The auditor pulled one thing from the file before she looked at a single chromatogram: the sample container log. Lot 22-0417, four 40 mL VOA vials, received Tuesday at 4.1°C. The log said 4.1°C. The cooler thermometer said 6.8°C. That 2.7-degree gap cost the lab three re-samples and a finding on the corrective action report, and it had nothing to do with the column, the method, or the instrument. It was a documentation failure on the front end of an EPA method, which is exactly where most of them happen.
If you run EPA methods, the HPLC filter in your sample path is a regulated component whether your SOP treats it that way or not. The EPA method requirements for an hplc filter and auditor expectations are not written as a filter specification. They are written as sample integrity, holding time, and chain of custody, and the filter is the physical object that either preserves or compromises all three. This article is about what SW-846 and the method-specific procedures actually say, and what a data reviewer asks for before they ever open your raw data.
SW-846 is a guidance compendium, not a regulation with a parts list. It tells you to prepare and analyze samples by a specified procedure, and it tells you the sample must be representative when it reaches the detector. It does not tell you which membrane, which pore size, or which housing to use. That silence is where inexperienced labs get comfortable, and where experienced auditors get specific. The absence of a filter clause is not permission to use anything. It is a requirement that you can defend whatever you used.
The relevant SW-846 chapters are Chapter One (quality control), Chapter Three (inorganic analytes, including metals and, by extension, the filtration steps in aqueous prep), and the individual 3000-series and 3500-series methods for organics. Read Chapter One’s QC section and you will find the language that governs filters without saying the word: the requirement for a method blank carried through the entire preparation, the requirement for a laboratory control sample, and the requirement that any deviation from the written procedure be documented and justified. A filter is a deviation the moment it is not in your SOP.
Here is the practical translation. If your SOP says “filter through a 0.45 µm PTFE membrane,” the auditor will ask which lot, which manufacturer, and whether the membrane was tested for analyte binding. If your SOP says “filter as needed,” you have already failed the documentation check, because “as needed” cannot be audited. The SW-846 test methods page is where you confirm the current revision of the method you cite, because auditors work from the revision in force at the time of analysis, not the one you learned in training.
What “representative” actually requires of the filter
The filter must not add, remove, or transform the analyte of interest. For a 0.45 µm PTFE membrane in an aqueous matrix, that means three things you can measure. First, analyte recovery through the filter must be within the method’s acceptance window, typically 70 to 130% for organic analytes and tighter for metals. Second, the filter must not shed extractable material that shows up as a ghost peak at your retention time. Third, the filter must not adsorb the analyte onto its surface. PTFE is chosen for aqueous and aggressive matrices because it is chemically inert and hydrophobic, but a hydrophobic membrane can hold back nonpolar analytes if you push a mostly aqueous sample through it dry. That is why the PTFE membrane filter you select for aqueous work should be pre-wetted or the method should specify a hydrophilic variant. I have watched a lab lose 18% of a semivolatile recovery to a dry PTFE membrane and spend two weeks blaming the extraction.
The sample container clause is the first thing an auditor reads
Every method in the 3500 series specifies the container, the preservative, the headspace, and the holding time. For volatile organics, the container is a 40 mL amber VOA vial with a PTFE-lined septum, zero headspace, and a 14-day holding time from collection to analysis if preserved with HCl to pH under 2 and stored at 4°C. For semivolatiles, the container is typically 1 L amber glass, extracted within 7 days and analyzed within 40 days of extraction. Those numbers are not suggestions. They are the audit checklist.
The auditor’s first question is not “did you filter.” It is “show me the container log.” They want to see the collection date, the receipt temperature, the preservation check, and the analysis date, and they want the arithmetic to close. A sample received at 4.1°C on day one and analyzed on day 12 passes. A sample received at 6.8°C fails the receipt criterion even if every downstream number is perfect, because the preservation was not demonstrated at receipt. This is the failure mode that catches labs that run clean chromatography but sloppy intake.
Where the filter enters this picture is the subsampling step. If you pour from the 40 mL vial into a filter housing and then into an autosampler vial, you have opened the container, exposed the sample to headspace, and created an unaudited transfer. For volatile methods, the correct move is to filter inline or to use a vial that accepts the filter directly so the sample never sees atmosphere. The COD vial format is worth knowing here because the same zero-headspace discipline that governs COD digestion applies to any method where oxidation state or volatiles matter. If your method requires a small-volume final extract, a vial insert keeps the transfer volume down and the headspace predictable, which is one fewer variable for the auditor to challenge.
What the auditor asks for first, in order
I have sat through enough audits to know the sequence is almost always the same. The reviewer works from the outside in: paperwork, then preparation records, then raw data, then the instrument. If the paperwork does not close, they never reach the chromatogram. Here is the order I prepare for, and the document that answers each request.
| Audit request | Document that answers it | Common failure |
|---|---|---|
| Container and preservation log | Receipt log with temperature and pH | Temperature recorded from cooler, not vial |
| Holding time compliance | Collection date vs. extraction/analysis date | Extraction date logged, analysis date missing |
| Chain of custody | Signed COC with every transfer | Relinquish signature blank on internal transfer |
| Method blank and LCS | QC summary carried through prep | Blank filtered separately from samples |
| Filter lot and specification | SOP plus lot certificate | SOP says “0.45 µm” with no membrane type |
| Deviation log | Documented and justified exceptions | Deviation noted in bench notebook only |
The filter lot request is the one that surprises labs. The auditor is not asking for a certificate of analysis because they doubt the membrane. They are asking because the lot number on the certificate must match the lot number in your prep record, and in a surprising number of files it does not. If you cannot tie the filter you used to a lot, you cannot demonstrate the method blank reflects the same materials the samples touched.
Method 3510 and the filtration step nobody documents
Method 3510 (separatory funnel liquid-liquid extraction) and 3520 (continuous liquid-liquid extraction) both begin with a filtration or clarification step for samples that carry particulates. The methods assume the analyst removes solids before extraction. They do not specify the filter, and that is the gap. If you filter a 1 L aqueous sample through a 47 mm, 0.45 µm membrane and the membrane loads with sediment, you have removed analyte along with the solids unless you extract the filter or account for the loss. The auditor who knows the method will ask whether the filter was extracted with the sample or discarded. If it was discarded, your recovery data has an unquantified bias.
The fix is procedural, not analytical. Write the filter into the method as part of the extraction, specify the membrane material and pore size, and state whether the filter is extracted or discarded. For a 47 mm disc in a vacuum manifold, a 0.45 µm pore is the default for aqueous work, but a loaded membrane at high solids will blind and force you to a 1.0 µm prefilter ahead of the 0.45 µm final. That two-stage stack is a defensible deviation if it is in the SOP. It is a finding if it is not.
Field case: the blank that came back at 0.9 ppb
A lab running Method 3510 for semivolatiles started seeing a method blank at 0.9 ppb for a phthalate on every batch. The samples were clean. The instrument was clean. The blank was filtered through the same membrane lot as the samples, and the membrane was the source: the filter housing carried a plasticizer that leached into the aqueous matrix during the 10-minute filtration. The measurement that identified it was a blank filtered through a membrane from a different lot, which came back at 0.1 ppb. The fix was a switch to a PTFE membrane with a documented low-extractable housing and a lot screening step: filter a reagent water blank through each new lot before it enters the prep batch. That screening step now appears in the SOP, which is what turned a finding into a controlled procedure.
Holding time and the filter’s effect on the clock
Holding time starts at collection, not at filtration. A sample collected Monday and filtered Friday has already spent four days of its 14-day window doing nothing, and the filtration step does not reset the clock. Auditors check this because labs sometimes treat prep as a reset. They do not. If your filter step adds a day because the membrane has to be conditioned or the manifold has to be cleaned, that day comes out of the analytical window.
The filter also affects the clock indirectly through preservation. A sample that is filtered and then stored in a partially filled vial has headspace, and headspace in a preserved volatile sample is a preservation failure. The 40 mL zero-headspace requirement exists because volatiles partition into the headspace and the concentration in the liquid phase drops. If your workflow filters into a smaller vial, you have created headspace unless the vial is filled to the neck. That is why the final container matters as much as the filter, and why the vial you filter into should be sized to the volume, not to convenience.
What the filter certificate needs to say, and what it does not
Auditors ask for the filter lot certificate, but they are not looking for the same thing a purchasing manager looks for. They want to see that the lot was manufactured to a stated specification and that the specification matches the SOP. A certificate that lists pore size, membrane material, and housing material is enough. A certificate that lists a full extractables profile is more than the method requires, and a certificate that lists neither is a finding waiting to happen.
