A lot of 500,000 9 mm screw-thread vials left our dock in March with a certificate that showed a mean crimp diameter of 8.42 mm and a standard deviation of 0.03 mm. Eleven weeks later a customer sent back 24 vials from that lot with septa pushed through the cap and into the vial body. The lot had passed outgoing inspection at AQL 1.0, general inspection level II. That is the problem with a chromatography supplier AQL sampling plan that was copied from a fasteners catalog: the plan told us the lot was acceptable, and the lot was not acceptable for the job it went into. The defect was real, it was concentrated in one molding cavity, and a 200-piece sample drawn at random across five cavities should have caught it. It did not, because the defect class was wrong. We had been counting “septum displaced” as a major defect. It should have been critical.
This is what I have learned about outgoing QC for chromatography consumables, and what I would put in a sampling plan today if I were rebuilding one from scratch. It is narrower than a general quality guide on purpose. Incoming inspection at the lab and in-process capability are separate problems, owned elsewhere. This is about the last gate before a pallet ships.
ANSI/ASQ Z1.4 (the current US designation; ISO 2859-1 is the international twin) is a sampling system built around acceptable quality limits. The AQL is not a target defect rate. It is the worst process average you are willing to accept as the basis for sampling. The subtlety that bites people: a plan at AQL 1.0 with a sample size of 200 accepts the lot at 5 defectives or fewer and rejects at 6. That means a lot running at a true 2% defective rate has a meaningful chance of passing. For a vial that will be pierced once by an autosampler needle and discarded, a 2% failure rate is fine. For a septum that will be pierced 50 times on a GC autosampler, it is not.
The fix is not to tighten the AQL across the board. It is to split the defect classes and give each one its own AQL, because the cost of escape is wildly different. A scuffed label costs a customer a phone call. A septum that cores on the third injection costs a customer a lost sequence and a re-run of 96 samples.
Defect classification for chromatography consumables
I use four classes. The boundary between critical and major is where most suppliers get it wrong, and it is the single most important decision in the plan.
- Critical (AQL 0.065, or zero acceptance): any defect that can cause a false result, a lost run, or a safety event. Septum core or plug visible after 10 punctures with a 23-gauge needle. Cap that will not hold torque. Glass fragment inside the vial cavity. Wrong septum material in a lot (silicone where PTFE-faced was ordered). Cross-threaded closure that leaks at 1.5 psi.
- Major (AQL 0.65): dimensional or functional defects that degrade performance without stopping the run. Crimp diameter out of tolerance by more than 0.10 mm. Septum durometer outside 45 to 55 Shore A. Vial height outside ±0.25 mm. Cap liner misaligned but sealing.
- Minor (AQL 1.5): cosmetic and handling defects. Label smudged, print misaligned, bag seal wrinkled, minor mold flash on the base that does not affect seating.
- Incidental (AQL 4.0): appearance-only defects with no functional consequence. I usually drop this class entirely; it consumes inspector time for no return.
Note the zero-acceptance call on critical. For a critical defect, the plan is c=0, not AQL 0.065. You inspect a fixed sample (I use 125 pieces for lots above 10,000) and reject on the first critical. That is not statistically elegant, but it is operationally honest: a critical defect found in sampling almost always means the process is out of control, and the right move is to quarantine the lot and investigate, not to accept it because the count stayed under a threshold.
Sample size follows the defect class, not a single table lookup
Most suppliers run one sample size per lot and apply all defect classes to it. That wastes inspection on minor defects and under-samples critical ones. I run a tiered plan: a large sample for the cheap, fast checks and a smaller, slower sample for the functional checks that take real time.
| Defect class | AQL | Sample size (lot 10,001 to 35,000) | Accept / reject | Test |
|---|---|---|---|---|
| Critical | 0 (c=0) | 125 | 0 / 1 | Puncture test, leak test, material ID |
| Major (dimensional) | 0.65 | 315 | 3 / 4 | Crimp diameter, height, thread gauge |
| Major (functional) | 0.65 | 125 | 1 / 2 | Torque, seal integrity at 1.5 psi |
| Minor | 1.5 | 200 | 7 / 8 | Visual, labeling, packaging |
These numbers come straight from Z1.4 single sampling, normal inspection, general level II for the dimensional and minor rows, and level I for the functional row because each functional test takes 40 seconds. If your process is running well, you can move to reduced inspection after 10 consecutive accepted lots; if you have two rejects in five lots, tighten to level III and doubled sample sizes until you have 10 clean lots. That switching rule is in the standard and almost nobody uses it. Use it. It is free protection.
