in the proposed participant population.1 Usability evaluations confirm that trial participants can use the DHT as intended, without errors, in a remote setting.1 FDA also notes that its use of “verification” and “validation” is not meant to match the definitions in the device quality system regulation or its software validation guidance.1

The V3 framework published in npj Digital Medicine in 2020 splits the same territory into three parts: verification, analytical validation, and clinical validation.13 An EFPIA paper on decentralized elements, published in Therapeutic Innovation & Regulatory Science in 2025, reports that current practice shows systematic use of V3 to assess the reliability and accuracy of data from digital tools.14 Whichever vocabulary your team uses, write it down in the validation plan so that quality, clinical, and biostatistics staff are not using the same word for different evidence.

The practical point for leaders is that measurement validation and system validation are two separate bodies of evidence, and a DCT program needs both.

Measurement Evidence

Does It Measure the Right Thing?

Bench and human studies comparing the DHT with a reference measure, in a population like yours, in settings like yours. Owned by clinical science and biostatistics, often drawing on manufacturer studies.

System Evidence

Does the Software Behave as Specified?

Requirements, configuration, functional testing, interface testing, audit trails, access control, change control. Owned by quality and IT, often drawing on vendor documentation.

Usability Evidence

Can Participants Use It at Home?

Evaluations showing the target population can wear, charge, sync, and respond to alerts without errors. Findings feed training and instructions.

Operational Evidence

Does It Keep Working for the Whole Trial?

Wear-time and transmission monitoring, device accountability, replacement procedures, and version tracking through close-out.

Population and Setting Matter More Than the Spec Sheet

FDA gives a clear example of why manufacturer accuracy claims are not enough: an algorithm that counts steps well in healthy participants may not work for participants with Parkinson’s disease who walk with a shuffling gait.1 The guidance lists the kinds of evidence that may be needed: comparisons with reference measurements, evaluation of device placement and activities that could be mistaken for the event of interest, validation of any calibration process, and justification of differences between measurements taken remotely and in the clinic. It notes that a run-in period may be appropriate to ensure reliable data collection.1

Sponsors can rely on verification and validation data that DHT manufacturers or other parties make public, that appear in device labeling, or that are available by right of reference to other FDA submissions, where appropriate.1 In practice, this is where most of the measurement evidence comes from. The sponsor’s job is to read that evidence critically against the protocol: the population, the wear location, the duration, the endpoint definition, and the specific model and version.

Specify the Device, the Version, and the Supporting Technology

FDA recommends that sponsors identify minimum technical and performance specifications for the DHT and, where applicable, the specific product, model, or version that meets them. If new models or versions become available during the trial, the sponsor could consider adding them if they meet those specifications.1 When a protocol allows more than one brand or model to collect the same data, FDA expects confirmation that measurements are consistent across them.1

The guidance also reminds sponsors that DHTs depend on general-purpose computing platforms and networks, and that those supporting technologies must be adequate for the DHT’s function.1 For bring-your-own-device designs, the EMA guideline adds that the sponsor should describe minimum specifications for participants’ devices, consider which operating systems are still supported by the manufacturer with security patches, and ensure data quality and integrity across all accepted models and versions.10 Both regulators expect a sponsor-provided device to be available so that participants without a suitable device are not excluded.1,10

Read the Terms of Service Before the Consent Form Is Written

One item in the FDA guidance is easy to miss and often discovered late. DHTs and the platforms they run on may carry end-user license agreements or terms of service that allow the manufacturer or other parties to access participant data. FDA says the consent process should explain this when it applies, and that sponsors may need to work with manufacturers to modify those terms for the study.1 The EMA guideline makes a similar point and notes that such terms may conflict with GCP and local law, and that in some cases an application may not be suitable for use at all.10 This is a validation question as much as a privacy one, because it determines who else can see or change the data before it reaches the sponsor’s repository.

A fit-for-purpose dossier for each DHT should hold: the protocol function the DHT serves and the endpoint it supports; the model, firmware, and app versions approved for use; the manufacturer and published evidence relied on, with a written assessment of how well it matches the trial population and setting; any additional verification, validation, or usability work done by the sponsor; the data flow to the first durable repository; the missing data plan; and the terms of service review.

eConsent System Validation and Part 11

eConsent is often treated as a document workflow. From a validation standpoint it is closer to a regulated signature system with strict version control, and it is the one DCT technology where a defect can call into question whether a participant was legally enrolled.

The System Supports Consent but Does Not Obtain It

FDA and OHRP’s eConsent Q&A is direct on the most important point: “The investigator cannot delegate authority to obtain informed consent to the electronic system.”5 Whether the process takes place at the site or remotely, the investigator and appropriately delegated staff remain responsible. The system design should make that responsibility visible, with a record of the investigator or delegate who conducted the consent discussion and confirmed the participant’s identity.

ICH E6(R3) Annex 2 adds that when consent is obtained remotely, the investigator should confirm the participant’s identity, for example by checking an official identification document over a video call, and that the verification method and the privacy safeguards should be prespecified.8 Annex 2 also says participants may be offered a paper or in-person option where feasible, and that the protocol should describe the consent modalities used.8

What Part 11 Requires of an eConsent Signature

For FDA-regulated trials, electronic signatures on consent must meet 21 CFR Part 11 to be treated as equivalent to handwritten signatures, and the system must capture the date the participant or legally authorized representative signed.5,15 Part 11 requires that an organization verify a person’s identity before it establishes or certifies that person’s electronic signature, but it does not prescribe a method. The eConsent Q&A gives examples such as official identification and security questions,5 and the 2024 electronic systems Q&A adds strong digital login credentials with multi-factor authentication, or video observation.2

Two details from the 2024 Q&A regularly cause rework. First, signed electronic records must show the printed name of the signer, the date and time the signature was executed, and the meaning of the signature.2 Second, FDA considers a signature drawn with a finger or stylus on a tablet to be a handwritten signature, not an electronic signature, and it must be linked to its electronic record.2 Your validation testing should confirm which type each signature step produces and that the link to the record cannot be broken.

Version Control Is the Highest-Risk Function

The eConsent Q&A asks that all versions of the IRB-approved consent be easy to retrieve, and that inspectors be able to see site-specific versions, IRB submissions, amendments, and every signed consent.5 The EMA guideline goes further on system behavior. Only versions approved by the ethics committee should be enabled, release to sites before approval should be prevented, and the system should prevent the use of obsolete versions.10 When a protocol amendment changes the consent, the system has to manage re-consent for every enrolled participant, and the EU recommendation paper asks sponsors to adapt existing re-consent procedures to electronically signed forms.11

The EMA guideline also sets expectations that are useful as test cases even for US-only programs: signature timestamps that cannot be changed by system settings, any alteration of the signed document invalidating the signature, an immediate downloadable copy for the participant, notifications to the sponsor that contain only non-identifying information, and investigator control over granting and revoking access for monitors, auditors, and inspectors.10 If a participant can withdraw consent inside the system, the withdrawal should generate an alert to the investigator,10 and the EU recommendation paper adds that withdrawal should always be possible outside the system as well.11

eConsent RequirementTest Worth RunningBasis
Only approved versions are presentedAttempt to consent on a version not yet marked approved for that site; confirm it is blockedEMA guideline10
Obsolete versions cannot be usedApprove a new version mid-test; confirm the prior version is no longer selectable and re-consent tasks are createdEMA guideline10; FDA eIC Q&A5
Identity verification is recordedConfirm the method, the person who performed it, and the time are captured for each participantAnnex 28; Part 1115
Signature manifestation is completeCheck printed name, date and time, and meaning on the rendered record and on any PDF copyFDA Q&A2
Time comes from a trusted sourceChange the device clock and time zone; confirm signature times do not changeEMA guideline10; E6(R3)3
Altered documents invalidate signaturesModify a signed record through any available path; confirm the signature shows as invalidEMA guideline10; EU paper11
Participant receives a copyConfirm a protected copy is available right after signing, including linked materialsEMA guideline10
Sponsor sees no identifying dataReview sponsor-role screens and reports for names, contact details, or ID imagesEMA guideline10
Withdrawal reaches the investigatorWithdraw in the app; confirm an alert, and confirm a site-recorded withdrawal outside the app is also handledEMA guideline10; EU paper11

Plan for National Differences Early

Global programs should expect consent rules to vary. The EMA guideline notes that some countries still require a wet-ink signature,10 and the EU recommendation paper keeps a national provisions overview for this reason.11 The EFPIA survey paper reports, for example, that Denmark allows eConsent if there is active dialogue with the participant and did not accept pre-recorded video, and that some sponsors removed eConsent from their trials because it proved hard to put into practice.14 A validated system that cannot be configured for each country’s rules is not fit for purpose in a multinational trial, so country configuration belongs in the requirements, not in a later change request.

Home Health Data Capture and the Device-to-EDC Data Flow

Home health capture covers two quite different kinds of data. One is machine data from wearables and connected devices. The other is human data recorded by home nurses, mobile phlebotomists, and local health care providers who perform protocol procedures away from the site. Both have to reach the trial record with attribution, timing, and metadata intact, and both depend on a documented data flow.

Start With the Data Flow Diagram

FDA’s DCT guidance asks sponsors to include, in a data management plan or other trial documents, the data origin and data flow from all sources to the sponsor, for example as a diagram from creation to final storage, along with the methods used for remote data acquisition and a list of the service providers involved in data collection, handling, and management.4 The EU recommendation paper calls a data flow diagram in the protocol “highly recommended” and asks sponsors to ensure that data acquisition tools are configured and validated for their intended use.11 E6(R3) asks sponsors to keep a record of important computerized systems with their use, functionality, interfaces, validation status, and owner.3

The diagram is not paperwork. It is the scope statement for validation. Every arrow on it is a data transfer that either needs validation or a reconciliation process, because E6(R3) says that validated processes or other appropriate processes such as reconciliation should be in place so that data and metadata keep their integrity when transferred between systems.3

Machine Data: Attribution, Metadata, and Timing

When a DHT sends data automatically with no human step, FDA’s Q&A says the DHT itself should be identified as the data originator, with a data element identifier created automatically, and that the date and time of recording and the participant identifier are part of the metadata to be kept in the durable repository.2 The sponsor should keep a list of authorized data originators available for inspection.2 Transfer to the repository should use a validated process, should happen at the time of recording or as soon as possible afterward, and the transfer date and time should be in the audit trail.2

E6(R3) adds that acquired data should be accompanied by relevant metadata, and that automatic capture of the date and time of entries or transfers should be unambiguous, for example using coordinated universal time (UTC).3 For ePRO, the EMA guideline asks that the timestamp record the time of data entry, not only the time of submission or transmission.10 These sound like small details. In a trial where participants travel across time zones, or where a phone stores readings offline for two days before syncing, they decide whether an endpoint can be computed correctly.

The EMA guideline also describes a sequence that is worth building into the requirements. Data on the participant’s device is the original record until it is transferred to a durable server by a validated procedure, after which the server copy is treated as a certified copy and the device copy can be removed. There should be procedures for failed or interrupted transmission, monitoring that transmission completed, an accountability log linking each device to a participant, and a way to merge data from replacement devices without losing traceability.10

Human Data From Home Visits and Local Providers

FDA’s DCT guidance describes two routes for local health care providers to submit trial data: entering it directly into an eCRF built for that purpose, or sending forms securely to investigators who enter the data and keep the records. Remote trial personnel or local providers who enter data directly into the eCRF should be on the sponsor’s list of authorized data originators.4 The guidance confirms that electronic systems used to produce and process required trial records are subject to Part 11, while live video and audio visits are not themselves electronic records but must still be documented.4

