Chromatography Forum

Hi Guys
can anybody confirm me Linearity and Accuracy are must for HPLC autosampler calibartion , if so, can you please share me the any refrence article from regulatory agencies like FDA, EU etc

What do you mean by Linearity and Accuracy for an Autosampler? Do you mean injection volume specifications?

There a ways of testing accuracy such as removing a number of fixed volume injections froma water filled vial. Knowing the density of water, the temperature and the starting and finish weights of the vial will get you accuracy.

Linearity can be assessed by performing the exercise at different injection volumes and plotting wt. loss/injection v volume setting.

can of worms suddenly opens! If you habitually make 5-10uL injections, is a balance that reads to 0.1mg adequate? If you take 10 injections from a vial and re-weigh, is your result relevant to assessing the error on a single injection (which is what actually matters in practice)? And some autosamplers (e.g. Thermo's Surveyor, and probably Accela) are designed such that they take a slight excess volume and only inject part of it, so on these instruments, the weight removed from the vial is not a measure of the volume injected.

Once you've established detector-linearity (by same-volume injections from reliable standards of different concentrations) you can test autosampler linearity by checking peak areas of different volume injections from a single standard. On an instrument such as the Thermo ones, which take the excess volume, after confirming linearity, you can also compare partial loop (excess volume) injections with full-loop, where the accuracy should depend only on the sample loop (which of course you buy commercially and is therefore fairly solid). Unless you have an instrument that is guaranteed to inject exactly what it takes from the vial, and you have a very good balance, it's quite hard to validate the instrument without things getting circular (validation of detector depends on autosampler, and validation of autosampler depends on detector).

Incidentally, I personally would be interested in how important people generally regard this sort of instrument testing as being? In a lab that doesn't work under regulations, but wishes to make decent measurements with limited hours for validation, I prefer to spend more time checking methods than instruments. The reason is that it's almost impossible to set up a reliable, repeatable and accurate method on an instrument that isn't repeatable, but it's very easy to take a perfect instrument and create a method that generates random numbers...

^ saved me a lot of typing!

We only test for linearity and precision, not absolute accuracy. If we get r²>0.99 for injections ranging from 2 (or 5) to 100µL, and we get good precision on replicates of the 10 or 20µL injection (RSD<1% for n=5), we are happy. We run this test using a C18 column and a 4 component test mix or we make up some methyl & propylparabens solution (anything w/ nice sharp peaks will do).

The only way to check absolute accuracy is to tare, fill and reweigh a loop. This requires a good balance, good technique and still leaves you guessing as to the volume associated w/ the channels in your valve (and I've heard that this can vary by over 1µ for popular 6 port Rheodyne valves).

Why would linearity or accuracy of injection volume (within sensible limits) be of any relevance at all ? Surely calibration standards and samples are, without exception, injected at the same volume (in the same batch and on the same instrument) and so any deviation from the nominal injection volume will cancel out perfectly.

Peter

Surely calibration standards and samples are, without exception, injected at the same volume

Actually, that's not true. In many cases, calibration curves are generated by injecting different volumes of the same solution.

The argument has been made that an autosampler may be more accurate than manual pipetting/dilution (or, at least, sufficiently accurate, and easier). That said, I prefer constant injection volume!

[quote="tom jupille"][quote]
The argument has been made that an autosampler may be more accurate than manual pipetting/dilution (or, at least, sufficiently accurate, and easier). [quote]

I confess myself astonished Leaving aside the possible contribution to true injection volume of extra little bits of pipe and channels in valves, and the oft repeated on this forum effects of different volumes of injection solvent on peak shape - if the technicians can't pipette and dilute accurately enough to make up standards then how can you trust them to prepare samples ?

If you calibrate by injecting different volumes of a fixed concentration you are calibrating response vs analyte mass. You then have to know the real injection volume in order to calculate concentration in the vial, so your injector now has to be accurate as well as precise, and as already discussed, determining its accuracy is not a trivial excercise.

OK, I'm going onto internet shopping not to buy a flame-proof suit .

Peter

And right on cue:

viewtopic.php?f=1&t=16194

Peter

Slightly off the original topic, but one could use the auto-additions function present on many autosamplers to have the auto injector mix the standards yet retain the same total injection volume. If you have a blank solution,a standard stock in the same solvent, and your desired injection volume is 10 ul, you could do 10+0, 8+2, 6+4, and so on. Some validation would be required but I don't see that as too tough.

I'm actually tempted to try it and run a head to head comparison to see whether my 2695 can beat me.

Slightly off the original topic, but one could use the auto-additions function present on many autosamplers to have the auto injector mix the standards yet retain the same total injection volume. If you have a blank solution,a standard stock in the same solvent, and your desired injection volume is 10 ul, you could do 10+0, 8+2, 6+4, and so on. Some validation would be required but I don't see that as too tough.

I'm actually tempted to try it and run a head to head comparison to see whether my 2695 can beat me.

Elegant ! I for one would be very interested in how good the autosampler is.

Peter

I'll try it this afternoon and let you know how it works out. Cheers!

I do calibration curves both ways, but I have no idea whether I'm acceptable...

My feeling is that making different volume injections from one vial shouldn't be a problem provided the injection solvent is weak. In fact this is something I do get upset about, and I very, very rarely inject samples in a solvent stronger than the start of the gradient. Also I don't tend to push limits on high-volume injections; I can't remember the last time I injected more than 20uL on a 2mm column.

The usual advantage people cite for making multiple standards at different concentrations and injecting the same volume of each, is that the calibration points are somehow more independant. I sort-of feel this is a bit of a spurious argument because they've usually been derived from a single stock by some manual dilution process. And yes, the autosamplers in my lab are definitely better than me at pipetting.

On the other hand, when using the Surveyor system in our lab, I always use multiple vials, because for various reasons I don't entirely trust its injection volumes to be accurate (although I do trust them to be repeatable).

Of course, the trouble with multiple vials is that they're more expensive on consumables!

If I understand gclc's autosampler question correctly (linearity, accuracy(?)...perhaps precision?) then we may consider strictly from a hardware performance perspective:

For basic autosampler performance, some key autosampler modular qualification performance tests performed (without in-line consumables (no column, using backpressure coil)) as part of the wholistic system qualification) by your vendor's service engineer will include in the following order:

Injection Precision: Multiple same volume injections from the same vial (e.g., 10 injections from the same vial of an NIST traceable single standard) to obtain %RSD (Relative Standard Deviation) based on peak area counts from each injection

where,
Relative Standard Deviation = 100 × [(standard deviation of array X) ÷ (average of array X)] = RSD expressed as a %

Injection (%) Carry Over: high concentration NIST traceable standard injection followed by 1 or 2 same volume injections of other sample liquid from a different vial (e.g., 10 uL of 1000 ppm uracil (or caffeine, nitrate, or some other highly miscible substance), then 10 uL of DI ASTM Type I water or HPLC spectrophotometric grade), then compare the residual peak area responses of the carryover injection vials to the high concentration peak area response expressed as a percentage

Injection Linearity: Same vial injections with different fixed volumes (e.g., 1-, 5-, 10-, 30-, 50-uL injections using a 100 uL injection loop) of a single NIST traceable concentration standard then plot results of area response as a function of injection volume then obtain a linear correlation factor (r or r^2) of 99.9XXX (# of X's dependent upon vendor's OQ/PQ specifications)

For autosamplers with sample preparation/autodilution capabilities, the performance test would include:

Autosampler Dilution Precision & Accuracy: this will vary depending upon the vendor

Some instrument vendors may also include an autosampler thermal compartment/Peltier performance test option and would include:

Thermal tray temperature accuracy/precision performance test: vial or vials filled with liquid (usually water) placed in the thermal vial rack and measured with a thermocouple inserted and recorded from a certified/calibrated multimeter

Each of the above performance test results will then be compared against the instrument vendor's qualification specifications for PASS/FAIL criteria. The GxP customer would then archive these reports for their compliance records.

The user can then modify the performance tests to apply to their application PQ (with columns, chemistry, etc.) as part of their SST criteria...

Hope this helps...

Not going to delve into the injection 'accuracy' part of the discussion here...

Tom:

Can you ban someone for the continual posting of an irrelevant link?

** No, but I can (and did) move it to the "Around the Water Cooler" section.
-- Tom Jupille / Admin
4/23/11 8:08am **

Autosampler Dilution Precision & Accuracy: this will vary depending upon the vendor

Mr John Lim, we perform the precission test for auto-dilution by using one stock solution and then diluted to same concentration for 6 injection. we also perform a 'linearity' test by diluting one stock solution to several concentration for each injection. is your auto-dilution accuracy test same with our 'linearity' test? thanks in advance.