Help with Quadratic Calibration Curves on Agilent 7000 GCQQQ

[Aviator]

Hi all,

Apologies if this is on the wrong board, I'm not sure whether this a prep or instrumentation issue!
I'm trying to get one of the methods developed by our development chemist suitable for a production environment in the laboratory (i.e. perform routine testing for polar and non-polar pesticides in potable and waste water matricies). Unfortunately whilst the method works, it doesn't work well and whilst we've made a number of improvements, there's an area I'm stuck on.

Basically, all of our compounds have a tendecy to behave quadratically. The extent to which the calibration curve is quadratic varies from run-to-run. If everything works well, the calibration curve (although set as quadratic on the software) is linear for the majority of the compounds and the QCs tie in. When it doesn't work so well, most of the compounds will show a degree of bend in their curve. In these cases, the flatter curves correlate with passing QCs whereas the ones that fail can be predicted as to the extent they will fail and whether it will be high or low judging purely by the shape of the calibration curve.

My question is: What can I do to keep my calibration curves nice and flat?

Thanks in advance!

Aviator

[bisnettrj2]

Is there a region of your curves where you have linear increase in response? Can you calibrate across a narrower range, where you will avoid the quadratic part of the curves?

[James_Ball]

Are your QC check standards at concentrations within the upper and lower calibration levels? If they are beyond those limits then when quadratic is used they will usually fail. Also if they are at the extreme limits have you tried different weighting of the curve to cause better fit? I have some compounds that are normally quadratic and I have to adjust the weighting of the points on the curve so that the lowest standard will be accurate when quanted against the curve they were used in. It is always a good check to requant all of your standards against the curve to make sure they all give accurate values, if not the curve is not fitting well. The very lowest point should at least be within +/-50% of the true value.

Instrument side, if you are using the same calibration levels and one time you get a linear fit and the next it is quadratic you may want to be sure you are not getting breakdown or adsorption in the inlet due to contamination or active sites.

[Aviator]

Hi both, thank you for the replies.

This is definitely an instrument problem. The analysis worked well durinig our performance testing and even now, can work really well. The problem with the calibration curves suddenly going very quadratic is intermittent and there deosn't seem to be any logic to it (i'm sure there is, we just need to work out what's going on)!

James: Very good points. We found out recently that our internal standard solution had inadvertantly become contaminated and remaking that has fixed a lot of the problems. We're still getting random issues with the majority of the cal points being fine, but for some compounds the points will either drop or raise by a significant margin (for example, a cal 4 (150ng/l) sometimes comes out as high as a cal 5 point (200ng/l) for some compounds (but not for others), and other times one of the high cal points (4 or 5, 5 being the highest) will drop; sometimes by a bit, sometimes by a significant margin. This is also a new instrument for us with a new design of liner so we're trying to work out how many injections we can get roughly per liner. It could just be a simple active site issue but the fact that it's only just started to be a problem now (with no change in our maintenance schedule) seems strange.

Our chromatography is starting to suffer so we're going to replace the column. That usually has the wonderful effect of fixing a lot of woes so we'll see if it does this time

All the best,

Aviator

[James_Ball]

I will be interested to see how this works out since we will be getting our first 7000QQQ in a couple weeks. I am not completely familiar yet with the software for this one but I have run LC/MS/MS and I know that if you are setting up your MRMs in segments and Retention Times drift enough to place your target peak in the wrong segment you will get similar problems.