Weighing of SD/average of LC retention time in RSD%

[holmerz]

Hello all
I'm in the process of qualifying our new LC-MS. The specification for repeatability of the retention time is set to a %RSD of <1%.
Initially the measured retention time was reported with one decimal, and in five runs the observed retention time was reported to be 2.8 mins and one was 2.7 mins, resulting in a %RSD > 1%.
Of course when the retention times are reported with three decimals, the standard deviation gets lower and hence the %RSD gets lower.
However, if the observed retention times, for some reason, were in the 8 mins area and the standard deviation was identical, the %RSD would be lower.
Is there a way to "take out" the impact of the magnitude of the retention time?

Of course, I assume that the variation in retention time is independent of the magnitude of the retention time.

In advance thanks
Erik

[lmh]

I think you've got two hurdles to overcome, both of which result from an incomplete specification of how to do the RSD.
(1) Someone needs to specify how you're using the LC system to generate the peak from which you measure the RSD. To take a really bad example, if you were to run a reverse phase method for a well-retained test compound, with a program that runs 1% acetonitrile for 40 minutes, then switches abruptly to 100% acetonitrile, the worst LC system in the world will give you an RSD better than 1%, because all it's doing is timing 40 minutes! The gradient-making could be complete rubbish, and the flow rate could probably vary 20% run-to-run.
For gradient methods, you will probably get a better precision from a longer method. For isocratic methods, you may find that the precision is more unchanging (because it depends very much on the pump flow; in an isocratic method, the peak will appear after X column-volumes of solvent; if the pump is out by 1%, the peak will be out by 1%, irrespective of whether it's early or late).
(2)… and the problem of early peaks is also why you're in a mess if the retention time is reported to only one decimal place. If you were to look at a very early eluting peak, say at 2 minutes, then with your measurement accuracy of 0.1min, your retention times are operating at 5% intervals, so it makes no sense to try to estimate whether they're accurate to 1%.

Basically, whoever wrote the specification/acceptance criteria needs to go away and specify the method that will be used to meet those criteria, including the precision of reporting. I would suggest choosing an easily-reproduced method that is representative of how you intend to use the instrument, and collecting data with a precision of at least one decimal place more than you need.

[holmerz]

Thank you a lot for your answer.
I think you're right in that the procedure needs to be specified in greater detail, however I only have guidelines from the
"Qualification of Equipment – Annex 7, 7.2. Qualification of Mass Spectrometers”, European Pharmacoeia 8.8, PA/PH/OMCL (10) 86 R6".

They suggest the use of a mixture of Betamethasone 17,21-dipropionate and Betamethasone-17-valerate, maybe with the intend of one of the two compounds acting as an internal standard. I guess you could then operate with a relative retention time?

Also I was looking for a statistical manoeuvre to take out the impact of the amplitude of the retention time.

Thanks again
Erik