ICH E6(R3) Annex 2 says investigator oversight of these activities should be proportionate to the importance of the data and the risk to participants, and for procedures that are part of usual clinical practice it asks for arrangements covering how data are shared, how records are kept (for example as certified copies), and how data integrity is maintained.8 When the sponsor arranges a service such as home nursing that falls under the investigator’s responsibility, Annex 2 says the sponsor should ensure the provider is suitable, while the investigator keeps the decision on whether that provider is appropriate and oversees its work.8

Safety Data Has to Reach the Investigator in Usable Form

Continuous data creates a review problem. The EU recommendation paper notes that wearables increase the volume of incoming data, which can strain investigators, and asks for a clear, risk-based procedure for how often investigators review it, with alerts for data that may signal a serious adverse event.11 Annex 2 asks sponsors to provide DHT safety information to investigators in a way that is relevant, meaningful, and manageable.8 FDA says it generally does not expect investigators to monitor DHT data continuously, but the safety monitoring plan should state how often continuous data will be reviewed and how abnormal findings are handled.1 Alert rules, thresholds, and routing are configured functions and need testing like any other critical function.

Investigator Access and Sponsor Control

One EU expectation shapes architecture decisions. The EMA guideline says the sponsor should not have exclusive control of data entered in a computerized system at any point, and that this requirement is not met if data are stored only on a server controlled by the sponsor or by a service provider that is not independent of the sponsor. The investigator should be able to download a contemporaneous certified copy of the data.10 The EU recommendation paper also asks sponsors to ensure the investigator has continuous and complete access to source data, whether generated on-site or off-site.11 If a device vendor sends data only to the sponsor, that design needs to change before EU sites are activated.

Data Flow StepTypical FailureEvidence That Shows Control
Device to phone or hub (local sync)Readings lost when the battery dies or the app is closed; duplicates after re-pairingStore-and-forward testing, duplicate detection rules, device-to-participant pairing checks
Phone or hub to vendor cloud (often the first durable repository)Partial uploads, device clock used instead of a trusted clock, records attached to the wrong participantValidated transfer, UTC timestamps for recording and transfer, transmission completeness monitoring
Processing inside the vendor platformAlgorithm version changes the derived value without noticeAlgorithm and firmware version stored with each derived record; release notes reviewed under change control
Vendor platform to EDC or data warehouseMapping, unit, or rounding errors; visits matched incorrectly; silent rejectsInterface validation, record-count and value reconciliation, error queues reviewed by data management
Home visit and local provider data to eCRFEntries by unlisted users; missing source documents; late entryAuthorized data originator list, role-based access, certified copy process, timeliness metrics
Trial data to analysis datasetsEndpoint derived from the wrong data window or wrong sourcePrespecified source and derivation in the statistical analysis plan; programming validation; traceability to source

On what reaches FDA, the DHT guidance says sponsors should discuss with the review division which DHT data will be submitted (complete, summary, sample, or abnormal data), that DHT data must be kept in human readable form, and that FDA generally does not intend to request raw machine data such as voltages during an inspection.1 That helps scope retention, but it does not reduce the need to retain the metadata that explains how each submitted value was produced.

Firmware and Software Updates Mid-Study

Of all DCT technology risks, mid-study change is among the most likely to damage an endpoint without anyone noticing. A traditional EDC release goes through the sponsor’s change control. A firmware update to a wearable, an operating system update to a participant’s phone, or a new version of a vendor’s step-count algorithm may go out on the manufacturer’s schedule, to some devices and not others, partway through enrollment.

The Evidence That Firmware Changes Measurements

This is not a theoretical concern. A 2014 study in Medicine & Science in Sports & Exercise compared firmware versions of a research-grade activity monitor and found that vertical counts during treadmill walking at 1.5 mph were significantly greater, by 75%, with one firmware version than with another. The authors recommended that the manufacturer run both human and mechanical shaker testing before releasing new firmware, to confirm outputs stay comparable.12

A 2025 editorial in Sleep raises the same concern for consumer devices: the manufacturer can change an algorithm with little or no warning, possibly halfway through a study, in ways that materially affect results. The authors also note that Oura added a feature to freeze firmware for devices used in research studies and gives months of notice before major software updates.16 That kind of feature, where it exists, is worth asking for in vendor selection.

What Regulators Expect

FDA’s DHT guidance has a section on exactly this. Sponsors should plan for unanticipated changes during the trial, such as updates needed to fix a security concern or a DHT being discontinued, whether devices are sponsor-provided or participant-owned. “Sponsors should keep a record of the timing and nature of any updates for each DHT.”1 When a DHT or its supporting platform is updated, the sponsor should ensure the DHT remains fit-for-purpose and that the earlier verification and validation still apply. Where measurements may be affected, it may be necessary to validate the measurements after the update, compare data from before and after, run sensitivity analyses, and specify how differences will be handled in the analysis before unblinding. The guidance closes the section bluntly: “Significant changes in the measurement after updates may invalidate the results from a clinical investigation.”1

The same guidance asks the protocol and statistical analysis plan to define events that could affect data collection and analysis, including changes in technology during the study, and to describe how they will be handled.1 ICH E6(R3) states the general rule for all computerized systems: “Subsequent changes to the system should be validated based on risk and should consider both previously collected and new data in line with change control procedures.”3 FDA’s Q&A lists software upgrades, security and performance patches, and component replacements among the changes to evaluate, and adds: “Changes should not adversely affect the traceability, authenticity, or integrity of new or existing data.”2

For bring-your-own-device designs, the EMA guideline acknowledges the limit of sponsor control directly: “The sponsor has no control over the implementation of updates to the operating system or over the applications on the trial participant’s device.” It asks sponsors to take this into account in their risk evaluation and validation activities.10

Freezing versions is not a complete answer. E6(R3) expects security controls, including security patching, to be maintained for computerized systems,3 and FDA expects safeguards against cybersecurity threats to DHTs.1 A sponsor that blocks every update to protect an endpoint can create a security problem instead. The workable approach is a decision process: classify each update, assess its effect on measurement before deployment where possible, and record exactly when each device received it.

A Workable Update Process

Update TypeExamplesTypical Response
No plausible effect on measurement or data flowDisplay text, language files, help screensDocument under change control; record version per device; no measurement testing
Possible effect on data flowSync logic, phone operating system, app permissions, cloud platform releaseRegression testing of transfer, timestamps, and integration before deployment; monitor completeness after
Possible effect on measurementSensor firmware, signal filtering, sampling rate, derived-metric algorithmBench or reference-dataset comparison before deployment where feasible; per-device effective date; pre and post comparison; sensitivity analysis planned before unblinding
Required for securityPatches for a known vulnerabilityDeploy on a risk basis even if measurement testing must follow; record timing; assess impact on existing and new data afterward
Device or model discontinuedManufacturer ends supply mid-trialAssess replacement against minimum specifications; confirm measurement consistency; plan data merging and analysis handling

Two practical points make this process work. First, the vendor contract must give you advance notice and release notes for firmware, app, and algorithm changes, and ideally the option to hold a version for the study. Second, the version history has to be data, not a memo. If firmware and algorithm versions travel with each record, biostatistics can run a before-and-after analysis in days. If they live in an email thread, the analysis may not be possible at all.

When Vendors Own the Platform: What Stays With the Sponsor

Almost every DCT program runs on platforms the sponsor does not own: an eConsent vendor, a device and analytics vendor, an eCOA vendor, a home nursing network, a direct-to-patient logistics provider, and often a CRO coordinating them. Regulators are consistent that this changes who does the work but not who is responsible.

The Sponsor’s Validation Responsibility Does Not Transfer

EMA’s notice to sponsors on validation and qualification of computerized systems, first issued in April 2020 after recurring inspection findings and updated in October 2026 to reference E6(R3), states: “The sponsor is ultimately responsible for the validation of the computerised system and for providing adequate documented evidence on the validation process.”17 Sponsors must be able to give EU inspectors access to qualification and validation documentation whoever performed the work. A sponsor may rely on a vendor’s documentation if it has assessed that documentation as adequate, and may need to add its own qualification and validation work based on a documented risk assessment.17 The notice warns that failure to demonstrate a system’s validated state may, depending on how critical the data are, lead GCP inspectors to recommend that the data not be used in a marketing authorization application.17

FDA’s Q&A takes a similar line. When an IT service provider performs validation, the sponsor may review the provider’s development processes, validation processes, functional testing, and change control procedures and logs to judge whether the system is fit for purpose.2 Sponsors should have a written agreement describing scope, roles, and quality responsibilities, plus a plan for sponsor access to data for the full retention period, and should be able to show FDA both the agreements and records of their oversight during the trial.2 FDA says it generally will not review the sponsor’s audit reports of IT service providers, may inspect those providers in some situations, and expects the sponsor to have access to all study records they hold.2

Agreements and Oversight Under E6(R3)

E6(R3) asks that agreements with service providers be documented before activities start and updated when activities change significantly, and that they provide for monitoring, audits, inspections, and direct access to source records and facilities, including those of service providers.3 For sponsor-deployed systems, it asks for a process through which service providers and investigators can report system defects to the sponsor.3 Annex 2 observes that sponsor oversight becomes more complex as the number of service providers grows, and asks for clear lines of communication and quality control measures matched to the trial’s critical-to-quality factors.8 The EU recommendation paper adds that any task delegated to a service provider should be in a written agreement with the party responsible for it, and that when the sponsor contracts a provider for tasks under the investigator’s responsibility, the investigator should be able to agree or not.11

The EFPIA paper is candid about the operational side: a single global vendor generally cannot provide local DCT services in every country, and sponsors may need to support vendor systems with fast-response help desks.14 That reality usually means more vendors, more interfaces, and more agreements, which makes a clear responsibility split essential.

ActivityPlatform VendorSponsorInvestigator or Site
Core platform validation and release testingPerforms and documentsAssesses adequacy; keeps access for inspectorsNone
Trial-specific configuration and interfacesBuilds; supports testingDefines requirements; approves testing; acceptsReviews site-facing workflows
Measurement fit-for-purposeProvides manufacturer evidenceAssesses against population and endpoint; fills gapsRaises usability issues
Firmware, app, and algorithm changesNotifies in advance; provides release notesClassifies and assesses impact; approves deploymentReports participant-facing problems
Consent processHosts system; maintains audit trailValidates configuration and versionsObtains consent; verifies identity; controls access
Safety alerts from DHT dataGenerates and routes per configurationDefines rules; tests routingReviews and acts per safety monitoring plan
Data retention and close-outRetains and returns data with metadataEnsures access for full retention periodKeeps or can download independent copy

Contract Terms That Make Oversight Possible

In our work, oversight problems in DCT programs often start with contracts signed before quality and data management were consulted. The terms below follow directly from the FDA, ICH, and EMA expectations above.

  • Advance notice of changes: written notice and release notes before any firmware, app, algorithm, or platform release, with a defined lead time and an option to hold a version for the study where the technology allows it.
  • Access to validation evidence: the right to review development, testing, and change control records, and to make them available to inspectors, including EU inspectors.
  • Audit and inspection rights: sponsor audit rights and cooperation with regulatory inspections, extending to subcontractors such as cloud hosts and device manufacturers.
  • Data and metadata access: access throughout the retention period, in a documented export format that includes audit trails, device identifiers, and firmware and algorithm versions.
  • Investigator copies: a way for investigators to hold or download an independent, contemporaneous copy of their participants’ data.
  • Defect and incident reporting: defined timelines for reporting defects, data loss, security incidents, and anything that could affect participant safety or trial results.
  • Terms of service alignment: confirmation of what participant data the vendor or device maker can access or reuse, with changes for the study where needed.
  • Close-out: procedures to end data collection, revoke access, return or archive data with metadata, and decommission accounts.

