A 200 µL lot of clear glass vials came back from the autosampler with a 1.8% RSD on the internal standard over 96 injections, against a method spec of 0.5%, and the first instinct was to blame the pump. It was not the pump. The closure torque on that lot measured 1.1 in-lb at the filling station instead of the 1.6 in-lb the method required, because the vendor had switched cap liner thickness by 0.2 mm without a change notification. That is the kind of failure a supplier audit is supposed to catch before the pallet ships. Below is the hplc vial supplier audit checklist I actually use, built around the quality-system items that predict this class of problem, the evidence to demand before the audit date, and a scoring method that produces a number you can defend to procurement.
Most supplier audits I have seen open with the quality manual and end with a walk through the warehouse. By then you have already spent the travel budget and the vendor knows you are there. The checklist has to run in two phases: a desk review of documents you request in advance, and an on-site review of the items the documents cannot prove. If the desk review fails, cancel the site visit. I have done this three times in fifteen years and each time it saved a week of my life.
What you request in advance, in writing, with a deadline of ten working days before the visit:
- Current ISO 9001 or ISO 13485 certificate with the accreditation body named, plus the scope statement. A certificate whose scope says “distribution of laboratory supplies” does not cover molding. If the vendor molds, the scope must say so.
- Control plan for the 12 x 32 mm vial family, showing the characteristic, the sample size, the frequency, the gauge, and the reaction plan for each. If the control plan lists “visual” as the only in-process check on dimensional characteristics, the process is not controlled.
- Three consecutive months of capability data (Cpk) on the four critical dimensions: body OD, body height, neck OD, and septum seat depth. Ask for the raw histogram, not the summary. A Cpk of 1.33 on a 12.00 ± 0.20 mm OD is the floor; anything below 1.0 means the process is running at the tolerance limit and will drift.
- Change notification procedure and the last twelve months of change notices issued to customers. This is the single most predictive document in the pack. A vendor that cannot produce a change log has never issued one, which means they change things silently.
- Resin and glass batch traceability from lot number to shipped carton. For glass, the forming machine and shift. For plastic, the resin lot and the molding press.
- Complaint log for the last 24 months, with root cause and corrective action status. Ask for it unfiltered. A vendor with zero complaints is either perfect or not recording them.
If the vendor pushes back on any of these, that is data. A supplier that treats a capability histogram as proprietary is telling you the process is not capable and they know it.
The three quality-system items that actually predict a bad lot
Over the years I have stopped weighting the quality manual heavily. Every vendor has a nice binder. Three items separate the vendors that ship consistent product from the ones that ship surprises.
1. Incoming inspection on glass and resin, with a gauge R&R
Ask to see the incoming inspection record for the last five glass deliveries. Then ask for the gauge repeatability and reproducibility study on the dimensional gauge they use. If the gauge R&R is above 30% of tolerance, their inspection cannot distinguish a good vial from a marginal one, and the outgoing certificate is decoration. I have seen a vendor measuring neck OD with a caliper rated at ±0.02 mm on a tolerance of ±0.10 mm. The gauge consumed 40% of the tolerance band before the part was even measured.
2. A documented mold or forming-tool maintenance interval
Vial defects cluster at the end of tool life. Neck flash, short shots on the septum seat, and out-of-round bodies all appear in the last 10 to 15% of a mold’s cycle count. Ask for the tool maintenance schedule, the actual cycle count at last service, and the reject rate trend against cycle count. A vendor that runs a mold to failure and sorts the bad parts out at the end of the line is a vendor that will eventually ship the bad parts. The HPLC vial forming line should have a preventive maintenance trigger tied to shot count, not calendar time.
3. A septum lot qualification protocol that includes puncture testing
The septum is where most field failures originate, and it is the component vendors most often buy in rather than make. Ask for the septum supplier’s name, the incoming lot test, and the puncture count at which the septum fails. For a standard silicone/PTFE septa on a 11 mm crimp or a 9 mm screw cap, the qualification should show no coring through at least 100 punctures with a 22-gauge needle at the method’s injection rate. If the vendor cannot state a puncture count, they have not tested it. This is the same failure mode the autosampler vial spec sheet should carry, and most do not.
What to measure on the bench, not read in the binder
The site visit still matters, but not for the reason most checklists assume. You are there to run three measurements yourself, on product pulled from the shipping dock, not from the sample cabinet the vendor prepared. The physical factory walk-through and the contract language are separate work; here I am only concerned with the bench evidence.
Dimensional and closure audit, 30 pieces minimum
Pull 30 vials from a sealed carton in finished goods. Measure body OD, height, neck OD, and the septum seat depth with a calibrated gauge. Then cap ten of them at the torque your method specifies and measure breakaway torque after 24 hours. The number that matters is the torque retention: if a 1.6 in-lb closure reads below 1.2 in-lb after a day at ambient, the cap liner is creeping and the seal will fail in the autosampler tray. That is exactly the failure that produced the 1.8% RSD in the opening case.
Extractables screen on the actual resin lot
Run a blank injection of the vendor’s vial filled with your mobile phase, held at the method’s autosampler tray temperature for the full run time, and compare against a fused-silica reference. Look for ghost peaks above 0.5 mAU at the method wavelength. If the vendor’s own resin lot produces peaks your reference does not, the lot is contaminated or the resin grade is wrong. The vial you approve should carry the resin grade on the certificate, not just the lot number.
Septum coring test at the method’s needle gauge
Puncture ten septa 100 times each with the needle your autosampler uses, then inspect the septum face and the vial interior for silicone fragments under 10x magnification. Count fragments per septum. Anything above zero on a 100-puncture run is a reject. This is the test that separates a qualified septum from a catalog part number.
Scoring the vendor so the number survives procurement review
A checklist that produces a narrative is a checklist that gets argued about. Produce a weighted score. I use five categories, each scored 0 to 5, with the weights below. The weights reflect what actually correlates with field failures in my complaint history, not what looks balanced on a slide.
| Category | Weight | What earns a 5 | What forces a 2 or below |
|---|---|---|---|
| Dimensional capability (Cpk on 4 critical dims) | 25% | Cpk ≥ 1.33 on all four, raw data provided | Cpk < 1.0 on any dimension, or summary only |
| Septum qualification and coring data | 25% | ≥ 100 punctures, zero coring, lot-level test | No puncture count stated, or < 50 punctures |
| Change control and notification | 20% | Written procedure, 12-month log, customer sign-off required | No log, or changes made without notice |
| Traceability (glass/resin lot to carton) | 15% | Full chain, retrievable in under 4 hours | Lot number on carton only, no link to batch |
| Complaint and CAPA history | 15% | Open log, root cause closed within 30 days | Zero complaints claimed, or CAPAs open > 90 days |
Multiply each score by its weight, sum, and you get a number out of 5. My rule: 4.0 and above is approved, 3.0 to 3.9 is conditional with a 90-day corrective action plan, below 3.0 is disqualified regardless of unit price. The headspace vial line gets the same scoring with the septum and closure categories weighted higher, because the crimp seal is the dominant failure mode there.
Two field cases where the audit caught what the certificate missed
Case 1: the drift that looked like a pump problem
Symptom: a 96-injection sequence on a 1260 Infinity II showed the internal standard RSD climbing from 0.4% at injection 20 to 1.8% at injection 96, with no change in retention time. The pump and autosampler were both serviced and cleared.
Measurement: we pulled the tray and measured breakaway torque on the remaining vials. The first 20 read 1.5 to 1.7 in-lb. The vials at positions 60 through 96 read 1.0 to 1.2 in-lb. The vendor had changed cap liner thickness from 1.0 mm to 0.8 mm on that lot without a change notice, and the thinner liner relaxed under the 1.6 in-lb set torque over the run.
Fix: the vendor was placed on conditional status and required to submit a change notice for any liner change going forward. We added a torque retention check (24-hour breakaway at 1.2 in-lb minimum) to incoming inspection. The audit checklist item that would have caught it in advance was the change log, which the vendor had never maintained.
Case 2: the ghost peak that was a resin grade
Symptom: a new lot of 9 mm screw-thread vials produced a reproducible peak at 4.2 minutes, 1.2 mAU, on a 210 nm method that had run clean for two years. The peak appeared only on vials from one carton.
Measurement: we ran a blank extraction on ten vials from the carton against ten from the previous lot. The new lot showed the 4.2-minute peak at 1.0 to 1.4 mAU; the old lot showed nothing above 0.2 mAU. The certificate listed “PP” as the resin, but the vendor’s incoming record showed a regrind content of 15% in that batch. Regrind carries carryover from the previous molding run.
Fix: the vendor’s control plan was amended to cap regrind at 5% and to require a resin grade declaration on every certificate. We added a lot-level extractables screen to incoming inspection for any new carton. The audit item that predicted this was the incoming inspection record on resin, which the vendor had been signing off without checking regrind content.
What to request in the audit report, and what to do with it
The audit is only worth the corrective action plan it produces. Every finding gets a severity, an owner, and a due date, and the plan goes back to the vendor within five working days. I classify findings three ways:
- Critical: a quality-system gap that can ship a nonconforming lot (no change control, no septum qualification, Cpk below 1.0). Corrective action before the next purchase order, no exceptions.
- Major: a gap that degrades confidence but has a detection control (gauge R&R above 30%, incomplete traceability). Corrective action within 90 days with a re-audit.
- Minor: documentation and housekeeping items. Corrective action within 180 days, verified by document review, no site visit.
The re-audit is a desk review of the corrective action evidence, not a second site visit, unless a critical finding is involved. That keeps the cost of the audit program proportional to the risk. For vendors supplying the Agilent or Thermo Fisher autosampler platforms, the dimensional spec is fixed by the instrument, so the audit can focus almost entirely on septum and closure behavior. For methods governed by U.S. EPA SW-846, the container is part of the method, so the extractables screen and the traceability chain carry more weight than the dimensional capability. The Restek technical library has good reference data on septum bleed and needle gauge effects that I use to set the puncture-count threshold for a given method.
The decision rule
Approve a vial vendor only when the desk review returns capability data on all four critical dimensions with Cpk at or above 1.33, a septum qualification showing zero coring through 100 punctures at the method’s needle gauge, a written change notification procedure with a retrievable 12-month log, and a traceability chain that resolves a carton to a forming batch in under four hours. Score the vendor against the weighted table, and treat anything below 4.0 as conditional and anything below 3.0 as disqualified no matter what the unit price says. The unit price is a line item; the audit score is the number that tells you whether the next 96-injection sequence runs at 0.4% RSD or 1.8%.
References & further reading
Related resources from our lab consumables network:
- Hplc Vial at hplcgc.com
- Autosampler Vial at autosamplerhplc.com
- Vial at vials-filters.com
- Headspace Vial at headspacevialsupplier.com
External references and standards:
