28.2 Critical Data, Acquisition Tools and Interim Findings

Critical Data, Acquisition Tools and Interim Findings

Data acquisition tools include CRFs/eCRFs, source worksheets, participant diaries, electronic patient-reported outcomes and other instruments used to capture trial information. E6(R3) expects fit-for-purpose tools, suitable procedures and proportionate controls. A tool should support the protocol's intended data and avoid unnecessary transcription or burden. A form containing every imaginable field can create more error without improving the trial's conclusions.

Compare the tool with the protocol: variables, definitions, units, assessment timing, response options and relevant instructions. Establish where the original record is created, how it reaches the sponsor and what metadata support interpretation. Some eCRF fields are transcribed from other records; others may be the first direct capture. Do not assume a CRF can never be source or that direct capture needs a duplicate paper original.

Critical variables and processes

A critical variable is important to participant protection or reliable interpretation, such as a primary endpoint, eligibility measurement or relevant safety result. A critical process might be consent, randomisation, dose authorisation or delivery of an urgent result. The exact set depends on the trial. A vital sign cannot be universally labelled noncritical simply because it is routine in another study.

Quality by design considers these factors before collection. Monitoring then tests whether the controls work and updates risk assessment when experience shows a gap. A large number of completed fields does not establish that a critical outcome was measured correctly.

Participant-generated data

Electronic questionnaires can use timestamps and controlled completion windows to reduce retrospective completion. They do not eliminate every risk: a caregiver might answer for a participant, a device clock may be wrong, a transmission may fail or the software may show the wrong language/instrument version. Provide suitable training/support and define how omissions and technical problems are handled.

Do not fill missed diary days from memory merely to make a completion rate appear satisfactory. Preserve actual entry timing and distinguish observation time from transmission time. A late upload can contain correctly time-stamped contemporaneous entries stored offline; conversely, a timely upload does not prove the intended person answered. Review relevant metadata and the designed workflow.

Bring-your-own-device and provided-device approaches have different access, compatibility, privacy and support considerations. Neither is inherently always more compliant. Assess intended users, instrument validity, accessibility and continuity. The protocol and system instructions should make participant responsibilities and urgent contact routes clear.

Data transfers and transmission

When data move between systems, use appropriate validated processes or reconciliation to preserve integrity and confidentiality. Verify participant/visit identifiers, units, mappings, missing records, duplicates and relevant metadata. A transfer can succeed technically while mapping a laboratory unit incorrectly or omitting an important status. Define who investigates and resolves such issues.

A monitor may notice that site source and central listings consistently differ by a factor of ten. Do not change every site record to match the listing. Establish the original values and units, assess whether the problem is transcription, transformation or mapping, and involve the responsible data/technical function. Evaluate affected safety decisions and analyses as well as the visible discrepancy.

Interim safety and efficacy findings

An interim analysis can influence continuation, modification or stopping. Its purpose, timing, access and methods should be prespecified as appropriate, with safeguards against bias. A DMC may review unblinded safety and relevant efficacy data under its charter. Other appropriately designated functions may also have limited unblinded access; a universal claim that only one committee ever sees treatment identity is inaccurate.

Maintain separation for people whose conduct or assessment could be influenced. Interim knowledge should not cause sites to change recruitment, assess outcomes differently or selectively collect data outside authorised changes. The sponsor evaluates recommendations and communicates operational instructions through controlled channels. Participant safety takes priority, but a concerning trend need not be publicised to every blinded staff member with treatment labels attached.

Monitoring a resulting action

If the sponsor pauses a cohort after a safety recommendation, verify scope, receipt, site implementation, participant care and relevant product handling. Preserve the decision and communication record and clarify conditions for resumption. If an analysis requires a data cut, verify the intended period and dataset version rather than assuming the entire global study must be complete.

The CRA helps ensure meaningful data are available and decisions become appropriate site action. They do not interpret an interim p-value independently to declare efficacy, unblind the site or waive continuing assessments. Reliable data quality combines sound design, suitable tools, traceable processing and controlled use of emerging information.

Official sources

Checked October 18, 2026: ICH E6(R3), Principles and Annex 1, ICH E9(R1).