A Validation Plan for a DCT Program

Pulling the pieces together, a DCT validation plan can follow seven steps. The order matters: teams that start with vendor testing before they have a data flow diagram and a criticality ranking tend to test the wrong things thoroughly.

1

Build the System Inventory and Data Flow Diagram

List every system, device, app, and service provider that touches trial data, with interfaces, owners, and validation status, as E6(R3) asks.3 Draw the data flow from creation to final storage and mark the first durable repository for each data type.4

2

Rank Data and Functions by Criticality

E6(R3) singles out randomization, dosing, and collection of endpoint data as critical functionality.3 Add consent, safety alerts, and any DHT value used to adjust treatment, where FDA flags the risk of erroneous measurements leading to inappropriate treatment.1

3

Assess Vendor and Manufacturer Evidence

For each system, assess the vendor’s validation documentation and record what you rely on and why. For each DHT, assess the measurement evidence against your population, setting, and endpoint. Record the gaps.

4

Test What Is Specific to Your Trial

Test protocol configuration, eConsent versions and country rules, alert rules, interfaces to EDC, timestamps, and data originator attribution. Use the five validation points to make sure every arrow on the data flow diagram has evidence behind it.

5

Pilot Under Real Conditions

A run-in period or small pilot in the target population shows wear-time, sync, and usability problems before they affect endpoint data. FDA notes a run-in may be appropriate to ensure reliable data collection.1

6

Monitor During the Trial

Track transmission completeness, wear-time, missing data, alert handling, and device and version history. Plan audit trail and metadata review as a risk-based activity, as E6(R3) requires,3 and route every vendor release through change control.

7

Close Out Cleanly

End data collection and transmission, revoke access, and archive data and metadata so they remain retrievable and readable for the retention period.1,3 Confirm investigators hold their copies before vendor accounts are closed.

Signs a DCT program is ready for inspection: a current data flow diagram that matches the systems in use; a fit-for-purpose dossier for each DHT; documented assessments of every vendor’s validation evidence; eConsent version history that ties every signature to an approved version; a list of authorized data originators; device, firmware, and algorithm versions stored with the data; reconciliation records for each interface; and a change log showing how each mid-study update was assessed.

Gaps we often see in DCT validation work:

  • The vendor’s validation summary is filed, but no one has recorded an assessment of whether it is adequate for this trial.
  • The EDC integration was tested with clean sample data, not with late, duplicated, or out-of-window records.
  • Firmware and algorithm versions are known to the vendor but not stored with the data the sponsor receives.
  • Consent re-versioning after an amendment was never tested end to end.
  • Safety alert rules were configured by the vendor and never reviewed by the medical monitor.
  • EU sites were activated on a design where only the sponsor receives the data.

Conclusion

Decentralized trials have not changed what good validation is for. It still exists to show that trial data are reliable and that participants are protected. What has changed is the shape of the system being validated. It now runs from a sensor in someone’s home, through a phone, through one or more vendor platforms, into EDC and the analysis datasets, and it keeps changing while the trial runs. FDA’s DHT guidance, ICH E6(R3) and its Annex 2, and the EMA computerized systems guideline all point the same way: the sponsor cannot hand that responsibility to the companies that own the platforms, but it can organize its oversight around a small number of points where evidence matters most. Measurement, attribution, transfer, integration, and change are those points. A program that has evidence for each one, and contracts that make that evidence available, is in a strong position with inspectors and with its own biostatisticians.

Sakara Digital works with pharma and biotech organizations planning and running trials with decentralized elements, from vendor selection and data flow design through validation and inspection readiness. If you are building a DCT program and want an independent view of where your validation evidence is strong and where it depends on assumptions about your vendors, we are happy to have that conversation.

For Further Reading