ALCOA+ and the analytical record: what data integrity means for the chromatogram behind a peptide purity figure

A purity figure on a certificate of analysis is a single number, but it is the output of a sequence of decisions: which injections were run, which were kept, where the integration baseline was drawn, which peaks were included in the denominator, and who looked at the result before it was reported. The number itself carries none of that history. Two certificates can both state 98.4% and reflect entirely different degrees of documentary support, because the figure is only as informative as the record that stands behind it.

Regulators have a name for the discipline of making that record trustworthy: data integrity. The term is sometimes treated as an IT concern or as a synonym for fraud detection, but its working definition is narrower and more useful. Data integrity is the degree to which data are complete, consistent and accurate throughout their life cycle. This article walks through the ALCOA+ framework that organizes those expectations, applies it to the specific case of a chromatography data system, and considers what it does and does not imply for someone reading a peptide analytical record.

The framework: ALCOA and what the plus adds

The acronym ALCOA originated in FDA inspectional practice and has since been adopted, with minor variations, in guidance from the World Health Organization, the Pharmaceutical Inspection Co-operation Scheme, the UK MHRA and the European Medicines Agency. The original five attributes describe a good record. It should be attributable, meaning it can be traced to the person or system that generated it; legible, meaning it can be read and understood for as long as it must be retained; contemporaneous, meaning it was recorded at the time the activity occurred; original, meaning it is the first capture of the observation or a certified true copy of it; and accurate, meaning it faithfully reflects what happened.

The extended form, usually written ALCOA+, adds four further attributes: complete, consistent, enduring and available. Completeness means all data are present, including repeat analyses and results that were not ultimately reported. Consistency means the sequence of events is coherent, with timestamps that follow a logical order. Enduring means the record persists in a durable medium for the whole retention period, and availability means it can be retrieved for review when needed. Some frameworks extend the list further still, but the nine-attribute form captures most of what inspectors look for in practice.

The attributes are not independent checkboxes. A handwritten note on a loose sheet can be legible and attributable and still fail on enduring and complete. A pristine electronic file can be original and enduring and still fail on accurate if the processing that produced the reported value was undocumented. The framework is best read as a set of questions asked of a record rather than a list of properties it either has or lacks.

Why chromatography is the pressure point

FDA’s 2018 question-and-answer guidance on data integrity and current good manufacturing practice, along with the equivalent documents from other authorities, gives disproportionate attention to computerized laboratory systems. The reason is structural. In a manual process, the raw observation is on paper and the interpretation follows it. In a chromatography data system, the raw observation is a digitized detector trace and the reported result is a derived quantity produced by software processing parameters that a person selected.

Between those two points sit several steps where the record can quietly diverge from the underlying measurement. Integration parameters, such as slope sensitivity, peak width and minimum area thresholds, determine where the software places baselines and which small features are treated as peaks. For a peptide purity assay the consequence is direct, because purity by area percent depends on the denominator, and the denominator depends on which minor peaks the integrator recognized. Reprocessing the same raw file with different parameters can shift the reported purity without any new injection being made. Manual integration, in which an analyst redraws a baseline by hand, can be entirely legitimate for a co-eluting or drifting-baseline case, but it is exactly the kind of intervention that should leave a reason and an author behind it.

Regulatory guidance therefore asks for several specific controls. Each user should have an individual login, because shared credentials defeat attributability. Audit trails should record the creation, modification and deletion of data, including changes to processing methods and integration events, and they should be reviewed as part of the review of the result rather than only during investigations. System clocks should be controlled so that timestamps are meaningful. And the original electronic raw data, not a printout or a PDF rendering of it, is generally regarded as the primary record, because only the electronic file can be reprocessed and audited.

Reprocessing, sample sets and the problem of the unreported injection

One of the recurring themes in inspection findings is the practice sometimes called testing into compliance: running a sample repeatedly, or reprocessing it repeatedly, until a result inside the specification appears, and then reporting only that result. The completeness attribute exists largely to address this. A result is only interpretable alongside the results that were generated and set aside, and the reasons for setting them aside.

For a peptide HPLC purity assay, this has a practical reading. System suitability injections, blank injections, standard injections and any repeat sample injections are all part of the sequence, and a complete record accounts for each of them. An injection that is excluded should be excluded for a documented, pre-defined reason, such as a failed system suitability criterion or a recognized instrument fault, rather than because its value was inconvenient. Pre-defined invalidation criteria are valuable precisely because they are written before the result is known. The related concept of result-independent decision-making, in which the rule for repeating or rejecting a run does not reference the outcome being sought, is one of the cleaner ways to tell a controlled process from a flexible one.

Audit trail review makes this concrete. A reviewer examining a sequence can see whether injections were aborted and restarted, whether the processing method was altered after the first pass, whether peak integration was edited manually, and whether results were deleted. None of those events is evidence of misconduct in itself. Methods get optimized, instruments fail mid-sequence and baselines do drift. The purpose of the audit trail is that such events are visible and explained rather than silent.

What this does and does not tell a reader of a certificate

It is worth being precise about the limits of the framework as applied to a peptide purity certificate, because data integrity is easily over-read in both directions. The ALCOA+ attributes describe properties of a record-keeping system. They do not describe the quality of the analytical method, the identity of the material, or the suitability of the specification. A record can be fully compliant with every data integrity principle and still report a purity from a method that lacks the resolving power to separate a critical impurity. The reverse also holds: a well-designed method run on a system with weak controls produces a result whose provenance cannot be demonstrated, whatever its underlying accuracy.

The framework is also scaled to context. The expectations in FDA, MHRA and PIC/S guidance were written for regulated manufacturing and testing under current good manufacturing practice, where 21 CFR Parts 11, 210 and 211 and their international equivalents apply. Research-grade material is not generally produced or tested within that framework, so a reader should not assume that a research certificate was generated under those controls, nor that its absence implies anything about the result. What the framework offers is a vocabulary for asking informed questions: whether raw data files are retained, whether the processing method is documented, whether the injection sequence is accounted for, and whether the person who generated the result is identifiable.

Those questions have a useful property: they can be asked of any analytical record without access to the laboratory. A chromatogram image that shows the integrated baseline, the acquisition date, the method identifier and the sample or lot identifier is more informative than a bare percentage, because each of those items corresponds to one of the attributes above. Attributable and contemporaneous are served by the date and operator context; original and complete are served by retention of the underlying file; accurate is served by a documented processing method.

Synthesis

The ALCOA+ attributes reduce a large body of regulatory text to a compact set of questions about any analytical result: who generated it, when, from what original data, with what processing, and with what accounting for everything that was set aside. Applied to a chromatography data system, they point to the places where a digital record can drift from the measurement it summarizes, namely integration parameters, manual edits, reprocessing and unreported injections, and they point to the controls that make such drift visible, namely individual accounts, reviewed audit trails and retained electronic raw data.

None of this converts a purity percentage into a guarantee, and the framework was never intended to. What it describes is the difference between a number and a documented measurement, and the degree to which a reader can reconstruct how the one became the other. Read this way, a reported value is only as portable as the record that allows someone else to reconstruct it.