A database lock is the moment a study's records stop changing. Before it, the last forms come in, the monitor's questions are answered, corrections are made with their reasons and the investigator signs off. At the lock, the data become the fixed set that the analysis and the final study report rest on. A reviewer proves nothing changed after it with two checks: the audit trail shows no entry dated after the lock, and a fingerprint of the locked data, a short code computed from every value and recorded at the lock, comes out the same when the reviewer recomputes it from the copy in hand. Change one value and the code changes.
What the rules ask
GL9 never uses the words "database lock". FDA's Center for Veterinary Medicine names it in Guidance #197, Documenting Electronic Data Files and Statistical Analysis Programs (November 2020). Section III.B.2 treats the audit trail of an electronic data capture (EDC) system as part of the raw data: the original value, every change, who made it, when, the reason, "and the date when the database was locked". Its sample audit-trail file has a lockdate column. Section III.B.3.b adds that "Submitted data files should be unmodified after export from the database"; when a file is changed for analysis, the sponsor describes the change and supplies the program that made it.
Guidance #287, Raw Data for Safety and Effectiveness Studies (June 2025), section III, says "The audit trail should also be submitted if raw data were collected by an electronic data capture system." Section IV asks the sponsor to describe how the data were collected, "including a record of all changes to the raw data", from the EDC files to the completed statistical analyses.
Guidance #85 (VICH GL9), the good clinical practice standard for veterinary studies, defines the final study report as one written "after the collection of all raw data is complete" (1.13). It keeps every change to electronic data in an audit trail (8.3.6), stores study documentation protected from "deterioration, destruction, tampering or vandalism" (8.4.1), and makes any correction to the final report a signed, dated amendment with its reason (7.4).
21 CFR Part 11 applies to records kept electronically under an FDA records requirement or submitted to FDA (11.1(b)). Section 11.10 asks for validation that includes "the ability to discern invalid or altered records" (a), accurate and complete copies in human-readable and electronic form (b), protection of records through the retention period (c), and secure, time-stamped audit trails whose changes "shall not obscure previously recorded information", kept as long as the records (e). Section 11.70 links each signature to its record so it cannot be "excised, copied, or otherwise transferred". FDA's October 2024 questions-and-answers guidance on electronic systems, which lists CVM among its issuing centers, says audit trails "should be protected from modification and from being disabled" (Q12), and that sponsor inspections look at the procedures that keep the records needed to reconstruct a study from being "altered in value or meaning" (Q8).
How a reviewer checks a lock
A monitor, an auditor or an FDA reviewer holding the export can run these checks without the system that produced it.
- The lock record: the date and time, who locked, and the fingerprint.
- The audit trail: no entry after the lock date, and every earlier change showing the old value, the new value, who, when and why (2024 guidance, Q12).
- The fingerprint: recomputed from the exported data, it matches the one recorded at the lock.
- The analysis files: the same as the export, or each change described with the program that made it (Guidance #197).
- The signatures: each shows the signer's printed name, the date and time, and the meaning (11.50), tied to the record it covers.
If all five hold, the data the report rests on are the data that were locked.
Steps to a lock
- Write the lock into the protocol or the data management plan before the first animal enrolls: who may lock, and what must be true first. GL9 6.3.18 asks the protocol to set out how raw data are recorded, processed, handled and retained.
- Bring in the last data: every visit, every owner diary entry, every laboratory panel, every end-of-study form.
- Close the monitor's queries. List any left open, with the reason.
- Reconcile adverse events, concomitant medications and deviations. GL9 3.2.5 asks for each deviation in a signed and dated statement.
- Have the investigator sign off the records.
- Lock. Record the date, the time, who locked and the fingerprint.
- Export at once: one file per form, a data dictionary and the audit trail with the lock date.
- In a blinded study, open the allocation for analysis only after the lock, so the lock date also shows the data were fixed before anyone saw the arms. GL9 3.2.18 lets a treatment code be broken only as the protocol says and with the sponsor's knowledge and consent.
- If a value must change after the lock, record the change, the reason and the date on the audit trail, and lock again with a new fingerprint. Never change a locked value silently.
What changes with the study design
- Observational study or registry, nothing assigned: GL9 2.6 puts studies "not intended to be used for regulatory support" outside its scope, and Part 11 reaches only records an FDA rule requires or that go to FDA. Unless the data are headed to FDA, no rule asks for a lock. Lock anyway: a fixed dataset with a fingerprint is something a paper, a thesis or a later submission can cite, and a co-author or examiner can check.
- Marketed product used on label: there is no investigational filing for the product. If the data are meant for FDA, GL9 and Part 11 apply as above. A lock does not pause safety reporting: under 21 CFR 514.80 the product's applicant files 15-day alert reports "within 15 working days of first receiving the information".
- Unapproved product under an investigational exemption: 21 CFR 511.1(b)(7)(ii) requires the investigator to keep complete records of the investigation and retain copies "for 2 years after the termination of the investigation or approval of a new animal drug application"; 511.1(b)(8)(i) has the sponsor keep investigator reports as long and make them available for inspection. Those are FDA records requirements, so Part 11 applies. The two years can start at an approval years after the lock, so the locked copy, its audit trail and its fingerprint have to stay readable that long; the 2024 guidance (Q17) asks that agreements with an IT service provider include a plan for the sponsor's access to data through the retention period. Findings that may suggest significant hazards are still reported promptly to FDA and all investigators (511.1(b)(8)(ii)), before or after the lock.
What stays with the sponsor and the investigator
The lock is a decision people make. When to lock, what the protocol says about it, whether a change after the lock is justified, the ethics review, the consent content, the INAD file, any submission and every adverse event report belong to the sponsor and the investigator. GL9 4.3.1 leaves the "ultimate responsibility for the quality and integrity of the study data" with the sponsor, and the 2024 guidance (Q17) keeps sponsors responsible for any obligation not lawfully transferred to an IT service provider. A data system can make a lock hard to break and easy to check. It cannot decide that the data are ready.
Signalment and the lock
Signalment is a data system. A study in it closes with a database lock that carries a fingerprint. The audit trail is append-only and hash-chained, and it is exported with the fingerprints in a JSON bundle next to CSV files per form and a data dictionary, so a monitor can verify the chain offline, on their own laptop. Signatures re-check the signer's password and bind to the hash of what was signed. There is more on the sponsor page, and on the records rules in 21 CFR Part 11 for veterinary studies.
What it does not do yet: SAS and SEND exports. Guidance #197 says data sent to CVM typically arrive as SAS transport (XPT) or XML files and expects a conversion from CSV to change nothing, so that step, and the submission itself, stay with the sponsor.
Sources
- FDA Guidance #197, Documenting Electronic Data Files and Statistical Analysis Programs (November 2020): https://www.fda.gov/media/75077/download
- FDA Guidance #287, Raw Data for Safety and Effectiveness Studies (June 2025): https://www.fda.gov/media/177901/download
- FDA Guidance #85, VICH GL9, Good Clinical Practice: https://www.fda.gov/media/70333/download
- 21 CFR Part 11, electronic records and electronic signatures: https://www.ecfr.gov/current/title-21/part-11
- FDA, Electronic Systems, Electronic Records, and Electronic Signatures in Clinical Investigations: Questions and Answers (October 2024): https://www.fda.gov/media/166215/download
- 21 CFR 511.1, investigational new animal drugs: https://www.ecfr.gov/current/title-21/section-511.1
- 21 CFR 514.80, reports on approved new animal drugs: https://www.ecfr.gov/current/title-21/section-514.80