A lab I audit twice a year failed its last ISO 17025 surveillance on a COD vial lot that had passed every in-house check we ran. The assessor asked one question: show me the calibration certificate for the balance that weighed the potassium hydrogen phthalate standard that set the COD value on the certificate of analysis. The lab had the balance log. It had the CRM certificate from the vial supplier. What it did not have was any document connecting the two. The traceability chain for ISO 17025 compliance on a cod vial broke at the point where a purchased standard entered a purchased consumable, and neither vendor owned that junction. That is the gap this article is about.
Most labs treat ISO 17025 traceability for a cod vial as a paperwork exercise: keep the CoA on file, keep the balance calibration certificate on file, done. The assessor will not accept that. Traceability is a chain of unbroken comparisons, each with a stated uncertainty, ending at a national metrology institute. For a sealed glass vial filled with a dried reagent, the chain has five links. Four of them are easy. The fifth is the one that sinks audits, and it is the one almost nobody thinks to demand from the supplier before the purchase order goes out.
ISO/IEC 17025:2017 clause 6.5 requires that measurement results be metrologically traceable to the SI through an unbroken chain of calibrations, each contributing to the measurement uncertainty. For a COD determination by closed reflux colorimetry or titration, the measurand is the oxygen demand of the sample. The vial does not measure that. The vial contains a reagent that reacts with it. So the traceability chain runs:
- Link 1: the mass of potassium hydrogen phthalate (KHP) weighed to prepare the calibration standard, traceable through a calibrated balance to the national mass standard.
- Link 2: the purity and water content of that KHP, traceable through a certified reference material certificate to a metrology institute.
- Link 3: the volume of digest solution dispensed into each vial, traceable through a calibrated dispenser or through the gravimetric fill weight recorded at manufacture.
- Link 4: the absorbance or titration endpoint measured on your instrument, traceable through certified absorbance or titrant standards.
- Link 5: the assignment of a COD value to the manufactured lot, which depends on links 1 through 3 being performed by the supplier and documented in a way that lets your lab adopt that value as its own.
Links 1, 2, and 4 are yours. You control them, you audit them, you can fix them. Link 3 is shared. Link 5 is the one that breaks. When a supplier ships a vial with a printed CoA showing “COD 1500 mg/L ± 5%”, that number is a supplier measurement. To use it as a traceable value in your own quality system, you need the supplier’s chain, not just the supplier’s conclusion.
What “unbroken” actually requires on a consumable CoA
A CoA is not a traceability document. It is a summary of results. The traceability document is the calibration certificate behind each result, and for a manufactured consumable you need three specific pieces of evidence to close the chain:
The KHP standard’s own provenance
The supplier’s KHP must come with a certificate that names the reference material, its certified purity, its expanded uncertainty, and the metrology institute it traces to (NIST SRM 84l is the common one, or a NIST-traceable secondary standard). If the CoA says “potassium hydrogen phthalate, ACS grade” and stops there, the chain is broken at link 2. ACS grade is a reagent purity specification, not a traceability statement. The difference matters: ACS grade KHP is typically 99.95% to 100.05% assay with no stated uncertainty on the water content, while a certified reference material carries an expanded uncertainty of around 0.02% to 0.05% and an explicit drying protocol.
The gravimetric fill record
For a sealed vial, the volume of reagent is the hardest quantity to verify after the fact. The supplier should record the net fill weight per vial (or per fill head, per lot) and the density of the solution at the fill temperature. With those two numbers, volume is calculable to ±0.1% on a 10 mL fill. Without them, the volume is nominal, and a nominal volume cannot enter an uncertainty budget. Ask for the fill weight distribution, not the mean. A lot with a mean fill of 10.002 g and a standard deviation of 0.015 g is a different product from one with the same mean and a standard deviation of 0.004 g, even though both datasheets read “10 mL nominal.”
The assignment uncertainty
The assigned COD value needs an expanded uncertainty (k=2) that combines the KHP mass uncertainty, the KHP purity uncertainty, the fill volume uncertainty, and the analytical measurement uncertainty on the supplier’s instrument. A typical combined expanded uncertainty for a well-run COD vial lot is ±1.5% to ±3% relative. If the CoA quotes ±5% but the supplier cannot decompose that figure into components, the number is a tolerance, not an uncertainty, and your assessor will treat it as such.
| Traceability element | What a weak CoA shows | What an auditable CoA shows | Why the assessor cares |
|---|---|---|---|
| KHP standard | “ACS grade” | SRM number, certified purity, expanded uncertainty, drying protocol | Links the standard to a metrology institute |
| Balance used | Not stated | Calibration certificate ID, calibration date, uncertainty at the weighed mass | Links the mass to the national mass standard |
| Fill volume | “10 mL nominal” | Net fill weight mean and SD, density at fill temperature, calculated volume with uncertainty | Volume is a measured quantity, not a design value |
| Assigned COD | Single value with ±5% | Value, expanded uncertainty (k=2), component breakdown | Uncertainty must be defensible, not asserted |
| Stability | Expiry date only | Real-time stability data, storage condition envelope, re-test interval | The assigned value must hold across the shelf life |
The link most labs cannot produce: the supplier’s fill volume record
In the audits I have sat through, the failure is almost always the same. The lab has the CoA. The lab has the supplier’s ISO 17025 or ISO 9001 certificate. The lab has its own balance and spectrophotometer calibration records. What it does not have is the supplier’s gravimetric fill data, and without that, the volume in every vial is an assumption. I have seen assessors push on this exact point three times in five years, and twice the lab had to re-establish the assigned value by running its own KHP curve against the vials, which is a week of work and a permanent note in the corrective action file.
The fix is procedural, not technical. Before the next purchase order, ask the supplier for a lot-specific fill record. A serious supplier will send a one-page table: fill head number, mean net fill weight, standard deviation, number of vials weighed, density at fill temperature, calculated volume, and combined uncertainty. If the supplier cannot produce this, you have two options. Either qualify the vials against your own KHP curve on every lot (which converts the vial from a traceable standard to a sample you measure), or change suppliers. There is no third option that survives an assessment.
Case 1: the lot that drifted 4.2% and the balance certificate that explained it
A water utility lab running closed reflux COD on a cod vial platform saw its QC check standard read 4.2% low against the assigned value on three consecutive runs. The vials were within expiry. The spectrophotometer had been calibrated the week before. The lab blamed the instrument and sent it out for service. The instrument came back clean.
The actual cause was in the supplier’s CoA. The assigned value had been set using a balance whose calibration certificate was 14 months old against a 12-month calibration interval. The balance had drifted +0.3% over that period, which propagated directly into the KHP mass and therefore into the assigned COD. The supplier’s internal audit caught it six weeks after the lot shipped. The lab’s fix was to add a CoA review step to incoming inspection: check the calibration date on every balance cited in the CoA against the calibration interval stated on its certificate, and reject any lot where the certificate is out of interval. That single check would have caught this lot before it entered the QC program.
Case 2: the 0.9% fill weight spread that failed a proficiency test
A contract lab failed an external proficiency test on COD by 0.9%, just outside the acceptable z-score band. The vials were from a supplier with a good reputation. The lab’s own KHP curve was solid. The problem was the fill weight standard deviation on that lot: 0.021 g on a nominal 10 g fill, which is a 0.21% relative standard deviation on volume. Combined with the KHP purity uncertainty and the instrument uncertainty, the total expanded uncertainty on the assigned value exceeded the proficiency test’s acceptance window. The lab had been treating the CoA’s ±3% as a fixed tolerance rather than an uncertainty that needed to be propagated into its own budget.
The fix was twofold. First, the lab added the supplier’s fill weight SD to its own uncertainty budget, which flagged the lot as marginal before the proficiency test ran. Second, it moved to a supplier who could hold fill weight SD below 0.010 g on a 10 g fill. That is a 0.1% relative SD, and it changed the total uncertainty enough to pass the next round comfortably. The lesson: a CoA’s stated tolerance is not the number you put in your uncertainty budget. The component uncertainties are, and fill weight is usually the largest one you do not control.
What to demand from the supplier before the PO, in writing
Here is the checklist I use. Every item is a document, not a verbal assurance, and every item is lot-specific. A supplier who can produce all seven is a supplier you can defend to an assessor. A supplier who can produce four is a supplier you will spend a week re-qualifying.
- KHP certificate naming the reference material, certified purity, expanded uncertainty, and the metrology institute it traces to. NIST SRM 84l or an equivalent is the benchmark.
- Balance calibration certificate for every balance cited in the CoA, with the calibration date, the calibration interval, and the uncertainty at the weighed mass. Check the date against the interval.
- Gravimetric fill record per lot: fill head, mean net fill weight, standard deviation, number of vials weighed, density at fill temperature, calculated volume, combined uncertainty.
- Assigned value with component uncertainty breakdown, expanded at k=2. If the supplier cannot decompose the figure, treat the CoA as a specification, not a traceable value.
- Stability data supporting the expiry date under the stated storage envelope. Real-time data on the same matrix and fill volume, not accelerated data on a different product.
- Scope of accreditation if the supplier claims ISO 17025 for the assignment. Check that the scope covers COD vials specifically, not just “reagents” in general. Scopes are narrower than certificates suggest.
- Change control notification for any change to the KHP source, the fill equipment, or the assignment method. A silent change to the KHP supplier breaks the chain retroactively for every lot in inventory.
For readers building out the rest of the consumable qualification file, the same logic applies to the vial insert and the 2ml vial in their HPLC workflows, and the broader HPLC consumables qualification follows the same five-link structure. The instrument vendors publish useful guidance on consumable qualification and uncertainty propagation; Agilent’s chromatography consumables documentation and Thermo Fisher’s sample handling resources are worth reading alongside the method text. For the regulatory frame, the EPA SW-846 test methods define the analytical requirement, and Restek’s technical resources cover the practical side of standard preparation and stability.
How to propagate the supplier’s uncertainty into your own budget
The assessor will ask for your uncertainty budget for the COD result, and the vial’s assigned value uncertainty is a term in it. The propagation is straightforward if you have the component breakdown and impossible if you do not. For a typical closed reflux COD determination using a pre-dosed vial, the budget looks like this:
| Component | Typical relative uncertainty (k=1) | Source |
|---|---|---|
| Assigned value of the vial | 0.75% to 1.5% | Supplier CoA, expanded at k=2 then divided by 2 |
| Fill volume variation | 0.1% to 0.3% | Supplier gravimetric record, SD divided by mean |
| Sample volume (pipette) | 0.3% to 0.8% | Your calibrated pipette, at the dispensed volume |
| Spectrophotometer absorbance | 0.2% to 0.5% | Your calibration certificate, at the working absorbance |
| Digestion temperature and time | 0.5% to 1.2% | Your block verification, at 150 °C ± 2 °C for 2 h |
| Repeatability | 0.5% to 1.0% | Your replicate runs, pooled SD |
Combine in quadrature. A well-controlled lab lands at a combined relative uncertainty of 1.5% to 2.5% (k=1), which gives an expanded uncertainty of 3% to 5% at k=2. If your budget comes out above 5%, the largest term is almost always the assigned value or the fill volume, and both of those trace back to the supplier’s documentation. That is the argument to take to procurement: the vial is not a commodity line item, it is a term in your uncertainty budget, and the term is only as good as the paperwork behind it.
Decision rule
Do not accept a COD vial lot into a 17025-accredited workflow unless the supplier can produce, for that specific lot, the KHP reference certificate, the balance calibration certificate with its interval, and the gravimetric fill record with mean, standard deviation, and calculated volume. If any of the three is missing, the vial is a sample you must measure, not a standard you can adopt, and your uncertainty budget has to say so. The chain breaks at the fill record more often than anywhere else, so ask for it first.
References & further reading
Related resources from our lab consumables network:
- Cod Vial at codvial.com
- Vial Insert at hplcinsert.com
- 2Ml Vial at 2mlsamplevial.com
- Hplc Consumables at hplcconsumable.com
External references and standards: