Lock is the line between collecting data and analysing it. This checklist turns it into a sequence of owned tasks with evidence: what must be true before lock, who confirms it, what to record and how to handle a request to unlock.
Free sandbox · No credit card · 21 CFR Part 11 aligned
Items complete
10 / 14
Open queries
3
Target hard lock
28 Oct
Lock in one page
Why it matters
Once a database is locked and, in a blinded trial, the treatment codes are released, any later change invites the question of whether it was influenced by knowing the results. The checklist is how you show that did not happen. It is also how you avoid the most expensive form of rework, which is unlocking, correcting and re-running an analysis. Teams that rely on memory discover missing items after lock. Teams that rely on a list discover them a week before.
The checklist below follows the order of work: data completeness, cleaning, reconciliation, review, approvals, lock mechanics and post-lock handover. It is a template to adapt. Your protocol, your data management plan and your sponsor contract decide the final content, and ICH E6(R3) data governance expectations around capture, verification, corrections, transfers and finalisation are a useful frame to check against. For the process leading up to it, see the clinical data management process and the clinical data cleaning guide.
Part 1
Owner is usually the data manager unless noted. Evidence means something you can file.
Evidence: missing-pages report compared with the visit schedule, with explanations for any gaps.
Evidence: last-entry dates per site; sites confirm no outstanding source to enter.
Evidence: query listing at zero open, or a documented list of unresolved items with justification. See query management.
Evidence: listings of range and consistency flags reviewed and signed by the data manager.
Evidence: SDV status report against the monitoring plan, including critical fields. See SDV.
Evidence: review notes with outcomes, covering outliers, impossible dates and visit window issues.
Part 2
Evidence: signed reconciliation showing matching cases, dates and terms, with differences resolved.
Evidence: coding report with reviewer sign-off. Coding is performed in your coding tool or by your provider.
Evidence: lab, imaging, device or randomization data compared with the database and mismatches resolved.
Evidence: deviation log with categories agreed before lock. See protocol deviation tracking.
Evidence: spot checks of calculated fields against manual calculation.
Evidence: confirmation that assignments match the randomization record, with unblinding kept closed until lock approval.
Use the free sandbox to practise queries, SDV status, exports and the audit trail on sample data, so the real lock holds no surprises.
Part 3
Evidence: dated electronic signatures with the meaning of the signature stated.
Evidence: completed checklist cross-referenced to the plan version.
Evidence: time-stamped soft lock, with a list of any controlled changes made after it.
Evidence: approval record from the sponsor, data management and statistics, as the plan requires.
Evidence: lock timestamp, user and the change in access rights. See database lock software.
Evidence: export with a data dictionary and an audit trail export, with checksums or file records. See audit trail export for inspectors.
Evidence: transfer log, and where used, CDISC datasets. See CDISC SDTM export.
Evidence: unblinding request and approval dated after hard lock.
Terms
| Term | Meaning in practice | Typical record |
|---|---|---|
| Soft lock | Data frozen for review; controlled changes still allowed with a reason | Soft lock timestamp, list of post-lock changes |
| Hard lock | Database closed to changes for analysis | Lock approval, timestamp, access change |
| Interim lock | Lock of a subset or snapshot for a planned interim analysis | Snapshot definition, approvals, analysis plan reference |
| Unlock | Exceptional reopening after lock to correct a justified error | Request, justification, approval, changes, relock |
Definitions vary by organisation. Align them with your data management plan.
If it goes wrong
Sometimes a real error appears after lock: a transposed value that changes a primary analysis, a missing serious adverse event report or a site that returns late data. The answer is not to avoid unlocking at any cost or to do it casually. Raise a written request that states the error and its impact, obtain approval from the people named in the plan (often sponsor, data management and statistics), make only the listed changes, record each in the audit trail with a reason, run the checks again for the affected data and relock with a new timestamp and approval.
Treat every unlock as a signal to ask why the checklist missed it, and update the list. If you are choosing software for this, check that unlocking requires a reason and is recorded. The database lock software page covers how Capture handles freeze, lock and unlock with a reason and an audit trail. Regulatory background for inspection readiness is in the FDA inspection checklist and audit-ready clinical trial data.
In Capture
Capture records query status, per-field SDV and a field-level audit trail that cannot be modified in the UI, and clinical data is soft-deleted only. Forms are approved and locked for live use, and exports come with an automatic data dictionary, including SDTM as SAS XPT with Define-XML. Run your whole checklist on sample data in the sandbox first.
Completeness of data, closed queries, SDV, safety and external data reconciliation, coding, deviation classification, investigator sign-off, lock approval, final exports with the audit trail, and a controlled unblinding step after lock.
A soft lock freezes data for final review while controlled changes with a reason are still possible. A hard lock closes the database to changes so the analysis runs on a fixed dataset.
It depends on the data management plan, but commonly the sponsor, data management and statistics approve lock, and investigators sign off their casebooks beforehand.
Yes, in exceptional cases, with a written justification, approval from named roles, recorded changes with reasons and a documented relock.
It depends on how much cleaning remains after the last visit. Continuous cleaning during the study shortens the time, and leaving it to the end lengthens it.
In blinded trials unblinding follows hard lock and approval, so that no data change can be influenced by knowing the treatment assignments.
Capture supports lock readiness with query status, per-field SDV, a field-level audit trail, locked approved forms and exports with a data dictionary. See the database lock software page for the freeze, lock and unlock workflow.
Keep exploring
Database lock (glossary)
The short definition.
Database lock software
Freeze, lock, unlock.
Clinical data cleaning guide
Getting to clean data.
Clinical data management process
The full sequence.
Audit-ready clinical trial data
Inspection readiness.
Data archiving and retention
After lock.
Clinical trial lifecycle
Where lock sits.
Free sandbox with every feature. No credit card, and you pay only when you go live with real participants.