Source documents in plain language, ALCOA without the acronym worship, and why you never erase. Written for new coordinators and research assistants, and for anyone who has tidied away a sticky note.

In this file:

·        What counts as source, and what surprisingly does

·        Why the first place a fact was written down wins

·        ALCOA in plain language, without the acronym worship

·        Why you never erase, and what you do instead

Marcus took Walter's blood pressure, wrote it on a yellow sticky note stuck to the edge of the monitor, finished the rest of the visit, and transcribed the number into the chart forty minutes later. He threw the sticky note away, which is the part he mentioned in passing, pleased with himself for tidying up. Nani asked him, without any edge in her voice, to go and get it out of the bin.

The definition is simpler than the anxiety around it. A source document is the first place a piece of study information was written down.

That covers the obvious things. Medical charts. Lab reports. ECG tracings. Pharmacy dispensing records. The signed consent form. The diary Walter fills in at home, which for a Parkinson's study is not a supporting document but often the primary endpoint itself.

It also covers things people don’t expect, and this is where new staff get caught. A nurse's note. A phone log of a call with a patient. An email, if the email is where a study fact first landed. A photograph of a rash. A screenshot. And yes, a yellow sticky note with a blood pressure on it.

The rule that makes sense of all of it: source is defined by being first, not by being official. You don’t get to decide after the fact that something was only a scratch pad. If a study value existed on that sticky note before it existed anywhere else, the sticky note is a source document, and it was one from the moment the pen left the paper. You cannot un-create source.

This isn’t a technicality invented to make life difficult. It exists because the alternative is a system where anyone can generate a record, decide retrospectively that it was informal, and discard it. That’s precisely the mechanism by which inconvenient data disappears, and every rule in this section is built to close it.

The sponsor's database is a copy. Every value in the electronic case report form was transcribed out of something that existed here first.

So when the two disagree, the question is not which one is correct in some abstract sense. The source is what happened. The database is what somebody typed. If they differ, the database is wrong, and it gets corrected against the source rather than the other way around. A coordinator who "fixes" a source document to match a database has inverted the entire structure, and that is a far more serious finding than the original discrepancy ever was.

The regulation behind this is quieter than people expect. 21 CFR 312.62 requires an investigator to prepare and maintain adequate and accurate case histories recording all observations and other data pertinent to the investigation, and to retain those records. Adequate and accurate is doing a lot of work in that sentence, and it is the hook everything else hangs from.

There is one legitimate way an original gets replaced, and it has a name. A certified copy is a copy that has been verified as a complete and accurate reproduction of the original, with that verification itself documented, by signature or by a validated process. Scanning a stack of paper source into a document system and shredding the paper is defensible if you did it as a certified copy process. It is indefensible if you just scanned things.

You will meet this acronym in your first week and hear it repeated with more reverence than it needs. It is a mnemonic for what makes a record trustworthy, and it appears in regulator guidance on data integrity rather than as a numbered rule you can cite by section. Knowing that is itself useful, because people quote it as though it were law.

·        Attributable. Someone can tell who recorded it. Initials, a signature, a user account.

·        Legible. Another human can read it, now and in fifteen years, including the corrections.

·        Contemporaneous. It was written when it happened, not reconstructed later from memory.

·        Original. It is the first record, or a certified copy of it.

·        Accurate. It matches what actually occurred.

The plus in ALCOA+ adds complete, consistent, enduring and available, which are mostly about the record surviving and staying findable rather than about how it was made.

Run Marcus's sticky note through that list and the problem stops being abstract. Attributable fails, because he did not initial it. Contemporaneous is where it really breaks: the number that went into the chart forty minutes later was written from memory and from a scrap, not at the moment of measurement, and nobody reading the chart would know that. Original fails once the note is in the bin, because the first record of that blood pressure no longer exists.

None of which means Marcus did anything dishonest. He wrote a number down on the nearest surface, which is what everyone's hands want to do. The habit that has to replace it is small and boring: write it where it is going to live.

Note to File: The most common source mistake a new person makes is not a wrong number. It is a right number written somewhere temporary. If you catch yourself about to record a study value on anything you intend to throw away, stop and put it in the source instead. And if you have already written it on the scrap, the scrap is now source: initial it, date it, and file it. The sticky note was never the real problem. Planning to destroy it was.

You will make errors in source. Everyone does, at every site, in every study. There is an established way to fix them and one absolute prohibition.

Never erase. No correction fluid, no scribbling a value into illegibility, no rewriting the page cleanly and binning the original. On paper, the correction is a single line through the wrong value, leaving it readable, the correct value written beside it, and your initials and the date. If the reason would not be obvious to a stranger, add a few words: transcription error, wrong unit, recorded on wrong patient.

The prohibition is not about neatness. A record exists so that a reader years from now can reconstruct not only what the answer was but how it got there. An erasure removes the history, and a record whose history has been removed cannot be trusted even when the surviving number is perfectly correct. The reader has no way to distinguish a corrected transcription error from a value someone did not like.

Electronic systems do this automatically, which is why the audit trail is such a big deal in validated study software. It records the original entry, the change, who made it, when, and often why, and it cannot be switched off by the person entering data. The paper convention and the electronic audit trail are the same idea wearing different clothes.

A source document in clinical research is the first place a piece of study information was recorded, whether or not anyone intended it to be official, which is why a sticky note with a blood pressure on it is a regulatory record. The sponsor's database is a copy of that source, so when the two disagree the source governs and the database gets corrected. ALCOA, attributable, legible, contemporaneous, original and accurate, is the plain-language test for whether a record can be trusted, and it is a guidance mnemonic rather than a citable rule. Corrections are made with a single line, the right value, initials and a date, and the one thing you never do is erase, because a record whose history has been removed cannot be trusted even when the number left behind is right. Marcus fished the sticky note out of the bin, initialed it, dated it, and filed it in Walter's chart, where it has been ever since.

·        Previous issue: Issue 8, the monitoring visit

·        Next issue: Issue 10, data queries

·        The whole file: roottorx.com/t/open-site-file

·        Root to Rx: the newsletter and The Root Room, the patient-facing side of the same story, at roottorx.com

·        21 CFR 312.62 — Investigator recordkeeping and record retention (eCFR)

·        FDA — Electronic Source Data in Clinical Investigations, Guidance for Industry (September 2013)

·        MHRA — GXP Data Integrity Guidance and Definitions, Revision 1, March 2018

·        FDA — E6(R3) Good Clinical Practice, Guidance for Industry

•  •  •

The Open Site File is general education about how clinical research sites work. It is not regulatory, legal, or medical advice, and it is not tied to any specific study, sponsor, or employer. Study OLS-114 and all names are invented. Always follow your own site's SOPs, your protocol, and your PI and QA team. Characters are composites.

Disclosure: Root to Rx is published independently by Open Label Media LLC. Views expressed are personal views of Justin Yamashita and do not represent his employer or any affiliated organization. No employer resources or proprietary information are used. Every claim is sourced from publicly available materials.

Root to Rx | 228 Park Ave S, #29976, New York, New York 10003, United States