Three change classes, and only one of them forces full revalidation

We swapped a C18 column from one lot to another in March and the system suitability tailing factor went from 1.04 to 1.31 on the same method, same mobile phase, same 250 x 4.6 mm, 5 µm format. Nothing else changed. That single number is what “revalidation triggers when a lab chromatography supplier or lot changes” actually looks like on the bench: not a regulation to read, but a decision you have to make by Friday about whether the run counts. The right answer depends on whether the change is a supplier identity change, a lot-to-lot drift within a qualified supplier, or a cosmetic change that touches nothing analytical. Get that classification wrong in either direction and you either revalidate for no reason or ship data an inspector will reject.

I classify every chromatography change into one of three buckets before anyone touches a document. The classification drives the work, and the classification itself must be written down with a reason.

Class 1: revalidation. The change alters a variable the method explicitly controls or that the validation study demonstrated was not critical. New column chemistry, new particle size, new dimensions, a different supplier of the column when the method names the supplier, a change in mobile phase modifier grade from HPLC to ACS, a switch from a PTFE to a nylon filter in a sample prep step the method names. These require a partial or full revalidation of the affected performance characteristics: specificity, linearity, accuracy, precision, LOD/LOQ if the method reports them, and robustness.

Class 2: verification only. The change is within a qualified supplier and within a specification the method already bounds. New lot of the same column part number, new lot of the same reference standard, new batch of the same mobile phase salt from the same vendor. You run system suitability plus a bracketing check against a retained standard. If it passes, you document and move on. No revalidation.

Class 3: documentation only. Packaging, label, carton count, catalog number suffix that maps to the same physical product. No analytical work. Log it and move on.

The failure mode I see most is labs treating a Class 2 lot change as Class 1 and burning three weeks, or treating a Class 1 supplier change as Class 2 because the part number “looks the same.” The second error is the dangerous one. A part number is not a specification.

What actually makes a supplier change a revalidation trigger

Here is the distinction that matters and that most change-control forms miss: a supplier change is a revalidation trigger when the supplier controls a critical material attribute the method relies on but does not name. Column chemistry is the obvious one. So is the endcapping strategy on a C18, the pore size (a 100 Å versus a 300 Å column with the same part number pattern behaves differently on large molecules), the base silica metal content (trace metals at 5 ppm versus 30 ppm shift peak shape for chelating analytes), and the surface area.

If your method was validated on a column from one manufacturer and the validation report names that manufacturer’s part number, a switch to another manufacturer is Class 1. Full stop. Even if the phase is nominally “the same C18.” The retention factor k’ can move 15 to 25 percent with the same nominal chemistry because the bonded phase density and endcapping differ. That is not a cosmetic difference.

Where it gets subtle: the supplier of a consumable the method does not name. A syringe filter. A vial. A septum. If the method says “filter through a 0.45 µm membrane” and does not name the polymer, then a switch from PTFE to nylon is a change you have to evaluate, because nylon adsorbs and PTFE does not. This is where I point people at the PTFE membrane filter specification sheet and make them read the extractables line, not the pore size line. A 0.45 µm PTFE filter and a 0.45 µm nylon filter are not interchangeable for a method that quantifies a basic analyte at 0.1 percent.

The consumables that quietly change a method

I keep a short list of consumables that have revalidation potential even though nobody puts them in the method text:

• Septum on the GC inlet. A change in septum material or a change in the lot can move siloxane bleed. If your method reports at low ppb, a new septum lot that sheds more cyclic siloxanes is a real change. I have seen a 3x increase in the 207/281 bleed ions from a septum supplier change alone.
• Liner deactivation. A switch from a deactivated liner to a plain glass liner changes peak shape for polar analytes. That is Class 1 for a method that reports polar compounds.
• Vial and cap. A change in septa durometer or in the liner of a crimp cap can change the headspace partition. For headspace GC, a cap change is a revalidation question, not a procurement question.
• Mobile phase water. A switch from one water purification system to another, or a change in the cartridge, can move the UV baseline and the gradient. I treat a water system change as Class 2 with a mandatory system suitability run and a blank injection.

None of these are in the method text. All of them are in the validation study as “materials used.” An inspector who reads the validation report and then reads your change log will find the gap if you did not classify them.

The verification protocol that lets you skip revalidation

You earn the right to skip revalidation by running a defined verification and passing it. The protocol has to be written before you run it, and it has to have a pass/fail number. A verification with no acceptance criterion is not a verification, it is a hope.

For a Class 2 lot change on a column, my standard verification is:

• System suitability on the new lot: tailing factor ≤ 2.0, resolution ≥ 2.0 between the critical pair, theoretical plates ≥ 2000 (or the validated value), and %RSD of five replicate injections ≤ 2.0 percent.
• One bracketing standard injection before and after a single sample run, agreement within ± 2.0 percent of the assigned value.
• Retention time of the main peak within ± 5 percent of the validated value.
• A blank injection to confirm no new ghost peaks above the validated LOD.

If all four pass, the lot is qualified, the change is closed as Class 2, and you write one paragraph in the change record. If any fail, it escalates to Class 1 and you revalidate the affected characteristic. This is the same logic the general lab chromatography qualification workflow uses, applied at the lot level instead of the instrument level.

For a filter change, the verification is different: run the method with and without the filter on a spiked sample at the LOQ, and confirm recovery within 95 to 105 percent and no shift in retention time greater than 0.05 minutes. If the analyte is basic and the filter is nylon, expect to fail this and expect to need the PTFE.

A lot-change decision table you can paste into the SOP

This is the table I use. The right column is the work, the middle column is the acceptance number. Copy it, put your own validated values in the acceptance column, and make the change-control form reference it.

The table is not the point. The point is that every row has a number in the acceptance column. An inspector reads the number, not the prose.

Two field cases where the classification decided the outcome

Case 1: the tailing factor that was a base silica change

A contract lab running a USP method for a small basic molecule saw tailing factor drift from 1.05 to 1.28 over three column lots. The method named the phase but not the manufacturer. The lab had been buying from two suppliers interchangeably because the part number matched. The measurement that identified it: they pulled the certificates and compared the base silica metal content. One supplier’s silica was 12 ppm total metals, the other was 28 ppm. The basic analyte was chelating to the higher-metal silica. The fix was to qualify one supplier, add the supplier name to the method materials section, and run a Class 2 verification on every subsequent lot. They also added a system suitability check at 1.15 tailing as an early warning. That is a supplier change that should have been Class 1 from the start and was being treated as Class 2 because the part number matched.

Case 2: the filter that ate the analyte

A method for a hydrophobic amine at 0.05 µg/mL used a 0.22 µm filter. Procurement switched from PTFE to a nylon filter to save a few cents per unit. Recovery dropped to 71 percent on the spiked LOQ sample. The measurement that caught it: the verification spike, run at the LOQ, not at mid-range. At mid-range the loss was invisible because the analyte concentration overwhelmed the adsorption sites on the membrane. The fix was to specify PTFE in the method materials section and to run the LOQ spike as part of the lot verification. The general HPLC filter selection logic applies here: polymer first, pore size second. A 0.22 µm nylon filter is not a drop-in for a 0.22 µm PTFE filter.

Writing the change-control entry so an inspector accepts it

The change record is where most labs lose the argument. The analytical work can be perfect and the record can still fail because it does not answer the four questions an auditor asks in order. Write the entry so each question has its own sentence.

1. What changed, specifically. Not “column lot change.” Write the old lot number, the new lot number, the part number, the manufacturer, the certificate number, and the date received. If the change is a supplier change, name both suppliers.
2. Why you classified it the way you did. One sentence that names the critical attribute. “Classified as Class 2 because the phase chemistry, particle size, and dimensions are unchanged and the supplier is qualified under SOP-xxx.” That sentence is what an inspector reads to decide whether you thought about it.
3. What you ran and what it returned. The verification protocol reference, the acceptance criteria, and the actual numbers. Tailing 1.08, resolution 2.4, %RSD 0.6, bracketing standard 99.8 percent. Numbers, not adjectives.
4. What you concluded and who approved it. Pass or fail, the disposition, the approver’s name and date, and the effective date of the change.

The one thing that gets labs in trouble: writing the classification reason after the fact, in language that matches whatever the result turned out to be. An inspector can tell. The record has to show the classification was made before the work, which means the form needs a timestamp or a sequential log number that precedes the run date.

