Data Integrity & ALCOA+

A hypothetical QC laboratory runs an HPLC assay on a finished-product batch. The printed chromatogram meets specification and is signed by the analyst and a reviewer. A week later, a routine audit-trail review shows the sequence was integrated three times and that the first two integrations were deleted from the processing queue. Nothing on the paper record shows this.

1. Establish which record is the original before deciding anything else

   The instrument generated an electronic record with metadata; the paper is a report of it. Treating the printout as the original is what made the deletions invisible, and it is the determination that has to be corrected first — every later step depends on which artifact the organisation holds as the source.

2. Reconstruct what actually happened from the audit trail, not from recollection

   Pull the full audit trail for the sequence: who processed, when, what changed between integrations, and on whose account. Record what the trail shows and, equally, what it cannot show — an audit trail with gaps is a finding in its own right, and papering over the gap is worse than reporting it.

3. Separate the data question from the result question

   Two distinct questions are live: is the reported result reliable, and is the record trustworthy? A reliable result on an untrustworthy record is still a data-integrity failure, and answering only the first is the most common way this investigation goes wrong.

4. Assess the attributes against the whole record, one at a time

   Attributable, Legible, Contemporaneous, Original, Accurate, plus Complete, Consistent, Enduring and Available. Deleted integrations fail Complete and Original; a review that examined only the printout fails the review control rather than the data. State which attribute failed and where.

5. Widen the scope before narrowing it

   Ask whether the same processing practice, the same permissions, and the same review procedure apply elsewhere — other analysts, other methods, other instruments. A single-sequence investigation that never asks this is how a systemic practice gets recorded as an isolated event.

6. Fix the control that failed, not only the record that showed it

   The record is repaired by documenting the true processing history. The control is repaired by changing permissions, by making audit-trail review part of second-person review with a defined scope, or both — and by stating how effectiveness will be measured.

The organisation holds a documented reconstruction of what happened, an attribute-by-attribute assessment naming the specific failures, a scoped question about how far the practice extends, and a remediation aimed at the review control rather than at the analyst. It does not hold a conclusion that the batch is acceptable — that determination sits with the quality unit, on this evidence.

WHAT WOULD CHANGE THIS

• If the system cannot produce a complete audit trail at all, the sequence above stops at step two: the question becomes whether the system is fit for GxP use, which is a validation and system-selection decision rather than an investigation.
• If the deletions were made under a documented, approved reprocessing procedure with the rationale recorded contemporaneously, this is not an integrity failure and treating it as one damages the culture the governance depends on.
• A paper-only laboratory has no audit trail to reconstruct from, so the equivalent investigation runs on controlled-copy issuance, error correction and retention records instead — a different evidence base with different weak points.