ALCOA+ on the floor: the reviewed record has to be what the plant generated

On the floor, a data integrity failure does not always start with a missing lab checklist. The shift generates injections, instrument printouts, equipment logbook entries and sampling records. The dossier that goes forward for release can show something else: injections that are missing, a copy instead of the original, raw data that does not match the batch record or the certificate of analysis, or a sampling entry with no matching sample. The inspection reads the dossier. The plant generated the data. When the two diverge, release rests on incomplete evidence.

Data integrity starts at the batch, the logbook and the instrument, during the shift. In practice, ALCOA+ on the floor means a record that is contemporaneous, attributable, original and accurate: what was done, by whom, at the time it was done, with no cleaned-up version later. It is a common pattern in recent FDA warning letters, and the problem is not the tool. It is execution and retention of the data the inspection will ask for.

What the inspection asks

The key distinction: the inspection does not ask whether staff completed data integrity training. It asks whether the batch record, the equipment logbook and the instrument output describe the same execution.

On the floor, contemporaneous and attributable mean the value was recorded when the step happened and you can identify who operated the instrument or completed the record. Original means the printout, the analytical sequence and the audit trail retained by the system or on paper, not the copy preferred at review. Accurate completes the set: if the record says sampling took place and there is no sample, the record is wrong, and the review inherits the error.

In the letters cited below, FDA cites the following 21 CFR Part 211 requirements: 211.194(a) (complete laboratory data), 211.22(a) (quality unit responsibilities) and, in the April 2026 letter [ref. 1], also 211.68(b) (access and change controls for computerized systems) and 211.192 (production record review). FDA's guidance Data Integrity and Compliance With Drug CGMP: Questions and Answers (December 2018) is interpretive, not a new rule. In Brazil, the equivalent requirements sit in RDC 658/2022 (Good Manufacturing Practices for Medicinal Products) and, for computerized systems, IN 134/2022.

If the recording system, paper or electronic, allows changes to what was already recorded, or instrument raw data cannot be cross-checked against the batch record before release, the dossier no longer reflects execution.

When data disappears or is selected afterwards

Two mechanisms are common.

Discarded trial injections. In April 2026 [ref. 1], FDA cited a finished pharmaceuticals manufacturer for, among other things, reporting trial injections when results passed and disregarding them when results were less favorable or OOS, and for deleted gas chromatography sequences found in the recycle bin. The data supporting release is no longer the full set generated at the instrument. The failing control is the workflow that lets an injection be discarded and only the convenient result be filed.

Original discarded, copy reviewed. In September 2026 [ref. 4], FDA cited a US API manufacturer for, among other things, thousands of unexplained injections in one chromatography data system over 30 months and more than 70 trial injections in another system in October 2025 alone, performed without adequate documentation or supervision to ensure product samples were not being tested; also for original laboratory printouts found in shred bins, and review performed on photocopies of laboratory notebook pages. Reviewing a copy is not reviewing the retained original. Without the original, the review looks at a substitute that may have been selected.

It is worth checking whether the lab or production workflow allows discarding injections, filing only passing results, or accepting a photocopy as the basis for release.

Logbook and raw data that do not match

Data integrity on the floor is not limited to HPLC. It includes the equipment logbook and the entry that declares a sample was taken.

Logbook that does not match the original. In July 2026 [ref. 2], FDA cited an over-the-counter drug manufacturer in India for providing the investigator with a falsified equipment usage logbook (the original was later found in the document room) and for discrepancies between raw data in analytical notebooks and values reported in batch records or certificates of analysis. If the logbook is the only link between equipment and batch, and that link can be rewritten, the dossier stops being evidence of execution.

Entry with no matching sample. In August 2026 [ref. 3], FDA cited a drug manufacturer in India because environmental monitoring samples were documented in logbooks as collected and incubated, but the plates could not be located in the incubators during the inspection. The case matters here as a failure of attributable and accurate records, not as a topic on environmental monitoring programs.

Both cases point to the same gap: raw data (instrument, plate, sampling) must be cross-checked against the batch or equipment record before release. The useful question is not whether a data integrity SOP exists, but whether instrument raw data can be cross-checked against the batch record with the original still available.

What the quality unit approves

When floor data does not reach review intact, the quality unit can approve release without the full set of what was generated, because the dossier arrived already filtered, copied or altered. In the July 2026 letter [ref. 2], FDA notes that the quality unit signed off on records without ensuring consistency between raw data and reported values.

What the recording system needs to allow

Where the electronic batch record captures data at execution (value, operator, time) and allows instrument raw data to be cross-checked against the batch record, with an audit trail, the plant gains the capability to produce contemporaneous, attributable evidence. That alone does not guarantee ALCOA+ or 21 CFR Part 11 compliance. What the system does is make changes to recorded data harder and make that cross-check visible before release.

A question for the next batch

Can you cross-check instrument raw data against the batch record and see who did what, and when, with the original available for review?

If the answer is "only later, from a copy" or "trial injections are not kept", the review is working from an incomplete record, and release may rest on it.

References

1. FDA — Warning letter, Ava Inc., 04/14/2026
2. FDA — Warning letter, Dabur India Limited, 07/24/2026
3. FDA — Warning letter, Reliance Life Sciences, 08/18/2026
4. FDA — Warning letter, Curia New York, Inc., 09/18/2026
5. FDA — Data Integrity and Compliance With Drug CGMP: Questions and Answers (2018)
6. ANVISA — IN 134/2022, computerized systems (English)
7. ANVISA — RDC 658/2022 and GMP rules