Data Integrity and ALCOA+ in Clinical Data Workflows
ALCOA+ asks that data be attributable, legible, contemporaneous, original, accurate, complete, consistent, enduring and available. The principles are easy to recite and the work is evidencing them in a specific workflow. Most findings trace back to one thing: nobody wrote down what the source record is.
Data integrity discussions tend to stay at the level of the acronym, which is why they produce agreement in the room and no change in the workflow. The useful move is to take each principle and ask what, concretely, would demonstrate it for the data this study collects.
This guide maps the nine principles onto a working EDC-to-export flow, shows which stage stresses which principle, and names the four places it commonly breaks. It sits under what you own and what the vendor owns in eClinical validation.
After reading this you will be able to:
- Translate each principle into evidence rather than a statement of intent
- Identify which workflow stage puts which principle at risk
- Write a source definition before anyone enters data
- Find the four common gaps before an inspection does
Nine Principles, Nine Questions
Read the third column as a list of things you either have or do not. That is uncomfortable and it is also how an inspection works: the question is not whether you believe your data is attributable, it is whether you can produce the unique account list, the audit trail and the delegation log that demonstrate it.
Two rows are routinely weaker than the rest. Contemporaneous, because almost nobody measures the gap between activity and entry. And Enduring, because retention is treated as a policy statement rather than as something to test.
Where Each Principle Is Tested
The distribution matters because effort tends to concentrate on the middle two stages, where the systems are. Entry and review get edit checks, audit trails and monitoring. Stage one gets a form design discussion, and stage four gets a retention policy nobody revisits.
Yet stage one decides Original and Contemporaneous before any system is involved, and stage four decides whether the whole record survives. Both are cheap to address early and expensive to address late.
The Source Question
If one thing in this guide is worth acting on, it is writing down what the source record is for each data point, per site, before the first patient is enrolled.
For some data points the source is obvious: a central laboratory result, a device reading, a patient-completed diary entry. For others it is genuinely ambiguous. Is the source a nurse’s paper note, the hospital record system, or the EDC entry itself? Different sites may answer differently for the same field, and in the absence of a written agreement, they will.
Direct entry into the EDC as the original record is a legitimate choice. What is not legitimate is assuming a source exists when it does not, which is what produces the monitoring visit where nothing can be verified against anything. The agreement needs to name the source per data point and be agreed with each site rather than issued to it.
Four Places It Breaks
The third one is worth a configuration check this week. If the reason-for-change field is optional for post-entry edits, it will be left blank, not through carelessness but because optional fields are what gets dropped under time pressure. Making it mandatory for changes after initial entry costs nothing and removes the most common audit trail question there is.
Finding the blanks that already exist is a review exercise, and it is one of the signals worth looking for routinely, as covered in writing an audit trail review procedure that survives an inspection.
Complete, Through to the Export
Completeness is usually assessed at the point of collection and then assumed through everything that follows, which is where it quietly fails.
A record can be complete in the EDC and incomplete in the extract, because the export specification omitted a field, because a repeat measurement was collapsed, or because a reconciliation against external data was never run. Expected-versus-received counts at each transfer, and a reconciliation report that runs on every transfer rather than at the end, are what keep the principle true downstream.
The same applies to the submission structure, where fields can lose meaning in mapping rather than being lost outright. The design decisions behind that are in where CDASH to SDTM mapping breaks.
Clinical and R&D AI covers where automated checks help with the volume of this work without becoming a substitute for the controls themselves.
References
- ICH E6(R3) Good Clinical Practice, data governance and records. International Council for Harmonisation, adopted 6 January 2025. www.ich.org
- 21 CFR Part 11, Electronic Records; Electronic Signatures. US Code of Federal Regulations, Title 21, Part 11. www.ecfr.gov
- Annex 11, Computerised Systems. EudraLex Volume 4, Good Manufacturing Practice guidelines. health.ec.europa.eu
- Good Clinical Data Management Practices. Society for Clinical Data Management. scdm.org
This guide describes process and regulatory expectations in general terms and is not legal or regulatory advice. Confirm the current version and applicability of any standard or guidance for your study and region.
Frequently Asked Questions
What does ALCOA+ stand for?
Attributable, Legible, Contemporaneous, Original and Accurate, with the plus adding Complete, Consistent, Enduring and Available. They describe what a record has to be in order to be trustworthy. Reciting them is straightforward, and the useful work is translating each one into something demonstrable in your specific workflow, because that is what an inspection actually tests.
What is the most common data integrity gap in clinical trials?
An undefined source. Teams know their data must be original, and nobody has written down what the original record is for each data point: whether it is a paper note, a device reading, a hospital record or direct entry into the EDC. Without that, questions about whether data is original or a transcription have no agreed answer, and the gap is invisible until someone asks.
Does direct entry into an EDC break ALCOA+?
No, provided it is a deliberate decision rather than an accident. If the EDC entry is the original record, that has to be stated and the implications accepted, including that there is no earlier record to verify against. What breaks the principle is direct entry where a source was assumed to exist, or where some sites keep a source and others do not for the same data point.
How do we evidence contemporaneous recording?
By tracking the gap between when an activity happened and when it was recorded, by site, as a metric you report rather than friction you absorb. A workflow that batches entry into a weekly session produces data that is accurate and not contemporaneous, which is a different finding from inaccurate data and a harder one to argue with because the timestamps are unambiguous.
What does Enduring actually require in practice?
That records remain readable and retrievable for the full retention period set by the applicable regulation, which for clinical records can outlast the system that created them. The part most often untested is readability over time: a format that opens today may not in a decade, and a system retired without a migration plan can leave data technically retained and practically unreachable. A periodic retrieval test is the cheap control.
Can Nirmitee Healthtech assess our data integrity posture?
Yes, and the assessment runs the nine principles against your actual workflow rather than against a policy document. In practice it usually surfaces the same few things: source definition gaps, entry lag that nobody is measuring, optional reason-for-change fields, and archival readability that has never been tested. Clinera EDC is configured with mandatory change reasons and a reviewable audit trail, but the assessment is system-agnostic.
