Mid-Study Edit Check That Would Re-Open 400 Queries
Six months into a 24-month Phase II study, the sponsor's medical monitor notices that a lab value (serum creatinine) is being entered in µmol/L by some sites and mg/dL by others - the CRF field has no unit control. Of 1,200 entered lab records, roughly 400 are suspected to be in the wrong unit. Data management proposes adding a hard edit check that rejects any creatinine above 200 and a unit dropdown, effective immediately. The EDC vendor says this is a 'configuration change' that can be deployed Friday without a formal change request because it 'only adds a check.' The study is mid-enrollment; 14 sites are actively entering data. The validation package for the current build was signed off at go-live.
1. Is the vendor's characterization of this as a non-controlled configuration change acceptable? Why or why not?
Read the worked answer
No. The change is not a trivial configuration tweak, for three reasons:
- It changes validation status. Adding a unit dropdown changes the data model (new field or controlled vocabulary) and the hard check changes entry behavior. The validated state documented at go-live no longer describes the system. Under computerized system validation principles, any change that affects data capture or processing requires impact assessment and, where warranted, re-validation or documented regression testing.
- It has downstream data consequences. 400 existing records may be wrong. A new hard check will not fix historical data and may generate a flood of queries or block legitimate entries (a genuine creatinine of 250 µmol/L is normal-ish; 200 as a hard ceiling is clinically wrong if units vary).
- 'Only adds a check' is the classic understatement. Hard checks can lock users out of completing visits. That is a high-impact change to site workflow.
The correct route: formal change request, impact assessment (data, validation, training, queries), risk-based testing, approval, controlled deployment, and communication to sites.
2. Design the remediation for the 400 suspect records. What must the plan include beyond re-entering values?
Read the worked answer
Re-entry alone is insufficient. The plan must include:
- Detection logic: define the rule that flags suspect records (e.g., value/unit implausibility given the reference range and the site's historical pattern), and document it so the rule is reproducible.
- Source verification: each suspect value must be checked against the source document by the site or a monitor - not corrected by assumption. Corrections require reason-for-change in the audit trail.
- Query management: raise queries with a clear, unit-aware question; track to closure; do not bulk-close.
- Root cause and CAPA: why did the field lack unit control? Update the CRF/build standard so future studies include unit enforcement.
- Impact on analysis: confirm with the biostatistician whether any interim analyses used these values; document any impact.
- Documentation: deviation/CAPA record, change request, updated validation evidence, and a summary in the data management report.
The exam point: remediation is a controlled process (detect → verify against source → correct with trail → root cause → prevent recurrence), not a data-cleaning exercise.