An AQL number without a defect class is decoration

A 10,000-piece lot of 2 mL clear vials arrived with a COA showing every dimension in the middle of tolerance. The first 200 injections on the lab chromatography bench ran clean. By injection 600 the autosampler started throwing “no sample” errors at a rate of 1 in 40, and a pulled vial showed a crimp that sat 0.4 mm high, enough that the needle guide glanced off the septum edge. The supplier’s outgoing record said “AQL 1.0, accept.” It did not say what sample size, what defect class, or which characteristic was actually measured. That gap between a stated AQL and a real outgoing QC plan is the whole problem, and it is what this article is about: the vial supplier AQL sampling plan and outgoing QC that should sit between the molding line and your dock.

Every vial supplier quotes an AQL. Almost none of them quote the sampling standard, the inspection level, the lot size basis, and the defect classification together. Those four things are the sampling plan. Pull any one out and the number means nothing.

The reference most buyers should insist on is ISO 2859-1 (the international twin of ANSI/ASQ Z1.4). It defines three inspection levels (I, II, III) and a set of AQL values from 0.010 to 1000 defects per hundred units. For a vial lot, the default is General Inspection Level II, normal severity, single sampling. That gives you a sample size code letter based on lot size, and from that a sample size n and an accept/reject number Ac/Re for each AQL.

Here is the table I hand to suppliers when we set up a qualification. These are the numbers I actually use for 2 mL and 4 mL crimp and screw vials, drawn from ISO 2859-1 Level II.

Read the critical column carefully. At AQL 0.065 with n = 80, the accept number is zero. One critical defect rejects the lot. That is the correct posture for anything that can cause a needle strike, a leak, or a false result. If your supplier’s plan has an accept number above zero on a critical characteristic, they are not sampling for critical defects. They are sampling for something else and calling it critical.

Three defect classes, and what belongs in each for a vial

The classification is where most supplier plans fall apart, because the supplier defines “major” as “looks wrong” and “minor” as “cosmetic.” That is a buyer’s definition, not a metrologist’s. Here is the split I use, and it is defensible against an audit.

Critical defects (AQL 0.065, Ac = 0)

• Any crack, chip, or star fracture in the neck or body, visible at 3x magnification under 1000 lux.
• Septum puncture-through or a slit longer than 0.5 mm in the PTFE facing before first use.
• Wrong septum durometer or wrong polymer (a silicone where a butyl was specified, or a PTFE facing under 0.10 mm).
• Cap that will not hold 1.0 in-lb of crimp torque without the seal failing a 30-second vacuum leak check.
• Vial with residual mold release or visible particulate in the interior after the supplier’s wash.

Critical is the class that maps to a patient result or a regulatory finding. A slit septum that lets a needle punch through on the first injection is a critical defect even though the vial looks fine on the bench.

Major defects (AQL 0.65)

• Neck OD outside the drawing tolerance. For a 11 mm crimp neck the typical call is 11.00 ± 0.20 mm; a 0.3 mm excursion is major because it changes cap seating.
• Crimp height above or below the gauge window. For an 11 mm aluminum seal, the finished crimp height is usually specified at 5.5 ± 0.3 mm; outside that the autosampler needle guide can graze the seal.
• Bottom thickness or concentricity out of tolerance. A 0.2 mm bottom runout on a 32 mm base is enough to tilt the vial in the tray and shift the injection depth.
• Thread defect on a screw vial that causes cross-threading on the first turn.
• Visible deformation of the septum after cap application (a “dimple” or “smile” that indicates over-crimp).

Minor defects (AQL 1.5 or 2.5, depending on the customer)

• Cosmetic scuffs, flow lines, or color variation that does not affect function.
• Label misalignment or ink smudge on the printed area.
• Bag seal or packaging print defect that does not compromise sterility or desiccation.

I set minor at AQL 1.5 for most catalog vials and tighten to 2.5 only for bulk industrial lots where the end user is not running a regulated method. The tighter number costs the supplier money in sorting, and it is worth it.

What the sample actually has to be measured against

An outgoing inspection record is only as good as the drawing it references. Before any AQL plan means anything, the supplier needs a controlled drawing with every characteristic that matters and a tolerance for each. For a 2 mL clear glass crimp vial, the drawing I expect to see carries at least these:

• Body OD: 12.00 ± 0.20 mm (the 12 x 32 mm footprint that fits the standard tray).
• Height: 32.0 ± 0.5 mm.
• Neck OD at the crimp ring: 11.00 ± 0.20 mm.
• Neck ID: 6.0 ± 0.2 mm.
• Bottom thickness: 1.0 ± 0.1 mm.
• Bore concentricity: ≤ 0.15 mm TIR.
• Glass: USP Type I borosilicate, hydrolytic resistance per USP <660>.

If the drawing is loose or missing, the AQL plan is measuring noise. I have rejected more supplier plans for a missing drawing than for a wrong AQL. The vial supplier should be able to hand you the drawing, the inspection method for each dimension, and the gauge or instrument used, in the same packet.

Attribute versus variable sampling, and why vials need both

ISO 2859-1 is an attribute plan. It counts go/no-go defects. That is fine for cracks and cosmetic flaws. It is a poor tool for a dimension that drifts, because attribute sampling only tells you whether a unit is inside or outside tolerance, not how far the population is from the edge.

For the dimensional characteristics (neck OD, height, bottom thickness, concentricity), the supplier should run a variables plan in parallel. ISO 3951-1 is the companion standard. In practice I ask for a sample of n = 10 to 20 per lot measured with a calibrated digital micrometer or an optical comparator, and a report of mean and standard deviation against the tolerance. If the mean is within 1.5 sigma of a tolerance limit, the lot gets flagged even if every unit passes the attribute check. That flag is the early-warning system. Attribute data alone will not catch a slow drift in neck OD until it is already producing rejects.

The same logic applies to septum durometer. A Shore A reading of 45 ± 5 is the common spec for a butyl septum; a lot that reads 50 on the high side will still pass a go/no-go durometer check but will punch harder and generate more coring. Record the actual number, not “pass.”

What a realistic outgoing inspection record contains

Most “COAs” I see are one page of boilerplate with a signature. A real outgoing inspection record for a vial lot has six sections, and I will not release a lot without all six. This is the part that separates a supplier with a working outgoing QC program from one that prints certificates.

1. Lot identification and traceability

Lot number, manufacturing date, glass batch number, septum batch number, cap batch number, and the line or cavity the lot came from. If the supplier cannot trace a septum batch to a lot, they cannot do a meaningful recall or a root-cause investigation. I want the cavity number because a single-cavity mold defect is a real failure mode, and a lot-level AQL will not localize it.

2. Sampling plan statement

The standard (ISO 2859-1), inspection level (II), severity (normal), sample size, and the AQL and Ac/Re for each defect class. It should name the characteristics inspected and the method used for each. “Visual” is not a method. “Visual at 3x under 1000 lux against a black and white reference card” is.

3. Measured data, not just pass/fail

A table of the actual readings for the dimensional sample: neck OD, height, bottom thickness, concentricity, septum thickness, durometer. Mean, min, max, and standard deviation. This is where a drifting process shows up before it produces a reject.

4. Functional test results

At minimum: a puncture test on a defined number of septa with a defined needle gauge, recording the puncture force and the number of injections before coring or leakage. A typical spec for a 20 mm butyl septum is ≥ 5 punctures with a 22-gauge needle at a defined insertion depth without a leak. Also a crimp pull-off or vacuum leak test on a sample of finished vials: apply the specified crimp, then pull the cap or apply 30 seconds of vacuum and check for loss of seal. A torque spec of 1.0 to 1.5 in-lb on the crimper is the setting; the record should show the resulting seal integrity, not just the tool setting.

5. Cleanliness and particulate

For vials used in trace analysis, a rinse or extraction test with a result in the record. The common method is a defined rinse volume of HPLC-grade water or methanol, filtered through a HPLC filter, and injected. The record should state the particulate count or the absorbance at 254 nm against a blank. A supplier that cannot produce this number is not doing outgoing QC for a trace-analysis customer.

6. Disposition and signature

Accept, reject, or accept-with-deviation, the name of the inspector, the date, and the reviewer. A deviation must name the characteristic and the reason. “Accepted per customer concession” with no detail is a red flag, not a record.

One more item that belongs in the record and almost never is: the retention sample. The supplier should hold a sealed sample of the lot, typically 20 to 50 units, for the shelf life of the product or at least 24 months. When a customer reports a problem six months later, the retention sample is the only way to run a side-by-side. Ask for the retention policy in writing.

Field case: the lot that passed AQL and failed the bench

A contract lab running EPA Method 8260 on a SW-846 workflow started seeing elevated and variable method blanks on a new lot of 40 mL VOA vials. The supplier’s outgoing record showed AQL 1.0, accept, with a visual inspection and a dimensional check. The blanks were the symptom; the measurement that identified the cause was a rinse test the lab ran themselves: 10 mL of reagent water per vial, 24 hours at 40 °C, then injection on a GC-MS. The blank showed a siloxane pattern at 2 to 4 ppb per vial, consistent with residual septum lubricant or mold release on the interior.

The fix was not a tighter AQL on dimensions. It was adding an interior cleanliness characteristic to the drawing and a rinse test to the outgoing record, with an accept limit of < 1 ppb siloxane response against a blank. The supplier had been sampling for the wrong defect class. Their AQL was fine; their characteristic list was incomplete. The lesson: an AQL plan is only as good as the characteristics it covers, and the characteristics have to come from the customer’s method, not the supplier’s catalog.

Field case: the crimp height that drifted inside tolerance

A different supplier shipped a lot of 11 mm aluminum crimp seals where the finished crimp height, measured on the customer’s bench, ran 5.2 mm against a spec of 5.5 ± 0.3 mm. Every unit was inside tolerance, so the attribute check passed. But the autosampler on the customer’s Agilent platform started reporting intermittent “vial not found” errors at roughly 1 in 60 injections, because the low crimp let the seal sit proud and the needle guide caught the edge.

The measurement that identified it was a variables sample: 20 crimps measured with a digital crimper gauge, mean 5.21 mm, standard deviation 0.08 mm. The mean was 3.6 sigma below the lower limit on the population, even though every sampled unit was inside the window. The supplier’s attribute plan could not see it. The fix was to move crimp height to a variables characteristic with a Cpk ≥ 1.33 requirement and a monthly gauge R&R, and to add a first-article crimp-height check to every lot. The AQL did not change. The inspection method did.

A working checklist for auditing a supplier’s outgoing plan

When I qualify a vial supplier, I ask for the outgoing QC plan as a document and walk it item by item. The checklist below is the one I use. Anything marked “no” is a finding.

• Standard named: ISO 2859-1 or ANSI/ASQ Z1.4, with inspection level and severity stated. (Yes / No)
• Lot size basis defined: the plan states the lot size ranges and the sample size code letter for each. (Yes / No)
• Defect classes defined: critical, major, minor, each with a written description and an AQL. (Yes / No)
• Critical Ac = 0: no accept number above zero on any critical characteristic. (Yes / No)
• Characteristics list: every dimension on the controlled drawing is inspected, with the method named. (Yes / No)
• Variables plan: dimensional characteristics have a measured sample with mean and SD, not just pass/fail. (Yes / No)
• Functional tests: puncture count, crimp seal integrity, and torque are tested and recorded. (Yes / No)
• Cleanliness test: rinse or extraction result is in the record for trace-analysis grades. (Yes / No)
• Retention sample: sealed units held for 24 months minimum, with a stated policy. (Yes / No)
• Signature and disposition: inspector and reviewer named, deviations documented. (Yes / No)
• Traceability: glass, septum, and cap batches traceable to the lot and cavity. (Yes / No)

If a supplier cannot produce this document on request, they do not have an outgoing QC plan. They have a certificate template. The distinction matters at the dock, not in the sales meeting.

The sampling plan is a contract term, not a brochure claim

Two practical notes on making this stick. First, put the sampling plan in the purchase order or the quality agreement, with the standard, the inspection level, the AQLs, and the characteristic list as an appendix. A plan that lives only in the supplier’s quality manual is a plan you cannot enforce. Second, ask for the outgoing record with every shipment, not on request. A supplier that batches records or sends them quarterly is telling you the record is generated for the audit, not for the lot.

On the instrument side, the same discipline applies to the consumables around the vial. A septum that passes puncture count but sheds a particle will show up on the column, and the fix often involves a PTFE membrane filter on the sample prep line to catch what the vial contributed. The Thermo Fisher and Restek technical libraries both carry application notes on septa and vial cleanliness that are worth reading against your own method before you write the characteristic list. The supplier’s AQL is not the method’s requirement; your method is.

The decision rule I use, and the one I recommend you adopt: accept no vial lot whose outgoing record does not name the sampling standard, the defect classes with their AQLs, the measured data for dimensional characteristics, and at least one functional test result, with a zero accept number on critical defects. If any of those four is missing, the lot goes on hold and the supplier gets a corrective action request, regardless of what the COA says. The AQL is not the guarantee. The record behind it is.

References & further reading

Related resources from our lab consumables network:

• Vial Supplier at vialsupplier.com
• Lab Chromatography at labchromatography.com
• Hplc Filter at hplcfilters.com
• Ptfe Membrane Filter at ptfemembranefilter.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography