The auditor asked for the injection log for run 47-113 and we produced it in ninety seconds. Then she asked for the balance printout that supported the standard weight recorded in that log, and the room went quiet. The printout existed. It was in a binder on the second floor, in a page protector, filed by month, not by run. Finding it took forty minutes. That forty minutes did more damage to the audit outcome than any out-of-specification result we had that year, because it told the inspector our analytical testing audit documentation was assembled for convenience rather than for reconstruction.

That is the whole lesson of this article. Inspectors do not grade you on whether records exist. They grade you on whether an independent person can rebuild a specific run from the record set without your help, and on whether you can prove the record has not changed since it was written. Everything below is about that reconstruction test and the retention rules that govern how long you have to keep passing it.

When an inspector pulls a run, she is not asking for “the data.” She is assembling a chain that starts at the sample container and ends at the signed report. I tell my staff to think of it as eleven artifacts. If any one is missing, the run is not reconstructable, and an unreconstructable run is treated as if it never happened.

• Sample receipt record: date, time, temperature at receipt (thermometer ID), condition, unique sample ID, chain-of-custody signatures with times.
• Login and storage record: where the sample sat between receipt and prep, at what temperature, and who moved it.
• Preparation worksheet: balances used, weights to four decimals, dilution volumes, lot numbers of every reagent and standard, and the analyst’s initials at each step.
• Standard and reagent traceability: certificate of analysis for each standard lot, expiry date, and the preparation record for working standards including the solvent lot.
• Consumable lot records: the vial, cap, septum, and column lot numbers tied to the run. This is the artifact labs skip most often, and it is the one that answers “why did the peak shape change in March.”
• Instrument logbook entry: run start and end time, method version, column serial number and install date, detector settings, and any maintenance in the preceding 24 hours.
• Sequence table: the exact injection order, including blanks, checks, and the position of calibration levels.
• Raw data files: the unprocessed chromatograms or spectra, with audit trail enabled, not the PDF export.
• Processing record: integration parameters, any manual integration with justification, calibration curve with residuals.
• QC record: method blank, continuing calibration verification, matrix spike, duplicate, and the acceptance criteria applied.
• Review and approval: second-person technical review signature with date, and the final report version that cites the run.

Eleven artifacts, and the inspector will spot-check the weakest link. In my experience the weak links are almost always the consumable lot records and the raw data files, because both live in systems that were never designed to talk to each other. If you buy vials and caps from a distributor who cannot give you lot-level certificates, you have already failed artifact five. That is one reason I am picky about where consumables come from; the chromatography consumables supply chain needs to hand you a lot number on every box, not just a catalog number.

Retention periods: what the regulations actually require, and where they conflict

Nobody remembers the numbers correctly, so here they are in one place. The catch is that the applicable period depends on which regime governs the work, and a lab that serves several clients is usually bound by the strictest one for every record it touches.

Two things about this table matter more than the numbers. First, 21 CFR 211.194 ties retention to batch expiry, which means a product with a five-year shelf life can push your chromatogram retention out to eight years from the test date. If your LIMS archives raw files to a five-year schedule, you are non-compliant and you will not find out until an inspection two years after you thought you were done. Second, ISO/IEC 17025 does not give you a number; it gives you an obligation to define one and follow it. An auditor will read your quality manual, find your stated retention period, and then ask you to demonstrate that the archive actually retains files for that long. A policy of “ten years” backed by a system that overwrites at seven is worse than a policy of “seven years,” because the first is a false statement.

My working rule for a multi-regime lab is to set one internal retention period equal to the longest applicable requirement plus two years of buffer, and to document the derivation. For most labs serving pharma and environmental clients, that lands at 10 years for raw data and 7 years for supporting records. Storage is cheap. Reconstructing a deleted chromatogram is not possible.

The three gaps that show up in almost every audit

I have sat on both sides of the table, and the same three failures recur. They are structural, not personal, which is why they keep happening to competent labs.

Gap one: the raw data is not the raw data

The inspector asks for the chromatogram, you produce a PDF with the integration result printed on it, and she asks for the file the software generated before processing. If you cannot produce it, the run is not documented, because a PDF can be edited and a raw file with an audit trail cannot. The failure mode is almost always a CDS configured to export to PDF and archive the PDF, with the raw files on a local drive that gets cleaned annually. The fix is to configure the chromatography data system to write raw files to a validated archive location at acquisition, with the audit trail enabled and locked. If you are running Agilent or Thermo Fisher systems, both vendors support this out of the box; the problem is that nobody configured it during installation and nobody revisited it. Audit it once a year by pulling a random run from eighteen months ago and proving you can open the raw file with its original audit trail intact.

Gap two: the consumable lot cannot be tied to the run

This gap shows up as a simple question: “Which lot of septa was in use on March 14?” If the answer requires a phone call to the warehouse, you have failed. The record must live in the run record, not in the purchasing system. The practical fix is a consumables log at the bench: when a box of vials or a column is opened, the analyst writes the lot number, date opened, and the run range it covers in a bound logbook, and the same lot number goes into the run worksheet. It takes fifteen seconds per box. It saves you when an out-of-spec peak shape appears and you need to know whether the HPLC vial lot changed between the passing and failing runs. I have seen a lab spend three weeks chasing a “column problem” that turned out to be a septum lot with a different silicone formulation; the lot record would have found it in an hour.

