Chromatography Forum

Hello. I have a problem about Ion trap :Agilent 1100 XCT, ESI. I use it for quantitative analysis. My calibration curve is 0.2-20 ng/ml and my samples are urine. My problem is about peak area, it is unstable. After we run for 20 samples the QC sample's peak area is decrease about 2 flod and afer cleaning ESI the peak are is increase. I dont know why and How to do for stable peak area. Another problem is about calibration curve. Some papers they use weighting factor. I want to know when we should to use weight factor and someone said in one analysis sometime we can weight for good calibration and sometime we dont weight if it is good enought. So it make me confuse.

On the issue of peak area drift: Is this urine being injected without any cleanup? If so, I highly recommend you use a divert valve, to send the salt and polar components to waste, during the first few minutes of the run, then switch back closer to the retention time of your analyte. I can't remember if the XCT had a built-in divert valve, but if so you would just use that. Or if you are using an Agilent 1100 LC system, and if it has the optional column-switching valve, you can use that as your divert valve.

Regarding weighting, I can't recall an LC-MS/MS analysis where I did not need to use at least 1/x weighting with my regression, and often I have to use 1/x^2 weighting. Compare the measured concentrations of your calibration standards to their actual concentrations, with no weighting of your curve (don't just look at R-squared). If the percent error for your low standards is greater than about +/- 15% to 20%, try 1/x weighting. If the error is still too high, try 1/x^2 weighting. If you are able to reduce the error for your low standards, but this causes your high standards to have unacceptable errors, then you have a linearity problem.

...someone said in one analysis sometime we can weight for good calibration and sometime we dont weight if it is good enought.

We have written some analytical protocols leaving the weighting up to the discretion of the analyst, when our QA department would let us get away with this. In other cases where we knew the analytes did not have great linearity, we would simply require 1/x^2 weighting in the protocol.

Thank you for your suggestion. It is good suggestion for me. I prepare sample by solid phase extraction and I use a divert valve. I only inject sample during 7.4-9.0 min into MS but it still has peak area drift. About peak area drift, I have consulted an expert about Ion trap in my country, he suggests me to have correction by standard, example if the area of standard that inject after 10 samples decrease 2 folds, I will correct by plus 2 with sample’s peak area. Do you think about this? I don’t sure is it ok to do that?

... don't just look at R-squared

Why you said that? I think R-squarded is important and I have to make 0.99. If I use 1/x^2 and percent error can accept but R-squarded is 0.97. Is it ok or ecceptable?

So you are cleaning up your sample and using a divert valve, which is good. But you are still getting drift. Other possible causes for the drift: (1) Inconsistent LC performance. Check retention times, and if they are also drifting, fix the LC method. (2) Buildup of strongly retained material on column, which begins to elute and cause ion suppression after a enough samples are analyzed. Wash column with a strong solvent during each run after the analytes elute. In my experience this is a more common problem with plasma than with urine. (3) Spray chamber parameters are not properly optimized, causing the source to become dirty prematurely. Make sure your spray chamber parameters are appropriate for the LC mobile phase and flow rate.

About peak area drift, I have consulted an expert about Ion trap in my country, he suggests me to have correction by standard, example if the area of standard that inject after 10 samples decrease 2 folds, I will correct by plus 2 with sample’s peak area. Do you think about this? I don’t sure is it ok to do that?

I have never heard of doing this. Normally if our QC is outside of specification, then all sample results bracketed by that QC are considered invalid, and the samples must be reanalyzed. If you cannot correct the drift problem, then I recommend you use an isotope-labeled version of your analyte as an internal standard. Then if your analyte area drifts by 50%, so will your internal standard area. Since your quantitation would depend on the ratio of analyte to internal standard, this would correct for the drift, and is a commonly accepted practice.

Why you said that? I think R-squarded is important and I have to make 0.99. If I use 1/x^2 and percent error can accept but R-squarded is 0.97. Is it ok or ecceptable?

You can look at both R-squared and percent error. What I meant to say is that you should not look only at R-squared. It is possible to have R-squared of 0.9999 with an unweighted regression, and still unacceptable percent errors at the low levels. On the other hand, I have never seen a calibration curve with percent errors less than 15% at every point (i.e. acceptable), and an R-squared of 0.97. I am not saying it is impossible, but I have never seen it with LC-MS.

Thank you for your suggestion again. It is always useful suggestion for me.
LCMSMS is a new technique in my country. I have no experience before. This is my first time so I need many consultants and try to discuss with people who work in this flied like you.

After read your suggestion, I turn back to review my method. I think may be spray chamber parameters are not properly optimized. It is very difficult but I will do it best.

On the other hand, I have never seen a calibration curve with percent errors less than 15% at every point (i.e. acceptable), and an R-squared of 0.97. I am not saying it is impossible, but I have never seen it with LC-MS.

It is possible with my calibration curve, percent error less than +/-15% at every point but r-squared = 0.97

About quantification by LCMSMS that publish in many journals. Most of them use triple quadropole but now I use Ion trap. What do u think about this?

About quantification by LCMSMS that publish in many journals. Most of them use triple quadropole but now I use Ion trap. What do u think about this?

Triple-quadrupoles will give you better linearity over a given concentration range, so a triple-quad is a better choice if the instrument will be used primarily for quantitation. On the other hand, ion traps are more sensitive than triple-quads for full-scan and product-ion scan experiments. So if the instrument will be used primarily for qualitative analysis, then an ion trap is often the better choice. But ion traps can be used for quantitation, they are just not ideal.