[Arkcon]

We've been having trouble, my managers and I, with a method being validated by an outside lab.  Briefly, when they run our HPLC assay, they get a lower result for the same sample as we do.

Our instruments are pretty comparable, and we've tried different instruments at both labs to remove the error.  We use a Water's UV detector, they've tried both the PDA and the UV detector.  We tested samples before we shipped them for their work, and when they had problems, we had them return the same sample for analysis.  We're both using the current USP standard, but they sent us an aliquot of their supply for use to evaluate.  But still, we get slightly higher assay results, no matter if we use our standards or theirs

Their peaks shapes are a little less symmetrical, but they did try new columns, and, (I'm pretty sure) usual systems checks.  And yes their areas are lower than ours.  But why I don't understand is a reason for why a standard and sample won't scale, despite possible differences.

Consider, we both build a standard shooting for 0.110 mg.  We expect are sample to fall with the range of 0.105-0.115 mg.  So our system gives peaks with 200,000 AU (arbitrary units, as I like to always say) and theirs gives 180,000 for the standard.  The response should stills scale for the unknown, giving the same answer.

*EDIT -- changed level*

[marquis]

Good question.  Maybe an audit, in person, would be the best approach.

Here are a few things to check:

Ask them to go over their standardization procedure, step by step.

Check their system parameters.  Waters and Agilent (best guess as to the other
instrument) used to have different parameters to do the same thing. And a slightly
different integration parameter would give the difference you're seeing.

Is there a dead volume somewhere (column, injector, whatever) giving the
difference in peak shape?  That could also lead to integration differences.

The instrument vendor tech support might be able to help.

There are a thousand other things that could do it.  Good luck.

[marquis]

Another suggestion.

Maybe a theoretical plate check on the two sets of columns would be a good idea?

[Arkcon]

Thanks, marquis:, my manager did go over many of those things and a trip halfway across the US to double check isn't in our budget.  There are a few differences, in the systems, and we can always check the plate counts.  We were thinking that threr might be a subtle quantitation error, that we hadn't accounted for, that would jump out at somebody.

[voidSetup]

I would say it sounds like a sample prep issue. I'm assuming its a solid sample. If the particle size or crystal form of your sample is different than what the USP standard is, it's possible they are just not fully dissolving the sample. Are you sonicating for a period and they are not?

[Arkcon]

The sample is a liquid formulation.  We simply mass out a liquid sample and dilute it to a certain volume in a volumetric flask.  That's one thing I'm left wondering.  One place I worked had analogous values they couldn't pin down, it turned out, the volumetric pipettes they'd purchased were simply not calibrated correctly at the manufacturer.  That's certainly a weird state of affairs, they're supposed to be verified.  But there's lots of steps where such errors can occur.  Maybe if they remain stuck, I'll suggest they do the dilution by mass once, just to be sure.

Sonication is a funny thing.  When building the USP standard from the powder, we do sonicate, and they give us a duration that we rigorously follow.  Not enough sonication could fail to dissolve solids, but excessive sonicatin can be bad too.  Sonication can break bonds that aren't usually broken thermally, so you can end up with loss or breakdown products that no suspects or even recognizes.

Our sonicator is in sad shape.  Its so weak, I often make the joke I wouldn't use it to clean my jewelry, and I don't even own jewelry.  (There's a meta-joke in there somewhere.)   So ours may be weaker than theirs.  Over sonication may just be the sort of experiment I can try to replicate their results.