Where it comes from
The ALCOA acronym emerged in 1990s US pharmaceutical inspection as a mnemonic for judging how trial data was kept. The "+" adds four attributes formalised later. The list was never a regulatory text: it became the common grammar in which inspectors, sponsors and methodologists discuss data integrity.
It stays useful outside regulated trials — in a thesis, a departmental registry, a retrospective study — because the questions it asks are the ones a reviewer will ask anyway.
The original five
| Attribute | Requirement | The question that tests it |
|---|---|---|
| Attributable | It is known who produced the data, and when | Who entered this value on 12 March? |
| Legible | The data reads, today and in ten years | Does this file still open without the original software? |
| Contemporaneous | Recorded at the moment of observation | Was this taken on the day of the consultation, or the night before analysis? |
| Original | The first source exists, or a certified copy | Where does this figure come from — the record, or a transcription? |
| Accurate | The data matches the observation | What does the source document say? |
The most frequently betrayed of the five is contemporaneous. Transcribing at the end of a study values observed along the way produces a dataset that looks clean and has lost what gave it value: the coincidence between observation and record.
The four in the "+"
- Complete. Nothing was removed silently. A discarded value stays visible, with the reason for discarding it — that is the difference between cleaning a dataset and arranging it.
- Consistent. Dates order correctly, units do not change from row to row, the same variable means the same thing throughout.
- Enduring. The data outlives its medium, its software, and the person who entered it.
- Available. It can be found and re-read on request — an inaccessible dataset cannot be verified.
What a spreadsheet satisfies, and what it cannot
The exercise is instructive, because a spreadsheet is the tool actually used in most work.
| Attribute | Shared spreadsheet | Why |
|---|---|---|
| Attributable | No | A cell says neither who wrote it nor when |
| Legible | Yes | The format reads widely |
| Contemporaneous | Not verifiable | Nothing distinguishes a same-day entry from a retrospective one |
| Original | No | The value is transcribed, and the link to the source is lost |
| Accurate | Not verifiable | No bounds, no type: "250" goes in where "25.0" was expected |
| Complete | No | A deleted row leaves no trace |
| Consistent | No | Nothing stops a unit changing mid-column |
| Enduring | Partly | The file survives; its successive versions do not |
| Available | Yes | As long as someone knows where it is |
In AnzarSeha
Every value carries its origin — registry, rule, model or human review — and every action enters the log with its author and timestamp. The database lock fixes the analysed state, and the audit trail hash-chains events, so that a past entry cannot be rewritten without the chain showing it.
Applicable frameworks
ICH E6(R3) addresses trial data integrity and traceability, and is the text ALCOA+ serves as a mnemonic for. The FAIR principles cover availability and reuse; the GDPR separately imposes retention periods that bound "enduring".