Where the standards attach to the record

The change-control entry is not a standalone document. It attaches to the validation report under EPA SW-846 methods when the lab runs environmental work, and to the method validation section under ICH Q2(R2) for pharmaceutical work. For a method that cites a SW-846 number, the method text is the contract and the change record has to show the change did not move the method outside its stated performance. If you cite a manufacturer’s application note as the basis for a method, the change record should reference the note’s revision. I keep the Restek technical resources and the Agilent column and consumable documentation bookmarked because the certificates and the phase specifications are what you cite in the record. The Thermo Fisher chromatography and sample handling pages are useful for the same reason: the specification tables are the evidence.

Revalidation scope: what to redo and what you can leave alone

When a change escalates to Class 1, the instinct is to revalidate everything. That is wasteful and it also creates risk, because a full revalidation on a method that has been running for years can surface old issues you now have to resolve. Scope it to the affected characteristics.

For a column chemistry change, revalidate specificity, linearity, accuracy, precision, and the system suitability limits. Leave robustness alone unless the change touches a robustness variable (pH, temperature, flow). For a filter change, revalidate accuracy at the LOQ and precision, and leave linearity alone because the filter does not affect the calibration curve shape. For a mobile phase grade change, revalidate the gradient and the baseline, and recheck the UV cutoff against the detection wavelength.

The scoping decision goes in the change record too. “Revalidated specificity, linearity, accuracy, and precision. Robustness not repeated because the change does not touch a robustness variable.” That sentence saves you an argument six months later.

Operational precautions that keep a change from becoming a deviation

A few habits that keep the change-control system from generating deviations:

• Quarantine the new lot until the verification passes. Do not put it into the analyzer’s hands before the paperwork is signed. I have seen a lot go onto the instrument on a Friday and the verification run on a Monday.
• Run the bracketing standard on the same instrument and the same day as the new lot. A bracketing standard from last week is not a bracketing standard.
• Keep a retained sample of the old lot. When the new lot fails, you need the old one to prove the method still works and the failure is the lot.
• Check the certificate of analysis against the specification, not just the lot number. A certificate that lists a 0.5 percent difference in carbon load is telling you something.
• For GC septa, track puncture count. A septum rated for 100 punctures at 250 °C is not a 200-puncture septum. If your autosampler is running 150 injections a day, you are changing septa more often than you think, and each change is a Class 2 event.

The puncture count and the torque value on the inlet nut (typically hand-tight plus a quarter turn, and follow the manufacturer’s spec, not your wrist) are the two numbers that decide whether a septum change introduces a leak or a bleed artifact. Both belong in the change record when the septum is the thing that changed.

Troubleshooting checklist when a lot change fails verification

When the verification fails, work through this in order before you declare the lot bad:

• Re-run the system suitability on the old lot. If the old lot also fails, the problem is the instrument or the mobile phase, not the new lot.
• Check the mobile phase preparation date and pH. A two-week-old buffer is a suspect.
• Check the lamp hours and the detector baseline. A drifting baseline mimics a failing column.
• Check the injection volume and the sample temperature. A change in the autosampler tray temperature changes the headspace.
• Only then compare the new lot’s certificate against the specification and against the old lot’s certificate. Carbon load, metal content, pore size, endcapping.

If the old lot passes and the new lot fails on the same day with the same mobile phase, the lot is the variable. Escalate to Class 1, revalidate the affected characteristic, and if the new lot cannot meet the validated limits, reject the lot and requalify the supplier. That last step is a procurement decision, and it is the reason the change-control record needs a procurement signature, not just a QA signature.

The decision rule

Classify every chromatography change before you run anything, and let the critical material attribute decide, not the part number. If the change touches a variable the method controls or that the validation study relied on, it is a revalidation. If it is inside a qualified supplier and inside a specification the method bounds, it is a verification with a written pass/fail number. Write the classification, the reason, the numbers, and the approver into the change record on the day the change arrives, and the inspector will accept it. Skip any one of those four and no amount of revalidation will fix the record.

References & further reading

Related resources from our lab consumables network:

• Lab Chromatography at labchromatography.com
• Hplc Filter at hplcfilters.com
• Ptfe Membrane Filter at ptfemembranefilter.com
• Cod Vial at codvial.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