Injector / Detector Linearity

[Oliver]

Anyone has ideas on this:
We are doing trials with a new performance check that we developed in house from several references. Two of the tests are injector linearity injecting 5µl, 10µl, 25µl, 50µl, 75µl, 100µl, and another test is injecting 10µl of an solution at 50%, 80%, 100%, 120%, 150% concentration.
The limit we set from litterature was R²>0.998
but we are failing all checks if we plot results in force through 0 mode, and fail some of the tests if we plot without force through 0.

My questions are:
is R²>0.998 too tight?
for prooving linearity for performance test should the data be plotted in force through 0 mode?

The HPLC is set up at 1.5ml/min, 90% ACN, 40ºC. injecting a solution of about 100µg/ml Anthracene in 100% ACN.
BTW is there something less hazardous than Anthracene to use but which gives short runs and has a defined lambda max that can be used for wavelength calibration (as anthracene is Lambda max at 251nm)?

I know its a bunch of questions but maybe i'm won't be the only one to learn from this. Thanks in advance.

[HW Mueller]

Another one for satisfying demigods? If you want to know whether your system is working you may look (and report here if not understood) whether the retention times and peak shapes are constant in the diff. volume injections (solvent incompatibility?) or wether only the areas are different (laminar flow problem in loop?). In both type of runs why not go to 0 concentration or 0 volume, then the question of whether one should force through 0 would be answered, and if it didn´t hit zero (within.......) you better find out why it doesn´t (carryover?).

[DR]

You are asking for trouble forcing the line through zero. You are also asking for trouble using 100% ACN in some pumps.

I would suggest using caffeine in 10% ACN, 0.1% TFA (if needed for peak shape) on nearly any C-18 column (about 5 min @ 2mL/min should do it). Caffeine has a nice local max, min at 272, 244nm. You will have to be careful about your concentration if you want to run all of those injection volumes with 1 solution... Maybe 3-4µg/mL if you are doing your linearity at a low wavelength (216-220nm area)?

[Consumer Products Guy]

" Two of the tests are injector linearity injecting 5µl, 10µl, 25µl, 50µl, 75µl, 100µl" - for sure, my lab would NOT do this, no reason as we do not vary injection volumes during a sequence

"another test is injecting 10µl of an solution at 50%, 80%, 100%, 120%, 150% concentration." - this is similar to a standard multi-point linearity study, this we would do. I would probably add some data points at 25% as well, and obviously, I'm sure you're also doing a 0% blank.

[Uwe Neue]

What is the response of your detector at the highest injection? (It could be that you are overloading the detector.)

[tom jupille]

I'll add additional commentary to both CPG's and Uwe's posts:

Changing injection volume is considered bad practice (CPG was being polite!), so injection volume accuracy is generally not relevant to HPLC performance (i.e., since you will inject the same volume for the calibrators and the samples, you don't care too much what that volume is, you just want it to be the same everytime).

Uwe's point is well taken, and it addresses a known weakness of non-weighed least squares to characterize chromatographic data over a wide range. The standard least-squares fit implicitly assumes that the data set is homoscedastic (the average errors have the same magnitude across the whole range). Chromatographic data tend to be heteroscedastic (the percentage errors have about the same magnitude across the whole range; the absolute errors therefore are larger near the upper end). In effect, errors at the high end tend to dominate the fit.

For narrow-range data (e.g., potency, where your range has to be 80 - 120% of stated value), this isn't a problem. For wide-range data (e.g., drug metabolites by LC-MS) it can be a very big problem, and a variety of weighted least squares fits are used to get around it.

[Mark Tracy]

Your autosampler probably has a manufacturer's spec on accuracy and precision. Find out what that is and how they test it. Don't set a spec tighter than that. You may have a problem with air in the autosampler, especially if the precision is poor at small volumes.

I vote for caffeine as your standard. That's what we use at Dionex. You can even buy our certified standard kit.

You don't need a column for these tests. We use 13 meters of 0.12 mm i.d. PEEK tubing in place of a column. The column is just one more thing that you need to validate, document and control; avoid that exercise if you can.

[Oliver]

Hi thanks for the feedback all of you

DR:
mobile phase is 90% ACN, only the anthracene solvent was 100% ACN. Thanks a lot for the Caffiene detials, is the lambda max in that same mobile phase.

Uwe:
50% (5µg/ml) ~250mV
150% (15µg/ml) ~650mV
blank 0.05mV
5µl ~250mV
100µl ~2000mV (peak not truncated)
Carry over less than 0.1mV

tom:
I got lost there in the homo and hetero data. I remember reading something about it in LCGC a while back, but i didn't have such a mathematical training before. I'll look it up again since now it seems more relevant, because i would like to verify detector response linearity over an extended range due to related substances tests.

Mark:
A restrictor tube is on the way but hasn't arrived yet. Thanks for the point.

I included injector linearity because it was in the old SOP. I understand the argument of having always the same volume, but I thought that this is required as a diagnostic for the injecor and left it in the draft updated SOP.

But now I dare put the argument that injecor linearity is not required, indeed the manufacturer PQ does not test for it but for accuracy at 50µl by weight. I will try to do it by weight also but would see the RSD between the weight loss from the vial of separate injections.

[Uwe Neue]

At 2000 mV, you are overloading the detector. You may not see a truncation of the peak (yet), but I am fairly sure that you are touching the limits of linearity.

Does your R^2 get better when you throw out the highest numbers?

[DR]

Doing injection accuracy by weight has its own problems.
With some injectors, the entire amount of sample withdrawn is injected, so you can do a gravimetric injection verification. To do this, you need a semi-micro balance and usually) a very light weight injection vial. Weigh the vial, do a 50µL injection from it and reweigh. If your balance is adequate, you may be able to see errors in the 0.5-1% range. Others have advocated weighing the injection loop empty & full but this approach fails to take the internal volume of the valve body into account. Per Rheodyne's own admission via a phone call (several years ago), there was something like a 17% RSD on the internal volume of their valve bodies for older 7 series 6 port valves.

Basically, it is most important that any given injection volume be highly reproducible. It is a luxury that injectors are also linear (like when you have only 1 prep of the last of something that is too dilute to see at a given injection volume, so you can increase the volume and not feel too bad about doing so).

[Oliver]

Uwe: I think thats it! the detector non linearity.

We had another HPLC of the same model and a couple of weeks ago passed the 5µl to 100µl test (R²: 0.9999) but using a pharmaceutical active as substance. Today we did the same linearity with the anthracene solution and it failed!
I have the reports here and i notice that the peak area of the 100µl injection using the pharamceutic on HPLC 2 was ~6M, while the 100µl injection using anthracene on HPLC 1 was ~16M. Tomorrow I'll check the peak heights... now I'll bet that this is one factor.
I recalculated the linearity using the values up to 50 µl where the area was ~9M the R² is 0.9977 or 0.998 that is just within limit.

Thanks a lot for your help every one!

[Victor]

DR- please elaborate on your statement about trouble using pure acetonitrile with some HPLC pumps.

What trouble have you experienced? With which type of pumps?

Thanks.

[DR]

Pumps that use a ruby ball and sapphire seat in their check valves sometimes have intermittent sealing problems when fed pure ACN. Either adding a bit of water to the ACN or switching to ceramic check valve components will generally cure the problem.