The specification table below is the one I keep in the SOP for aqueous EPA work, because it answers the auditor’s question before they ask it. The numbers are the ones I can defend from method language and from my own lot screening.
| Parameter | Specification for aqueous EPA work | Why the auditor cares |
|---|---|---|
| Pore size | 0.45 µm nominal, 1.0 µm prefilter for high solids | Method convention; blinding causes analyte loss |
| Membrane material | PTFE for aqueous and aggressive matrices | Inertness; low extractables |
| Housing | Polypropylene, no plasticizer | Blank contamination source |
| Diameter | 47 mm disc for 1 L; 13 mm for autosampler volumes | Loading capacity vs. volume filtered |
| Lot screening | Reagent water blank under 0.2 ppb target analyte | Demonstrates blank is representative |
| Temperature rating | Stable to 100°C for hot aqueous matrices | Membrane integrity at prep temperature |
One number worth adding to the SOP that most labs omit: the maximum volume per membrane. A 47 mm, 0.45 µm PTFE disc will pass roughly 500 mL of clean aqueous sample before flow drops appreciably, and far less with sediment load. If your method filters 1 L through a single disc, you are either changing discs mid-sample (document it) or you are forcing the last 400 mL through a blinded membrane at elevated pressure, which is how you rupture a membrane and send solids downstream. The Agilent consumables catalog and the Thermo Fisher sample handling pages both publish volume and pressure guidance by membrane type, and it is worth pulling the numbers into your SOP rather than guessing.
Field case: the 12% recovery that was a membrane rupture
A different lab had a semivolatile recovery of 12% on a batch that had passed every previous round. The extraction was fine, the spike was fine, the instrument was fine. The measurement that found it was a visual inspection of the filter housing after filtration: the membrane had a pinhole near the edge, and the filtrate carried fine sediment that the extraction then failed to recover efficiently. The cause was over-pressure on a blinded disc during the last third of a 1 L filtration. The fix was a 1.0 µm prefilter ahead of the 0.45 µm final and a pressure limit written into the SOP at 15 inHg on the vacuum manifold. Recovery came back to 94%. The auditor’s question, had they seen the original file, would have been “where is the filter inspection record,” and the answer would have been “there isn’t one.”
Chain of custody is where the filter record lives or dies
The filter is a consumable that touches the sample, so it belongs in the custody trail the same way a container does. That does not mean a signature for every membrane. It means the prep record names the filter lot, the analyst initials the filtration step, and the date matches the extraction date. If the filter lot changes mid-batch, the record shows it. An auditor who finds a single lot number across a batch that spanned two weeks will ask whether the lot was actually constant or whether the record was backfilled. Both answers are bad if the second is true.
The custody question gets sharper for methods that require the sample to be filtered in the field. If your filter step happens at the collection site, the filter lot and the filtration time go on the COC, not just the bench sheet. That is a level of documentation many labs skip, and it is the first thing a reviewer checks when a holding-time question arises.
The SOP language that survives review
I have rewritten enough SOPs after findings to know what auditors accept. They accept a procedure that names the filter, states the pore size and membrane material, states the volume per membrane, states the pressure limit, states whether the filter is extracted or discarded, and states the lot screening criterion. They do not accept “filter through a suitable membrane.” That phrase is the single most common cause of a documentation finding in EPA work, because it hands the decision to the analyst at the bench and leaves no trace of what was decided.
Write the SOP so a new analyst can follow it without asking a question, and so an auditor can read it without asking one either. The HPLC filter you standardize on should appear by name and specification, and the specification should match the certificate the vendor ships with the lot. If the vendor cannot supply a lot certificate with pore size and membrane material, that vendor is a finding. The Restek technical library is a reasonable cross-check for membrane selection and extractables data when you are writing the specification, and it is the kind of reference an auditor recognizes.
The decision rule for EPA-facing filter selection
If the method names a filter, use that filter and document the lot. If the method is silent, which most SW-846 methods are, the rule is this: choose a membrane that is chemically inert to the matrix, screen every lot against a reagent water blank before it enters a prep batch, write the pore size, material, volume per membrane, and pressure limit into the SOP, and record the lot on the prep sheet so the blank and the samples share a traceable source. The auditor is not testing your chromatography. They are testing whether your paperwork can prove the sample you analyzed is the sample you collected, and the filter is one of the few components that touches both ends of that chain. Get the lot number on the sheet and the temperature on the vial right, and the rest of the file holds together.
References & further reading
Related resources from our lab consumables network:
- Hplc Filter at hplcfilters.com
- Ptfe Membrane Filter at ptfemembranefilter.com
- Cod Vial at codvial.com
- Vial Insert at hplcinsert.com
External references and standards: