Calculating limit of detection for nonresolved peaks

[degenchem]

I have an application were I need to calculate the limits of detection and quantifiaction for a small peak which comes on the tail of a much larger peak.

Given that the LOD is 3x and the LOQ is 10x the noise, it seems to me that just taking the baseline noise of a blank would underestimate the LOD/LOQ, or am I wrong? The peaks are integrated by dropping a line to the baseline (no "tangential" or "valley-to-vallley" integration).

How should I calculate the LOD/LOQ in such cases?

[AdrianF]

If you can't use skimming you are wasting your time - the integration is not fit for purpose. Your only hope is to improve the chromatography to resolve the peak or use a different integration package!

[bartjoosen]

another approach to LOD and LOQ is to use the standard error and slope of the calibration line at near LOQ level.

See page 15: http://www.ich.org/LOB/media/MEDIA417.pdf

Bart

[Jumpshooter]

So, inject increasing amounts of your sample extract on your LC column, then collect the eluates at the waste end of your LC setup that correspond to the Tr for the "low area peaks". In this way you can collect increasing amounts of the compound that is giving the "low area peak". Put the different "amounts" in autosampler vials---use these as your Calibrants on the next analytical run to assay the "low peak area compound". Then, you will have generated a standard curve that displays "low peak area compound" amount vs. "peak area". Voila! your calibration that will enable you to quantify (LOQ). You can also do serial dilutions to further examine the LOQ

[AdrianF]

It would be helpful if you could post a typical chromatogram.

[bartjoosen]

So, inject increasing amounts of your sample extract on your LC column, then collect the eluates at the waste end of your LC setup that correspond to the Tr for the "low area peaks". In this way you can collect increasing amounts of the compound that is giving the "low area peak". Put the different "amounts" in autosampler vials---use these as your Calibrants on the next analytical run to assay the "low peak area compound". Then, you will have generated a standard curve that displays "low peak area compound" amount vs. "peak area". Voila! your calibration that will enable you to quantify (LOQ). You can also do serial dilutions to further examine the LOQ

Maybe I missed something, but you can never be sure about the exact amount in your vial? It's diluted with mobile phase, so if you inject it again from a collected fraction, it's still more diluted on the column...

Bart

[Noser222]

Using the peak height instead of the area may be more beneficial for your purpose, assuming the situation isn't too bad.

[degenchem]

Thanks for your input so far. Here is a sample chromatogram:



I have the "small peak" compound available in pure form, so I can easily calculate the LOD/LOQ with no obstructing peaks. It just seems like the value is too low for the sample matrix shown in the chromatogram above.

As for the slope of the calibration line/standard error method -- wouldn't that method also underestimate the LOD/LOQ in this case?

[philippem]

Hi,
Is it possible to change your mobile phase compoition or gradient profile ? I assume you are using gradient.

Philippe

[degenchem]

Hi,
Is it possible to change your mobile phase compoition or gradient profile ? I assume you are using gradient.

Philippe

Unfortunately the method can't be changed at this point, so we'll have to live with the poor separation.

[AdrianF]

If you are unable to change the chromatographic conditions you must change the integration method to give a more accurate idea of the area. The current area will be about double the true area of created by that substance.

PS why are you unable to change the conditions? If it is important to quantitate that the conditions need changing.

[HW Mueller]

But you know that your results are wrong, so why continue?

[philippem]

taking into account the last two remarks, one can say that the method was not (yet) completely developped, and probably also not validated !

so why couldn't you change it to have a better separation between the peaks

anyhow succes


Philippe

[bartjoosen]

Thanks for your input so far. Here is a sample chromatogram:

I have the "small peak" compound available in pure form, so I can easily calculate the LOD/LOQ with no obstructing peaks. It just seems like the value is too low for the sample matrix shown in the chromatogram above.

As for the slope of the calibration line/standard error method -- wouldn't that method also underestimate the LOD/LOQ in this case?

Do you have the "big" peak also in pure form?
Otherwise you can do an accuracy test, and see if there is a significant bias. If so, the method isn't sufficient to quantitate the minor peak, and you can't determinate an LOD/LOQ (or an amount....)

If it is accurate, the slope/se method wil be good enough to estimate the LOD/LOQ.
If you are afraid to underestimate, you can integrate valley to valley, this will overestimate your LOD/LOQ.


Bart

[tom jupille]

Dyson put the question into perspective almost a decade ago:

“. . . errors arising from peak overlap are introduced by the algorithms of perpendicular and tangent separation and cannot be eliminated by anything but better chromatography. Integrators are able to generate a highly precise and totally inaccurate set of results for all the foregoing examples.â€