Question 2 of 6 Fundamentals Concept
What should be checked before a clinical database lock?
Short answer
Before lock, verify predefined data quality requirements, outstanding queries, safety and external-data reconciliation, coding, deviations and approvals. Lock is a controlled milestone, not a declaration of perfection.
Why
In a multicenter trial, review the completeness and consistency of critical endpoint data, adverse-event and serious-adverse-event reconciliation, central laboratory imports, randomization information, medical coding and known protocol deviations. Check that edit checks have run as planned, unresolved issues are documented with risk-based decisions, and audit trails and access permissions are controlled. Ensure final planned transformations and extraction specifications are versioned and reproducible, and obtain approvals from authorized functions according to the data management plan.
Lock procedures should define how post-lock discoveries are triaged, authorized, changed and potentially relocked without erasing the original state. Not every trivial missing field has equal effect on analysis. The key question is whether the database is reliable and sufficiently complete for the prespecified analysis, with residual limitations transparent.
Common mistakes
- Claiming a database is safe to lock simply because every automated edit check is green
The interviewer might ask next
How would you document a known missing noncritical field at the time of lock?