Why 125 and not 32
A 32-piece sample at AQL 0.65 accepts on zero defectives. That sounds strict until you realize a 32-piece sample cannot distinguish a 1% defective lot from a 4% defective lot with any confidence. The 125-piece sample at the same AQL accepts on 1 and rejects at 2, and it gives you an ~90% probability of detecting a 3.5% defect rate. For a functional test that costs 40 seconds, 125 pieces is 83 minutes of inspector time. That is the cost of not shipping a bad lot. I have never regretted paying it.
The tests that actually find defects, and the ones that only look like they do
An outgoing inspection record full of “visual: pass” lines is theater. The inspections that catch real escapes in chromatography consumables are physical and instrumented, and there are five of them that earn their place in the plan.
Puncture test on septa
Take 10 septa from the sample, mount them in a test fixture, and pierce each with a 23-gauge needle at 20 mm/s to a depth of 8 mm, 50 times, with the needle withdrawn and re-inserted each cycle. Inspect the septum face at 10x. Any visible plug, tear, or material displaced into the vial side is a critical defect. This is the test that would have caught the March lot: the cavity that produced the bad septa had durometer drifting toward 40 Shore A, and 50 punctures at 40 Shore A with a 23-gauge needle will core reliably. A durometer check on the raw sheet would have flagged it, but the molded part is where the problem lives.
Leak test
Seal a sample vial with its closure, pressurize to 1.5 psi (about 10 kPa) with dry nitrogen, and submerge in water for 30 seconds. Any continuous bubble stream is a major defect. For crimp vials, this test also catches insufficient crimp force; for screw vials, it catches cross-threading and liner misalignment. I have seen suppliers skip this because it is messy. It is the only test that catches a whole class of sealing failures that dimensional checks miss.
Crimp diameter and torque
For 11 mm crimp vials, measure the crimp diameter with a calibrated pin gauge or optical comparator at four points around the cap. Target 8.40 mm ± 0.10 mm per the GPI 11 mm neck finish; anything below 8.30 mm is under-crimped and will leak, above 8.50 mm is over-crimped and will crack the neck. For 9 mm screw-thread vials, apply 0.9 to 1.3 N·m with a torque screwdriver and verify the cap does not back off after 10 minutes. Under-torque leaks; over-torque strips the thread or deforms the liner.
Material identification
FTIR on a single septum from each lot, compared against a retained reference spectrum. This catches the substitution defect that no dimensional check will ever find: a supplier swapping silicone for a cheaper EPDM, or a PTFE-faced septum where the facing is actually a thin film of something else. A 30-second scan per lot, and it has saved me twice.
Glass dimensional check
Vial height, body diameter, and neck ID measured on 30 pieces from the sample with calibrated calipers or a vision system. For a 2 mL vial, height 32.0 mm ± 0.25 mm, body OD 11.6 mm ± 0.20 mm, neck ID 6.0 mm ± 0.10 mm. Neck ID is the one that matters most: it sets how the septum seats, and a 0.15 mm drift there is enough to cause intermittent autosampler pickup errors.
Two field cases where the plan caught what the certificate missed
Case 1: the cavity that ran hot
Symptom: a customer running EPA Method 8260 on a 50-position autosampler reported intermittent “no injection” errors on positions 30 through 45 of every tray. The pattern was positional, which pointed at the autosampler, and the customer spent two days chasing it.
Measurement: we pulled 40 vials from the customer’s remaining stock and measured neck ID. Mean 6.02 mm, but the distribution was bimodal: 6.00 mm and 6.14 mm, with the 6.14 mm group clustered by tray position. The 6.14 mm vials were from one molding cavity running 4°C hot. At the wider neck ID, the septum sat 0.12 mm lower, and the autosampler needle was bottoming out before full penetration on certain positions.
Fix: the outgoing plan now includes a neck ID measurement on 30 pieces per lot with a reject at ±0.08 mm from nominal, and a cavity-stratified sample so each cavity is represented. The customer’s autosampler was fine. Our sample was not stratified.
Case 2: the lot that passed on paper
Symptom: a lot of PTFE-faced silicone septa passed outgoing inspection with zero major defects, then generated a customer complaint of ghost peaks at retention times consistent with a phthalate.
Measurement: we ran a blank injection sequence on 20 septa from the lot against our retained reference. The suspect lot showed a 340 ppb phthalate signal at m/z 149. The reference showed 18 ppb. The septa had been molded on a press that had just run a phthalate-plasticized compound for another customer, and the changeover had not included a purge.
Fix: the plan now requires a solvent extractables check by GC-MS on 5 septa per lot, with a reject threshold at 100 ppb total phthalate signal. It costs about 25 minutes of lab time per lot. It caught a second contaminated lot four months later before it shipped. This is one of the places where analytical testing capability inside the supplier’s own lab pays for itself; a supplier that has to send samples out for this check will not run it often enough.
What a real outgoing inspection record contains
The certificate of analysis most suppliers ship is one page and says “conforms.” That is not a record. A record that a procurement manager can actually audit has these elements, and I would ask for every one of them before approving a supplier:
- Lot number, manufacturing date, and cavity or line identifier for each sampled unit. Without cavity traceability, a reject tells you nothing about where to look.
- Sampling plan reference: the standard (ANSI/ASQ Z1.4 or ISO 2859-1), inspection level, AQL per defect class, and sample size actually drawn. “Per our internal standard” is not a reference.
- Measured values, not pass/fail. Crimp diameter as mean and standard deviation across the sample, not “within spec.” Neck ID as min/max. Torque as a range. If the record shows only “pass,” you cannot see drift.
- Test method and instrument ID for each measurement. Calibration due date on the instrument. This is what makes the record auditable under ISO 9001 clause 7.1.5 and, for pharma customers, under 21 CFR 211.160.
- Critical defect result stated explicitly, including the puncture count and needle gauge used. “Puncture test: pass” without the gauge and count is meaningless.
- Inspector ID and date, and the disposition (accept, reject, rework, sort).
- Retained sample reference. A sealed retention sample from each lot, held for at least 24 months, is the only way to investigate a complaint a year later.
If you are buying from an OEM and private-labeling, the record has to travel with the goods under your name. That is a contractual point, not a courtesy, and it belongs in the quality agreement. The OEM vial supply chain is exactly where records get thin, because the brand on the box is not the plant that made the glass.
The decision rule I would hand a new QC manager
Set critical defects at c=0 with a 125-piece sample and a puncture test at 50 cycles with a 23-gauge needle. Set dimensional major defects at AQL 0.65, level II, and stratify the sample by cavity. Run the leak test at 1.5 psi on every functional sample. Require measured values, not pass/fail, on the certificate. And when a critical defect appears anywhere in the sample, stop the lot and find the cavity before you ship anything from it. The AQL tables are a sampling system, not a warranty; the record is what you actually own. If a supplier cannot show you a cavity-stratified sample and a measured crimp diameter distribution, the AQL on their certificate is a number they chose for your comfort, not one their process earned. For the consumables side of the same question, the chromatography supplier page covers what the product line looks like; the sampling plan is what decides whether it ships. Cross-check septum and closure specifications against published reference data from Restek and the instrument makers (Agilent, Thermo Fisher) before you write the tolerances into the plan, and confirm the container requirements in EPA SW-846 for the methods your customers actually run.
References & further reading
Related resources from our lab consumables network:
- Chromatography Supplier at chromatographysupplier.com
- Oem Vial at oemsamplevials.com
- Vial Closures at vialclosures.com
- Analytical Testing at labanalyticaltesting.com
External references and standards: