I had a lot of 17,400 injections on a batch of 9 mm natural rubber septa when the auditor asked for the traceability chain on the batch certificate. The certificate had a lot number, a cure date, and a signature. It did not have a link to a calibrated thickness gauge, a reference standard for the durometer reading, or a temperature record for the autoclave cycle that preceded the cure. The auditor wrote it up as a nonconformity against ISO/IEC 17025:2017 clause 6.5 (metrological traceability). The injections were fine. The paperwork was not. That gap, between a certificate that looks like a record and a chain that survives scrutiny, is what ISO 17025 traceability for hplc septa actually turns on, and it is the part most labs cannot produce on demand.
The traceability chain for a consumable has more links than the certificate shows
A septum is not a measuring instrument, so labs assume 17025 metrological traceability does not apply. It does, but not in the way it applies to a balance or a pipette. The clause requires that measurement results be traceable to the SI through an unbroken chain of calibrations, each with a stated uncertainty. For a septum, the “measurement results” that matter are the ones your method depends on: the septum’s dimensional conformance (thickness, diameter, penetration depth), its physical properties (durometer, compression set), and the extractables profile that feeds into your blank and your detection limit. Each of those is a property with a number, and every number has to be traceable to a reference.
The chain, in practice, has six links:
- Reference standard: the certified reference material or reference artefact against which the property is measured. For thickness, a gauge block traceable to NIST or an equivalent national metrology institute. For durometer, a certified rubber block of known hardness.
- Measuring instrument: the gauge, micrometer, durometer, or balance, with a current calibration certificate that itself names a reference standard.
- Calibration event: the date, the technician, the as-found and as-left readings, and the uncertainty budget. A certificate that says “calibrated” without as-found data is a link you cannot defend.
- Production record: the batch record that ties the instrument reading to a specific lot of septa. Without this, the calibration is a fact about the gauge, not about your septa.
- Certificate of analysis: the document that carries the lot forward to the customer, with the measurement values and the instrument references on it.
- Receiving record: your incoming inspection or supplier qualification that confirms the lot you received is the lot the certificate describes.
Five of those six links are routine. The one that breaks is the fourth: the production record that connects a calibrated reading to a specific lot. Suppliers who mold septa in high volume often record a lot number and a pass/fail, not a measured value with an instrument ID. When your auditor asks “show me the calibration that supports the thickness on this CoA,” the answer is a gauge block certificate for a gauge that may or may not have touched this lot. That is the link most labs cannot produce.
What clause 6.5 actually requires of a supplier’s CoA
Read 6.5.2 carefully. It requires traceability for “measurement results” and lists the routes: calibration by a competent laboratory, certified reference materials, or “mutual consent” between parties on a documented standard. The last route is the one that saves most consumable supply chains, and it is underused. If you and your supplier agree in a written specification that thickness is measured by a defined method (say, a digital micrometer with 0.001 mm resolution, three readings per septum at defined points, averaged), and that the gauge is calibrated annually against a traceable block, that documented agreement is a defensible chain. What is not defensible is a CoA that reports “9.0 mm ± 0.1” with no method, no instrument, and no calibration reference.
I have a standing rule for incoming septa: a CoA without an instrument reference and a calibration date is a purchasing defect, not a quality defect. You send it back to the supplier before you send it to the auditor. The HPLC septa you buy should arrive with a CoA that names the gauge, the calibration due date, and the as-found readings on the reference block. If the supplier cannot produce that, the traceability chain for every injection on that lot has a hole in it, and no amount of method validation downstream closes it.
The link that breaks: production records that report pass/fail instead of measured values
Here is the failure mode in detail, because it is the one I see most. A molder runs a lot of 500,000 septa. In-process inspection uses a go/no-go gauge: the septum either fits the fixture or it does not. The batch record says “in-process dimensional check: pass.” The CoA says “dimensions conform.” Everyone is happy until an auditor asks a specific question: what was the actual thickness of septum number 250,000, and what instrument measured it? The go/no-go gauge cannot answer that. It has no reading, and its calibration (if it has one) certifies a pass/fail threshold, not a value.
This is a traceability break even though every septum in the lot may be within specification. The problem is not quality; it is the inability to demonstrate a measurement result. Clause 6.5 needs a number with a chain, and a pass/fail stamp is not a number.
The fix is not exotic. Require, in your purchase specification, that the supplier record a measured value (thickness, penetration force, durometer) on a sampling basis, with the instrument ID and the calibration due date on the batch record. A sample of 10 septa per 100,000, measured and recorded, is enough to give the auditor a number to trace. It costs the supplier a few minutes per lot. It costs you nothing to specify. I have seen this single clause turn a nonconformity into a closed item in one audit cycle.
Field case: the durometer reading that had no reference block
A pharma QC lab was qualifying a new supplier of PTFE-faced silicone septa for a headspace method. The CoA reported durometer 50 ± 5 Shore A. During supplier qualification, the auditor asked for the calibration of the durometer used. The supplier produced a calibration certificate from a third-party lab, dated two years prior, with no as-found readings and no reference block identification. The chain stopped there.
The measurement that identified the problem was not the durometer reading itself. It was the calibration interval: the certificate was 26 months old against a stated 12-month interval, and the reference block named on it was not the block in the supplier’s lab. Two breaks, one document. The fix was a re-qualification: the supplier sent three septum samples to an accredited lab for durometer measurement against a certified block, and the resulting report, with the block ID and the uncertainty, became the traceable reference for that material grade. Subsequent lots were compared to that reference on a 12-month recalibration cycle. The septa did not change. The chain did.
Extractables traceability is the link nobody budgets for
Dimensional and physical properties are the easy half. The half that sinks audits is extractables. Your method blank, your detection limit, and your system suitability all depend on what the septum does not release into the mobile phase or the headspace. If an auditor asks how you know the extractables profile of lot 4471 matches the lot you validated on, the answer has to be a traceable measurement, not a supplier assurance.
Traceable extractables data means: a defined extraction protocol (solvent, temperature, time, septum area), an analytical method with its own calibration traceable to certified standards, and a result reported with uncertainty. The reference standards here are things like certified solutions of the target extractables (plasticizers, antioxidants, curing agents) at known concentrations. The instrument is your HPLC or GC-MS, calibrated against those standards. The production record is the lot number. The chain is real, but it is expensive, and most suppliers do not run it per lot. They run it once at qualification and then rely on process control.
That is acceptable if the process control is documented and the qualification data is traceable. It is not acceptable if the qualification report is a summary table with no method and no standards. When I audit a supplier’s extractables package, I look for three things: the certified standard lot numbers, the calibration curve with R² ≥ 0.999 and the standard concentrations, and the extraction conditions in enough detail to reproduce. If any of those is missing, the extractables claim is an assertion, not a measurement. The analytical testing lab that runs the extraction should be able to hand you the raw data behind the summary.
A decision checklist for incoming traceability review
This is the checklist I use at receiving, before a lot goes into the validated workflow. It is deliberately short. If a lot fails any line, it goes to quarantine and the supplier gets a corrective action request, not a concession.
| Check | What to look for | Concrete threshold |
|---|---|---|
| CoA instrument reference | Gauge or instrument ID named on the CoA | ID present, calibration due date within 12 months |
| Measured values, not pass/fail | Thickness, penetration force, durometer reported as numbers | Sample ≥ 10 per 100,000, values recorded |
| Reference standard named | Gauge block or durometer block with certificate | Certificate ID and expiry on file |
| Lot-to-certificate match | Lot number on CoA matches lot on carton and on receiving record | Exact match, no suffix or rework codes unexplained |
| Extractables qualification | Method, standards, extraction conditions | Certified standard lot numbers, R² ≥ 0.999 |
| Temperature exposure record | Autoclave or cure cycle record for the lot | Cycle record with date and temperature within stated range |
| Uncertainty stated | Measurement uncertainty on dimensional values | Stated, and smaller than the specification window |
The uncertainty line is the one people skip. A thickness spec of 9.0 ± 0.1 mm with a measurement uncertainty of ± 0.08 mm is barely a specification at all; the uncertainty eats most of the window. If the supplier cannot state the uncertainty, the number on the CoA is not a measurement result in the 17025 sense. It is a claim. Ask for the uncertainty budget. A competent supplier has one.
How to close the chain when the supplier cannot
Some suppliers, especially smaller molders, cannot produce a traceable chain and will not build one for your volume. You have three options, and I have used all three.
Option one: qualify the material, not the lot. Send a representative sample to an accredited lab for dimensional and extractables measurement, get a traceable report, and treat that report as the reference for the material grade. Then require the supplier to demonstrate process control (SPC on thickness, say) rather than per-lot traceability. This is defensible if your method is not sensitive to lot-to-lot variation and you can show the process is in control. It is not defensible for a method with a tight detection limit.
Option two: bring the measurement in-house. Buy a calibrated micrometer (0.001 mm resolution, traceable block, annual calibration) and a durometer with a certified reference block, and measure incoming lots yourself. Your measurements are traceable because you control the chain. This is the route I recommend for any lab running more than about 50,000 injections a year, because the cost of the instruments is trivial against the cost of an audit finding. It also gives you a number to put in your own records, which is what the auditor actually wants to see.
Option three: switch suppliers. If the traceability gap is structural, and the method is critical, the cheapest long-run fix is a supplier who already runs a 17025-aligned quality system. The vial closures and septum supply chain has enough qualified vendors that you do not have to carry a supplier who cannot document a measurement.
Where the septum sits relative to the rest of the traceable system
One thing worth saying plainly: the septum’s traceability chain does not stand alone. It plugs into the vial, the closure, and the autosampler. If your HPLC vial has a crimp dimension that varies, the septum’s compression and the seal it forms vary with it, and the traceable thickness on the CoA no longer predicts the seal you get. The chain has to include the closure torque and the crimp geometry, or the septum number is traceable to a measurement that does not describe your system. I have seen a lab chase a septum thickness nonconformity for two weeks before finding the real cause: a crimper that had drifted to 0.3 mm out of adjustment, changing the compression on every vial. The septum was fine. The chain around it was not.
This is why the metrological traceability question is never just about the septum. It is about the measurement system that includes the septum. The auditor who asks for the septum chain is usually working toward the vial and the closure, and the lab that has all three documented closes the finding in one pass. The Agilent consumables documentation and the Thermo Fisher sample handling resources both publish dimensional and material specifications you can use as reference values when you build your own incoming specification, and the Restek technical library has application notes that show which septum properties actually affect a given method. Use them to set your thresholds, not to replace your own measurements.
Field case: the extractables standard that had expired
A contract lab running EPA SW-846 methods had a clean audit history until a client audit pulled the extractables qualification for a septum lot. The qualification report was complete: method, extraction conditions, calibration curve, results. The problem was the calibration standards. The certified standard solutions used to build the curve had expired four months before the qualification run. The curve was good (R² 0.9994), but the standards behind it were out of date, so the traceability link to the certified reference material was broken. The whole extractables claim for that lot was unverifiable.
The measurement that caught it was a date check on the standard lot numbers against the certificate expiry. Simple, and easy to miss because the report looked complete. The fix was a re-run of the qualification with in-date standards, plus a standing rule: every extractables report carries the standard lot numbers and expiry dates in the header, and the reviewer checks them before sign-off. That single field has caught three more near-misses since.
The decision rule
If your method is critical and your auditor is serious, the traceability chain for a septum lot must include a measured value, a named instrument, a calibration reference, and a production record that ties all three to the lot. If any of those four is missing, the lot is not traceable in the 17025 sense, regardless of how good the septa are. The cheapest reliable fix is to specify measured values and instrument references in your purchase agreement, and to bring the dimensional measurement in-house with a calibrated micrometer and a certified reference block if your injection volume justifies it. Do that, and the link that breaks becomes the link you own.
References & further reading
Related resources from our lab consumables network:
- Hplc Septa at hplcsepta.com
- Vial Closures at vialsclosures.com
- Analytical Testing at testinganalytical.com
- Hplc Vial at vialforhplc.com
External references and standards: