Novimed

The Novimed Journal · Clinical research

eCOA Translation and Migration: Protecting Meaning on the Device

By NovimedGroupUpdated 11 min read
Editorial illustration comparing a paper questionnaire with its electronic implementation
Approved wording and the final device experience need to be reviewed together. Original AI-assisted editorial illustration.

eCOA translation and migration should preserve the approved assessment’s meaning in its final electronic form. Reviewing the language file is only one part of that task. Teams also need evidence that the correct wording, response options and instructions appear and behave as intended in the relevant implementation.

Distinguish translation from electronic implementation

Translation produces an approved language version. Migration carries assessment content into an electronic implementation, sometimes with adaptations that need explicit agreement. The two processes interact, but their approvals answer different questions. A linguist can approve a sentence without having seen the screen on which it will appear; a developer can successfully load a file that contains an obsolete translation.

At kickoff, identify the authoritative source and target files, the eCOA provider, the instrument owner and the people responsible for language and technical review. Agree how string identifiers map to items and screens. Establish whether there are existing migration instructions or approved electronic versions. This prevents a team from inventing layout or wording conventions that conflict with the instrument’s requirements.

Connect permissions to the exact implementation

NovimedGroup’s November 11 discussion highlights the operational chain between instrument selection and the version a participant completes. Start that chain with a permission record identifying the measure, owner, source edition, authorized translations and intended mode. Confirm the project’s actual conditions with the responsible owner rather than assuming that possession of a translation authorizes adaptation, reproduction or electronic use.

An editorial handoff register can connect each approved locale file to its release identifier, the implementation package and the person accepting it. Where files are replaced, keep the earlier relationship visible. If the technology provider receives a new language package, record whether it supersedes every locale or only selected items. A clear handoff reduces the chance that a correct new screen is paired with an older instruction elsewhere in the same workflow.

Define the review configuration

A review record should name the instrument version, locale, device or responsive configuration and software build. “French screenshots approved” is not enough if multiple countries or builds are in scope. Make the coverage explicit, including conditional screens, help text and any administration-specific instructions. Confirm what is supplied as a screenshot and what requires an interactive session.

The configuration also affects how a finding is reproduced. A line-wrap problem seen at one viewport may not appear at another. A right-to-left issue may involve both translated text and embedded numbers. Record enough detail for the implementation team to find the problem without relying on an email description or the reviewer’s memory.

A practical reading lens

  1. Define the concept

  2. Review the evidence

  3. Record the decision

An editorial reminder, not a substitute for the instrument-specific methodology or required approvals.

Compare against approved wording

Review every displayed assessment element against the controlled language version. Look for omissions, unintended substitutions, altered punctuation where meaningful, missing qualifiers and inconsistent response options. Preserve the distinction between a wording problem already present in the approved file and a migration error introduced during implementation.

If the approved wording appears too long for a field, do not shorten it locally to make the screen fit. Raise an issue that describes the constraint and possible options. A layout adjustment may solve the problem without changing language. Where a wording adaptation is needed, obtain the appropriate linguistic and instrument-owner review and assess its downstream consequences.

Examine the response task as a whole

An isolated string does not show how a patient encounters an item. Read instructions, item and response options together. Check whether all choices are visible, their order is preserved and the relationship between labels and controls is clear. A visually prominent navigation button should not be mistaken for a response option.

Consider a hypothetical scale where the source presents low-to-high severity. If the implementation reverses the visual order in one locale without the required approval, matching text alone will not establish fidelity. Likewise, an instruction displayed only on the previous screen may be missed when the workflow changes. These are reasons to inspect the implemented task rather than a list of translated strings.

Review eCOA metatext in its actual context

Doraid Matani’s November discussion draws attention to interface language outside the questionnaire itself. This metatext includes navigation labels, completion messages and instructions for using the application. A correctly translated assessment can still be difficult to complete if a participant cannot distinguish saving a response from submitting it. Include these strings in the review scope while keeping their ownership and approval process distinct from licensed instrument wording.

ISPOR 2026 abstract PCR161, presented in Philadelphia in May, describes glossary development across 60 commonly requested languages and subsequent use in 10 studies. The authors report contextual adjustments during on-device review in 30% of languages, followed by a reported 1% metatext change rate in later studies. These are provider-reported operational observations, not a controlled comparison or proof of improved retention. The abstract does not establish that the two percentages share an identical denominator or justify a universal reduction estimate.

For a practical review, connect each glossary entry to its function and screen context, rather than storing an isolated word pair. Record the locale, source definition, approved translation, relevant control, context notes and version. A label used to leave a screen may need a different explanation from a label that permanently submits an assessment. Reuse is appropriate only when the intended action and context match; a glossary should not force different functions into one translation.

Test the path around the questionnaire

In a hypothetical participant journey, someone completes a diary, sees a translated confirmation and closes the application. If that confirmation refers only to a local save, the participant may reasonably believe the task is finished even though a separate submission remains. Review the full sequence with the implementation team: the prompt, control label, resulting state and message. Confirm what the application actually does before deciding whether the language is misleading.

Include relevant scripts, display sizes and accessibility needs in the agreed review plan. Check whether longer text obscures controls, whether right-to-left layouts preserve the intended relationships, and whether a reminder refers unambiguously to the pending task. Qualified language review, usability work with intended users and functional testing contribute different evidence. None should be described as a replacement for the others.

Feed verified contextual corrections back into the controlled glossary with the reason and applicable scope. Before applying them elsewhere, identify where the same string is used and whether the function is truly equivalent. Retain exceptions where justified. This turns a screen finding into reusable knowledge without assuming that approval in one study covers every device, population or workflow.

Plan multilingual eCOA use in decentralized trials

Doraid Matani’s November discussion extends the language question to participation at home. For eCOA teams, the relevant scope includes the assessment, onboarding instructions, reminders and the route to assistance. A participant should not need an unrecorded explanation from a family member to understand the intended task. Remote delivery makes the relationship between approved wording, usable technology and available support especially important.

FDA’s September 2024 guidance on trials with decentralized elements addresses training for applicable remote visits and the suitability and availability of digital health technologies. It recommends sponsor-provided devices and telecommunications support where needed to avoid exclusion for lack of access. These recommendations provide context for planning remote participation; they do not certify a particular translation or prescribe the localization workflow below.

Build a locale-specific journey with the implementation team: receiving an invitation, accessing the correct language version, learning the task, completing the assessment and obtaining support. Check who owns each message and which version is approved. Keep eConsent review, safety communication and clinical escalation under their responsible teams; approval of questionnaire wording does not approve those separate materials. Translation alone also cannot resolve an unsuitable device or unavailable connection.

Check the handoff from remote task to language support

In a hypothetical home-use review, a participant can read the diary but encounters a connection error followed by help available only in another language. Record the displayed message, locale, build, attempted action and actual support route. Ask whether the participant can distinguish a saved entry from one still waiting to transmit, and whether the support instructions match the implemented process. Functional testers must verify the underlying behavior; language reviewers assess how that behavior is communicated.

Agree how assistance will be documented without coaching the participant’s response. A technical explanation of how to open a screen is different from suggesting what a symptom item means or which answer to choose. Review translated support scripts alongside the participant-facing instructions, identify unresolved gaps and assign an owner before release. Repeat the relevant checks when instructions, platform behavior or support arrangements change. These practical checks complement the study’s broader oversight and do not guarantee recruitment, retention or equivalent measurement.

Know what screenshots cannot establish

Screenshots are useful evidence of displayed content and layout at a specific state. They cannot demonstrate every branch, validation rule, navigation action or saved-data behaviour. Interactive user acceptance testing and other technical checks remain necessary where those functions are in scope. Language review should feed into that work instead of being presented as a substitute.

C-Path’s eCOA resources include recommendations on user acceptance testing and electronic implementation. They are useful starting points for assigning responsibilities across the sponsor and technology provider. The practical review plan should make clear which evidence comes from linguistic review, which comes from technical testing and which questions need joint resolution.

Design an issue log that survives revisions

For each finding, record the locale, item or string identifier, screen, build, observed behaviour, expected content and proposed next action. Attach the relevant visual evidence where permitted. Assign an owner and status, and identify whether the issue affects one language, multiple languages or a platform component.

A status of “fixed” should mean the implementation team has made a change; it should not automatically mean the reviewer has verified it. Keep correction and verification distinct. When a new build arrives, compare the changed content against the approved decision and check related elements that may have been affected. Preserve the original finding so that the rationale for the revision remains visible.

Use AI-assisted review as a controlled aid

A tool may help extract screen text, compare it with an approved file or flag possible truncation. The reviewer should know the limitations of that extraction, including missed small text, incorrectly recognized characters and assumptions about reading order. An absence of automated findings is not evidence that a screen is suitable for respondents.

Use identifiers and controlled versions to connect an automated finding to the right source. If a system suggests a correction, retain the original and route the proposal through the agreed review path. In his presentation, Dr. Matani describes AI-assisted screen issue detection while explicitly retaining a human walkthrough and approval. That boundary is particularly useful when a model appears confident about a visual result.

Put later changes through a documented release decision

EMA’s 2023 guideline for computerised systems and electronic data in clinical trials addresses formal change control in Annex 2.10. It calls for authorized change requests, risk and impact assessment, appropriate testing and documentation, approval before production, version control and retention of earlier validation records. Its scope includes systems used in trials of investigational medicinal products; the project team should assess the applicable requirements for its own study.

The following is an editorial way to organize a multilingual change discussion, not an official EMA template. Name the trigger and baseline first: for example, a newly approved instruction, a display defect or a platform update. Separate the proposed language decision from the software change. Identify which locales, devices, screens, training materials and review records depend on each. An apparently local change may touch shared code or a common source instruction.

Next, agree how the team will establish that the change is correct. A revised translation may need language review and respondent evidence; a modified display needs checks in the affected configuration; a change to branching needs functional testing. Document the rationale for the review scope, including related areas that could be affected. A clean screenshot cannot show that the correct value was saved, just as a passing technical test cannot establish that a patient understood the wording.

Work through a late instruction change

Consider a hypothetical diary in which the instrument owner clarifies an instruction after initial screen approval. The sponsor should first identify the authorized clarification and decide whether it changes the source wording, explains an existing concept or requires further investigation. The language team should not silently treat the email as permission to alter every translation. Attach the decision to the relevant source version and identify the affected language-country versions.

The implementation team can then prepare a controlled build from the approved inputs. Reviewers should compare the changed instruction with the item and response options, inspect the relevant layouts, and confirm that an obsolete string has not remained on another screen. If the same component appears across several instruments, the technical team should assess that dependency too. Keep findings, corrections and verification results connected to the exact build.

Before rollout, the responsible team should decide how the change will be communicated and how versions in use will be identified. If data were already collected, preserve the record of what participants encountered and escalate potential interpretability concerns to the appropriate study experts. Do not rewrite historical evidence to make the new and old versions appear identical. Close the issue with the approved disposition, the implementation date and the remaining limitations.

Release a traceable implementation

Before release, confirm the approved language version, final build and closure of in-scope findings. Record outstanding limitations and the responsible decision where an issue is accepted rather than corrected. Ensure that the files handed over for production correspond to the build that was reviewed, and that later changes trigger a proportionate reassessment.

For the next update, the team should be able to identify what changed and why. This makes reuse more efficient without assuming that previous approval covers new instructions, layouts or devices. A complete migration record connects the linguistic decision to the actual implementation, giving sponsors a clearer basis for oversight and future maintenance.

Sources and editorial scope

Updated 18 November 2026 from first-party LinkedIn discussions. ISPOR PCR161 provides operational observations, not a controlled effect estimate. FDA guidance provides remote-trial context; the localization journeys are editorial examples. Language review complements usability testing, technical validation and study oversight.

A worksheet for your next project

Capture the instrument version, language-country pairs, review responsibilities and delivery evidence before agreeing the scope.

Download the project brief (plain text)

Continue the conversation

Related on LinkedIn

The discussion connects remote participation with understandable assessments, instructions and language support. Earlier implementation discussions remain in the references.

Doraid Matani on localization in decentralized trials ↗

Put the discussion into practice

Discuss your COA/eCOA project

Explore our linguistic validation services and agree the languages, review stages and documentation your project needs.

Request project support →