Last February, a contract lab lost its ISO 17025 accreditation for a single method because an auditor asked for the traceability chain on a 0.45 µm PTFE syringe filter. The lab had the certificate of analysis from the filter vendor. They had the purchase order. They did not have the link that connected the filter lot used in every validation run to the filter lot sitting in the drawer during the audit. Three weeks later, the method was still valid on paper. The accreditation was not.
ISO 17025 traceability for hplc filter is not about whether the filter is clean. It is about whether you can prove which filter lot touched which sample, which standard, and which validation data set. Most labs cannot produce that proof. This article covers what an unbroken chain actually looks like for a consumable, which link breaks first, and what to change before your next audit.
The traceability chain for a consumable has five links, and most labs only keep two
An unbroken traceability chain for an HPLC filter looks like this:
- Link 1: The filter manufacturer’s certificate of analysis, tied to a specific lot number, with acceptance criteria and test results.
- Link 2: Your receiving record: purchase order, packing slip, and the lot number you physically received.
- Link 3: The internal inventory record that tracks where that lot went: which analyst, which instrument, which date.
- Link 4: The method or run log that ties the filter lot to specific sample IDs, standard preparations, and injection sequences.
- Link 5: The data file or validation report that references the same lot number, so an auditor can move from raw data back to the physical filter in under five minutes.
Most labs have Link 1 and Link 2. A few have Link 3. Almost none have Link 4 and Link 5 for a filter. That is the problem. The filter is treated as a generic supply item, not as a component in the measurement chain. But ISO 17025 clause 6.4.6 says equipment and consumables that can affect the validity of results must be traceable. A filter that changes recovery by 15% affects validity. So does a filter that sheds particles into the flow path.
The link that breaks first is Link 3. It breaks because filters are stored in a common drawer, pulled by whoever needs one, and never logged. The second break is Link 4: the run log says “filtered sample” but does not record the lot. Without those two links, Link 1 and Link 2 are useless. A certificate of analysis for a lot you cannot connect to a run is just a piece of paper.
What an auditor actually asks for, and why the certificate is not enough
An ISO 17025 auditor will not ask “do you trace your filters?” They will pick a data point. Usually a calibration standard, a QC sample, or a validation run. Then they will say: show me every consumable that touched this sample, and show me the lot number for each one. The filter is on that list. When you cannot produce the lot number, the finding is written as a failure of clause 6.4.6 or 7.2.1.5, depending on the auditor’s interpretation.
The certificate of analysis from the filter vendor is necessary but not sufficient. It proves the filter lot met the manufacturer’s specification when it left the factory. It does not prove that the filter you used in your method was from that lot. It does not prove the filter was stored correctly after receipt. It does not prove the filter was not expired, contaminated, or swapped. The auditor knows this. So the certificate buys you about thirty seconds of conversation.
What the auditor wants is a chain that moves from the certificate to the physical filter to the run log to the data file. Each step must be dated, signed or initialed, and cross-referenced by lot number. For a filter, that means you need a receiving log, a usage log, and a notation in the method or run record. Three documents. Most labs do not have even one of them.
The filter lot number is the only unique identifier you have, so use it
Every reputable filter manufacturer prints a lot number on the unit package and on the case. The lot number is the only field that connects a certificate to a physical item. Without it, you have nothing. With it, you can build a chain in under an hour per lot.
Start with the receiving step. When a case of Hplc Filter units arrives, log the lot number, quantity, date received, and the certificate of analysis file location. This is Link 2. It takes two minutes per case. If you use a Ptfe Membrane Filter for aggressive solvent work, log that lot separately. PTFE filters have different failure modes than nylon or PVDF, and the lot number is the only way to trace a solvent compatibility issue back to a specific batch.
Next, create a filter usage log. A simple spreadsheet or a page in the instrument logbook works. Columns: date, analyst, filter lot number, instrument ID, method ID, sample or standard IDs. When an analyst pulls a filter, they write the lot number. This is Link 3. It takes ten seconds per use. The resistance is always “we do not have time.” You have time to rerun a failed sequence, which costs more.
Finally, add the filter lot number to the method or run record. If your LIMS has a consumables field, use it. If not, add a line in the sample prep section: “filtered through 0.45 µm nylon, lot 240118A.” This is Link 4. It is the link that lets an auditor move from the data file back to the filter drawer in under five minutes.
Two field cases where a missing filter lot number caused a real problem
Case 1: The recovery drift that took four weeks to diagnose. A pharmaceutical QC lab ran a USP assay for a small-molecule API. Recovery on the system suitability standard dropped from 98.2% to 91.4% over three weeks. The column was replaced twice. The mobile phase was remade three times. The autosampler was serviced. Nothing fixed it. Finally, an analyst noticed that the filter lot in the drawer had changed. The old lot was a 0.45 µm nylon filter with a polypropylene housing. The new lot was the same part number from a different manufacturing site, and the housing material had a different extractables profile. The method had been validated with the old lot. The new lot was causing a matrix effect in the sample prep step. The fix was to qualify the new lot before use. But because the filter lot was not logged, the lab spent four weeks and roughly $18,000 on column replacements and instrument service before finding the problem.
Case 2: The audit finding that survived a CAPA. An environmental lab running U.S. EPA SW-846 test methods for semi-volatiles was audited. The auditor selected a batch of soil extracts and asked for the filter lot. The lab had the certificate of analysis. They had the purchase order. They could not show which filter lot was used for that batch. The finding was written against clause 6.4.6. The lab’s corrective action was to create a filter usage log. But the CAPA only applied to new lots going forward. The auditor returned six months later, selected a batch from the previous year, and the lab still could not produce the link. The finding was escalated. The lesson: traceability is retroactive. You cannot fix it going forward only. You must be able to produce the chain for any data you claim is valid.
Which filter parameters actually matter for traceability, and which ones do not
Not every filter parameter needs to be traced. The ones that matter are the ones that affect the measurement. Here is a decision checklist:
| Parameter | Traceable? | Why it matters |
|---|---|---|
| Lot number | Yes, always | Unique identifier for the chain |
| Pore size (0.22 µm, 0.45 µm) | Yes | Affects recovery of analytes near the size cutoff |
| Membrane material (nylon, PTFE, PVDF, PES) | Yes | Affects analyte binding, solvent compatibility, extractables |
| Housing material (polypropylene, acrylic, glass) | Yes | Affects extractables and pressure tolerance |
| Sterility (sterile vs. non-sterile) | Yes, if method requires it | Sterile filters have different packaging and handling requirements |
| Certificate of analysis | Yes | Links lot to manufacturer’s acceptance data |
| Expiration date | Yes | Expired filters may have degraded membrane or housing |
| Color of housing | No | Cosmetic, does not affect measurement |
| Packaging format (bulk vs. individually wrapped) | No, unless method specifies it | Affects handling but not the filter’s physical properties |
The parameters that matter are the ones that can change between lots. A filter manufacturer can change the housing polymer, the membrane casting process, or the wetting agent without changing the part number. The certificate of analysis may not flag these changes. The lot number is the only way to detect that a change happened. If you do not log the lot number, you cannot see the change until your data drifts.
How to build a filter traceability system that survives an audit
You do not need a LIMS module. You need three documents and a habit. The documents are:
- Receiving log: date, lot number, quantity, certificate file location, received by. One line per case.
- Usage log: date, analyst, lot number, instrument, method, sample or standard IDs. One line per use.
- Run record notation: filter lot number in the sample prep section of the method or in the LIMS consumables field.
The habit is the hard part. The habit is: no lot number, no filter. If an analyst cannot write down the lot number because the package is already in the waste bin, that is a process failure. Put a small notepad next to the filter drawer. Put a barcode scanner on the bench if your filters have barcodes. Make the lot number part of the sample prep checklist. Do not accept “I will log it later.” Later does not happen.
For labs that use Vial Insert products or Cod Vial consumables, the same rule applies. Any consumable that touches the sample before injection is part of the measurement chain. The filter is the most common one to be missed because it is used once and thrown away. But the insert and the vial are also consumables. If your method is sensitive to glass surface chemistry, the insert lot matters. Log it the same way.
What a filter certificate of analysis should contain, and what to do when it does not
A useful certificate of analysis for an HPLC filter includes: the lot number, the pore size, the membrane material, the housing material, the bubble point or pressure test result, the extractables test result if applicable, and the date of manufacture. If the certificate does not include the lot number, it is not a certificate of analysis. It is marketing material. Send it back.
If the certificate is missing extractables data, ask the vendor for it. Thermo Fisher chromatography and sample handling and Agilent chromatography consumables and columns both publish extractables data for their filter lines. If your vendor cannot produce it, that is a red flag. A filter with unknown extractables is a filter that can contaminate your sample in ways you cannot predict or trace.
For regulated methods, you may need to qualify each new filter lot before use. A simple qualification is: filter a blank mobile phase through the new lot, run it on the HPLC, and compare the baseline to the previous lot. If the baseline noise increases by more than 20%, investigate. If a new peak appears, identify it. This is a ten-minute check that can save you from a four-week investigation like the one in Case 1.
Operational precautions that keep the chain intact
Filter handling affects traceability. If an analyst touches the membrane with bare fingers, the filter is contaminated before it is used. The lot number will not help you trace a contamination event that was caused by handling. So the precautions are:
- Store filters in the original packaging until use. Do not dump them into a common drawer.
- Use forceps or wear powder-free gloves when handling the filter. Finger oils contain compounds that can appear as ghost peaks.
- Do not use a filter past its expiration date. The membrane can degrade, especially for PTFE and nylon.
- Do not exceed the pressure rating of the filter housing. A polypropylene housing typically fails above 75 psi. A syringe filter used with a manual syringe can generate over 100 psi. If you are filtering a viscous sample, use a larger diameter filter or a lower flow rate.
- Do not reuse a syringe filter. A 13 mm filter with a 0.45 µm membrane is designed for one sample, roughly 10 to 50 mL depending on the sample. After that, the pores are clogged, the recovery drops, and the filter itself becomes a source of contamination.
- Record the puncture count if you are using a septa-piercing filter. A filter with a needle on the outlet side is designed for a specific number of vial punctures. Exceeding that number tears the septum and introduces particulates.
These precautions are not optional. They are part of the traceability chain because they affect whether the filter performs the same way in your lab as it did in the manufacturer’s test. If you mishandle the filter, the certificate of analysis no longer describes the filter you used. The chain is broken by your own hand.
The decision rule for filter traceability
For any method that is accredited under ISO 17025, the filter is a consumable that affects the validity of results. That means it must be traceable. The decision rule is simple: if you cannot produce the filter lot number for a given data set within five minutes, you are not compliant. Not almost compliant. Not compliant enough. Not compliant.
Build the three documents. Log the lot number at receiving. Log it again at use. Put it in the run record. Qualify new lots before they touch a sample. Treat the filter like a column, not like a paper towel. The filter is the cheapest component in the flow path, but it is the one that breaks the accreditation chain most often. Because it is cheap, it is ignored. Because it is ignored, it is not logged. Because it is not logged, the chain breaks. That is the entire problem in three sentences.
References & further reading
External references and standards: