US FDA · Data Integrity
FDA Findings
What FDA Warning Letters say about data integrity, which ALCOA+ principle broke, and what a compliant control looks like. Firms are anonymised in the text; every post links the original letter.
Failure patterns
- FDA Findings: Analysts With Admin Rights and Shared Logins Three recent FDA Warning Letters where analysts could delete or alter lab data, or shared one login. What went wrong, why it keeps happening, and what good access control looks like.
- FDA Findings: Audit Trails That Are Off, Gappy or Never Reviewed Three recent FDA Warning Letters on audit trails: not enabled, silent during testing, or full of deletions nobody assessed. What went wrong and what good looks like.
- FDA Findings: Backdated, Pre-Signed and Non-Contemporaneous Records Three FDA Warning Letters on records not made at the time: pre-signed forms, backdating, sticky notes and pencil overwritten in ink. Why it happens and how to fix it.
- FDA Findings: Deleted Data and Missing Backups Three FDA Warning Letters where lab data was deleted, never backed up, or sat on one PC that left the building. What went wrong and how to make data enduring.
- FDA Findings: Discarded and Shredded Original Records Three FDA Warning Letters where original GMP records ended up in shred bins, garbage bags or were replaced. Why 'original' matters and how to control record destruction.
- FDA Findings: Fabricated Records and Falsified Data Three FDA Warning Letters where records described work that never happened. How fabrication shows up, why controls miss it, and what detection and culture look like.
- FDA Findings: Trial Injections and Testing Into Compliance Three recent FDA Warning Letters on trial and unreported injections: passing results reported, failing ones left out. Why it is a data integrity failure and what good looks like.
- FDA Findings: Reviewing Printouts Instead of Electronic Data Three FDA Warning Letters where the quality unit reviewed printouts or photocopies and missed OOS injections and data changes. What a real second-person data review covers.