Gap three: the review signature has no evidence behind it

A second-person review signature is worthless if the reviewer cannot show what she reviewed. Inspectors increasingly ask for evidence that the review was substantive: a checklist, a review date, and a record of any corrections made. If your review is a signature on a cover sheet, you have documented that someone was in the room, not that anyone checked the data. The fix is a review record that names the specific items checked (calibration curve, QC acceptance, integration, calculation, transcription) with initials per item and a date. It adds ten minutes per run and it is the single most effective defense against a data-integrity finding, because it demonstrates control rather than asserting it.

Field cases: what the gap looks like when it bites

Case one: the balance printout that was filed by month

Symptom: During a routine inspection, an auditor selected a passing run from eight months earlier and asked for the balance printout supporting the standard weight recorded in the preparation worksheet. We had it, but it was in a monthly binder organized by date, not by run. Retrieval took forty minutes and required two staff.

Measurement that identified the problem: The auditor timed the retrieval. Forty minutes for one record, extrapolated across the sample she intended to pull (twelve runs), would have consumed most of a day. She wrote it up as a records-retrieval control weakness, not a data-integrity finding, which was the difference between a corrective action and a serious citation.

Fix: We changed the filing rule so that every balance printout carries the run number in the header at the time of printing, and the binders are filed by run number within month. Retrieval now averages under two minutes. The change cost nothing except a habit.

Case two: the column log that did not exist

Symptom: A method showed a gradual loss of resolution over six weeks, with the USP resolution between two critical pairs drifting from 2.1 to 1.4. The analyst replaced the column twice with no improvement, which is the classic signature of a problem that is not the column.

Measurement that identified the problem: We pulled the injection logs and found that the failing period correlated with a change in the mobile phase preparation record: a new lot of buffer salt had been opened, and the pH of the aqueous phase had shifted from 3.0 to 3.3 as measured on a calibrated meter. The pH shift changed the ionization of the analytes and compressed the selectivity. We had no column log to check whether the column itself had drifted, because nobody had recorded install dates or plate counts, so we spent two weeks ruling out the column by replacement rather than by record.

Fix: We added a column logbook with install date, serial number, initial plate count and tailing factor, and a monthly plate count check. When resolution drifts now, the first question is answered in thirty seconds: did the column change or did the chemistry change? The log also gives you the data to justify column retirement at a defined plate-count threshold rather than at the analyst’s discretion.

Building the record set so it survives an inspection you did not schedule

The labs that pass cleanly are the ones where the record set is a byproduct of the work, not a separate task. That means the analyst creates the record as she works, in the system that will store it, with the identifiers already attached. Here is the checklist I use to audit a lab’s readiness, and I run it quarterly:

• Reconstruction test: pull one run from each of the last four quarters and rebuild it from records alone. Time the rebuild. Anything over ten minutes per run is a finding.
• Raw file test: open the raw data file, not the PDF, and confirm the audit trail shows the original acquisition timestamp and no post-acquisition edits without justification.
• Consumable traceability test: for the same runs, confirm the vial, cap, septum, and column lot numbers are recorded in the run record.
• Retention test: confirm the archive actually holds files for the stated period. Query the oldest file in the system and compare it to your policy date.
• Review evidence test: confirm the second-person review record names the items checked and carries a date.
• Version control test: confirm the method version cited in the run matches the version in the controlled document system, and that superseded versions are retained.

Two of these deserve a number. For the reconstruction test, I set the threshold at ten minutes per run because that is roughly the time an inspector will tolerate before she starts writing. For the retention test, I check the archive against the policy date and flag any gap over 30 days, because a gap of a month usually means a backup job failed and nobody noticed.

Where the record set connects to the consumables you buy

I keep coming back to consumables because that is where the documentation chain breaks most often, and because it is the part of the record set you can fix by changing a supplier rather than changing a process. If your vial supplier cannot give you a lot-level certificate with the dimensional data (neck finish per GPI 13-425 or GPI 8-425 depending on the vial, septum thickness, and the compression set of the cap), you cannot build artifact five. If your column supplier cannot give you a serial number and a performance certificate, you cannot build the instrument log entry. The lab vials you buy are part of your record set whether you treat them that way or not.

This is also where the external method references matter. If you are running EPA SW-846 methods, the method itself specifies the QC records that must accompany the data package, and the retention period flows from the contract. If you are running USP methods, the record requirements flow from 21 CFR 211.194. The Restek technical resources are useful here because they publish the lot-to-lot performance data that lets you document why a consumable change did or did not affect the method, which is exactly the kind of evidence an inspector wants when she asks why the peak shape shifted.

The decision rule I give every lab manager

If you take one thing from this, take the rule: a record is not compliant until a stranger can reconstruct the run from it in under ten minutes without asking you a question. Apply that test to one run per quarter, fix whatever fails, and you will never be the lab that loses forty minutes on a balance printout. The retention period is the easy part once the reconstruction works, because a record set that can be rebuilt is a record set that was built to be kept. Start with the consumable lot numbers, because that is the gap you can close this week by asking your supplier for a certificate they should already be sending you.

References & further reading

Related resources from our lab consumables network:

• Analytical Testing at testinganalytical.com
• Hplc Vial at vialforhplc.com
• Chromatography Consumables at chromatographyconsumables.com
• Lab Vials at labvials.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography