Regulations and compliance · Practical guide

eCOA offline synchronisation: did every response arrive correctly?

With offline eCOA, a response can exist on a local device while the central system shows it as missing. Repeated transmission attempts can also create more messages than responses. Test capture, transfer and assignment separately, and handle failures in a way that preserves the origin of the data.

Table of contents

At a glance

How do you test offline synchronisation of eCOA data?

Compare uniquely identified local responses with the confirmed target records. Preserve the original recording time, content and relevant change history. Test interrupted transfers, retries, delayed uploads and device changes against defined outcomes. Message counts alone do not demonstrate success. Assess the usability of a response and clinical suitability of the instrument separately from technical synchronisation.

  • Data entry time and upload time address different questions.
  • A verified receipt requires content and identity attribution, not merely message count.
  • Offline tests do not replace the assessment of clinical instrument suitability.

Offline function and professional suitability are separate issues

eCOA refers to electronic procedures for capturing clinical assessments. In ePRO, participants enter their responses directly. A functional offline synchronization does not establish that a questionnaire, its translation, or the selected data collection timing is clinically appropriate. The EMA explicitly distinguishes this clinical suitability from the technical aspects of its corresponding Annex.

Sources: [1] [3]

Recorded, stored, and transferred are different states

Define when a response is considered locally stored and how successful receipt in the intended server inventory is recognized. A display of "sent" without confirmed receipt creates a gap in traceability. Similarly, a later upload timestamp must not substitute the actual entry timestamp. The EMA expects, in particular, that the time interval between local entry and permanent storage be short and clearly justifiable.

The following internal state matrix helps align functional and IT departments on common definitions. It does not specify a particular software architecture. Where data originates locally, the transition to an appropriate central repository must be controlled, prior to the local data being retired according to the established procedure.

Status of an offline recorded response
StateRequired differentiationAssessment question
Inputted Entry is not yet definitively and permanently stored. What occurs immediately after an app termination?
Locally stored Data capture timestamp and origin are locally available. Is the dataset retained along with relevant history after a restart?
Transfer attempt attempted A study may fail or remain without notification. How are interruptions and missing verification detected?
Receipt confirmed The target set has taken over the data set. Is identity, content, and required time stamp correct?
Retry transmission The same local dataset may reappear. Will a duplicate be detected without losing a genuine second response?

Sources: [1]

The right clock answers the right question

For interpretation, data capture, subsequent modification, and upload may each be relevant. Specify the time source and time zone to be used, and define how local time shifts will be visible. In the case of BYOD devices, the significance of a user-modifiable clock must not be assumed implicitly.

Distinguish a technically delayed submission from a professionally delayed response. If the inspection protocol specifies a data capture window, a response recorded offline within that window may be evaluated differently from one completed and submitted later. The usability of data ultimately depends on the study protocol requirements and documented generation timeline, not solely on a server timestamp.

Sources: [1] [3]

Example: A file upload is missing, another one appears twice

Example: The predefined data capture window runs from 18:00 to 20:00 hours in the documented time zone. Three responses, A, B, and C, are locally recorded at 18:10, 18:20, and 20:05, respectively. In this case: A and B fall within the window, while C falls outside. The assignment of response identifiers is independently defined regardless of the transmission attempt.

Following a connection failure, A is transmitted twice, B is transmitted zero times initially, and C is transmitted once. The three received messages therefore do not represent three fully received responses: prior to the retransmission of B, only two unambiguous data sets are available. The test must clearly demonstrate this distinction, without inadvertently discarding C due to its delayed capture.

Defined expected matrix for the synchronization test
AnswerLocal sourceExample transferExpected result
A 18:10; Identifier A; unchanged content Two identical transmission tests are carried out. A clear answer A; repetition detectable, no duplicate analysis line.
B 18:20; Identifier B; unchanged content First attempt missing; resupply scheduled at 22:00. Detect the missing receipt and retain the 18:20 capture time when it arrives.
C 20:05; Identifier C; unchanged content Successfully transferred once. Dataset received; flag if outside defined window.
Reconciliation Three unambiguous local responses First three messages, but only A and C; later B Before delivery 2 out of 3 unique responses, thereafter 3 out of 3; no mere message count.

Sources: [1] [3]

Corrections and conflicts require a defined pathway

A subsequent upload may be identical or may contain content that has been modified during the interim period. Such cases must be distinguished. A blanket overwrite with the most recently received value may obscure a relevant prior state; a blanket rejection may result in the loss of a legitimate correction. Define how conflicts are detected and resolved on a professional basis.

The EMA does not generally exclude traceable participant corrections. It outlines conditions such as timely processing, justification, and reliable contemporaneous sources. For data stored locally, it is also essential that changes prior to upload remain traceable. The specific user interface should not prompt participants to enter a new response, even when a correction is intended.

Sources: [1] [3]

Device changes must not interrupt the data chain.

Plan a scenario for a lost, damaged, or replaced device. Which data has already been confirmed and transferred? Which data may still reside only locally? How is the replacement device assigned to the same user without accessing data from another user? A successful login alone does not address these questions.

For private devices, supported operating systems, app versions, settings, and assistance features are included. The EMA considers BYOD as a distinct context and provides alternatives for participants who are unable or unwilling to use such devices. Technical support must ensure data confidentiality and must not cause uncontrolled modifications to responses.

Sources: [1]

Outsourced monitoring requires a designated responsible person.

Define who identifies missing transfers, when an alert calls for action and who contacts the participant or trial site. Timelines depend on data purpose, protocol and participants' interests. The EMA describes early transfer according to a defined procedure and timetable. It does not prescribe one upload frequency for all trials.

The inspection site requires timely access to data and relevant metadata pertinent to its responsibilities. At the same time, blinding and confidentiality must be protected. A central dashboard is therefore only beneficial when the display scope, responsibilities, and escalation procedures align with the study requirements. Document the handling of a non-reconstructible response as a result; do not invent a substitute value.

Sources: [1] [3]

Further reading:

Checklist for a Robust Offline Verification

Review normal workflows and failure scenarios that impact the specific study process. A successfully transferred sample response neither captures repetitions nor device loss nor inconsistent versions.

  • Data capture, storage, and transmission timestamps are clearly defined.
  • Local records and relevant change history are retained until appropriate transfer.
  • Interrupted transfer, missing confirmation and repetition have been verified.
  • True second response, identical duplicate and content conflict are distinguished.
  • Data capture windows are evaluated based on the professionally appropriate time reference.
  • Device changes, account assignments, and data loss require a predefined exception procedure.
  • The trial site and responsible data owners receive appropriate notifications and access.
  • The clinical suitability of the device is evaluated separately from the technical transfer.

Sources: [1]

Project experience and collaboration

Test data capture and transmission under difficult conditions

With study owners and IT, we run an offline workflow using controlled test data. We examine capture time, device state, reconnection and confirmation that the data reached the target. One successful transmission cannot show what happens after interruption or multiple retries. Different devices and long outages may require additional cases. We agree how to detect and handle missing transmissions, taking account of the link to the individual and study context as well as technical processing.

Frequently asked questions about eCOA offline synchronisation

Is the upload time equivalent to the capture time?

No. A response may be locally stored for an extended period before being transmitted. The EMA requires a clear and traceable time reference for both data capture and permanent storage. Maintain the significance of both time points to ensure that technical delays are not erroneously interpreted as delayed responses.

Sources: [1]

Can the app delete local responses after transmission?

Not solely based on the initiation of a transmission test. The EMA describes a controlled transfer into a central inventory, including the possible removal of local data following an appropriate validated transfer. Therefore, their procedure must specify how success and completeness are verified prior to the retirement of the local source.

Sources: [1]

How do we verify that a response is duplicated?

Compare the unique professional response identity and content, not merely the number of received messages. A retry attempt may still return the same dataset. A truly separate data collection must not be treated as a duplicate and should be distinguished even in the ideal case. Define this distinction in the planned scenario.

Sources: [1] [3]

Are late-uploaded responses automatically invalid?

No. Separate the time of response from the time of transmission. Professional suitability is determined by the inspection protocol, data collection rules, and documented workflow. A late-entered response may require a different evaluation than a timely recorded response with delayed upload.

Sources: [1]

Can participants correct previous ePRO responses?

The EMA prohibits uncontrolled changes, but requires a controlled, timely, and justified approach with feasible source documentation. Both original content and modifications must be traceable. A silent overwrite of a locally stored response does not meet this requirement.

Sources: [1]

What is important when switching to a backup device?

Verify the most recent confirmed inventory receipt, potential locally stored data, and the unambiguous assignment of the replacement device. Then assess data consolidation and access controls. A new device identifier must neither generate a second participant identity nor inadvertently expose foreign responses.

Sources: [1]

Must each participant use their personal smartphone?

No. The EMA addresses BYOD with additional requirements and takes into account participants who are unable or unwilling to use their own devices. Plan an appropriate alternative and support both approaches. The decision should not unnecessarily impair participation or data quality.

Sources: [1]

How often must synchronization occur?

The EMA does not specify a fixed interval for each study. However, for locally stored ePRO responses, it expects early transfer to a permanent server via a validated, predefined procedure at scheduled times (A5.1.1.2). The time between data capture and upload should be short and clearly justified (6.2.1). Define rhythm and monitoring considerations in alignment with the protocol, data purpose, and participant considerations, and also evaluate extended interruptions and failed uploads.

Sources: [1]

Does a successful technique test demonstrate the suitability of the questionnaire?

No. In the eCOA annex, the EMA distinguishes technical assessment from clinical validation and the suitability of the instrument for the intended assessment. Data transfer, display and operation need to work. Assess separately whether the instrument captures what is needed for the clinical question.

Sources: [1]

Which data should the trial site see when an upload fails?

The data and relevant metadata required for accountability must be provided in a timely manner, maintaining blinding and confidentiality. This may include the distinction between no response and pending data transfer. The scope of data visibility must align with the study requirements and must not permit indiscriminate, unbounded access.

Sources: [1] [3]

Sources and review date

Sources checked on 7 October 2026. ICH E6(R3) Principles and Annex 1 have applied in the EU since 23 July 2025. The final Annex 2 has been published, with an EU effective date of 15 January 2027. The 2023 EMA computerised systems guideline explains GCP expectations and is not an EU regulation. This article covers clinical trials of human medicines. The examples are our working materials. Applicable regional law and trial documents remain decisive.

  1. EMA: Computerised systems and electronic data in clinical trialsFinal Guideline dated 9 March 2023. Sections 6.1.2 and 6.2.1; Annex 5: A5.1, A5.1.1.1–A5.1.1.9 and A5.1.3.
  2. EMA: ICH E6 Good Clinical Practice – current versionStatus page, accessed on 7 October 2026. Principles and Annex 1: EU-effective from 23 July 2025; Annex 2: EU-effective from 15 January 2027.
  3. EMA/ICH: E6(R3), Principles and Annex 1, Step 5Principles and Annex 1 (Step 5), EU-effective from 23 July 2025. Sections 2.12.3–2.12.4 and 4.2.1–4.2.5.

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Purposefully verify offline data paths

We support you in translating professional collection rules into clear synchronisation requirements, error cases, and traceable system justifications.

  • Computer system validation →

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

Send a project enquiry Customer testimonials →