Chromatography Forum

Hi Experts

Who can help me out here!!

Recently, i am doing a method development for impurities test in our product by HPLC. The y-intercept of my linearity is negative. And i am pretty sure it is the bias of intercept. Some of my analyte seems disappeared in the system. Initially, i thought there might be some absorption by injection vial. but when i tried to inject different volume to the HPLC system from one vial, the y-intercept is still negative.

Then, where does the analyte go? That really bother me a lot

Thanks

It's almost impossible to tell based on the very limited information you have given. Some thoughts:
- how negative is the intercept? If you are doing a multi-point calibration, statistical software (even Excel) will calculate the standard error of the intercept. If your intercept is within that standard error, then it's not a problem except insofar as it affects the LOD and LLOQ
- if you are covering a wide range (2 orders of magnitude or more), your data points at the low end should be more closely spaced than those at the high end (a given percentage error at the high end has more effect than one at the low end.
- are there visible problems with peak shape that might affect quantitation (for example, severe tailing(?
- is your analyte stable under your chromatography conditions? Is there a chance that some might be adsorbed by active sites on the column packing?

Yes, titration of some of your analyte onto binding sites in the system is possible. But also, some detectors can be non-linear in weird ways (especially things like electrospray MS). And yes, check integration. If, for example, you're losing a roughly constant lump of the bottom of the peak because the integrator is a bit slow to detect the peak start, this will translate to a negative intercept (I assume throughout that you've got signal on the Y-axis and amount on the X-axis, which is Traditional but Wrong).

tom jupille wrote:
It's almost impossible to tell based on the very limited information you have given. Some thoughts:
- how negative is the intercept? If you are doing a multi-point calibration, statistical software (even Excel) will calculate the standard error of the intercept. If your intercept is within that standard error, then it's not a problem except insofar as it affects the LOD and LLOQ
- if you are covering a wide range (2 orders of magnitude or more), your data points at the low end should be more closely spaced than those at the high end (a given percentage error at the high end has more effect than one at the low end.
- are there visible problems with peak shape that might affect quantitation (for example, severe tailing(?
- is your analyte stable under your chromatography conditions? Is there a chance that some might be adsorbed by active sites on the column packing?

Hi Tom

Thanks for replying

--The standard error is about 0.028, the intercept is -0.087. So it is not just error of the intercept.

--The range i am covering is from 0.05%~1.5% (0.05%, 0.20%, 0.50%, 1.0% and 1.5%), very normal range for impurity.

-- The peak sharp looks good.

-- The analyte is stable.

-- I am also suspecting there might be some absorption on the column or in the LC system including the tube and injection system. If that, how can i do to remove the absorption?

Thanks

Yijun


The LINEST results are :

308.3437111 -0.086832727
5.706067321 0.028049147
0.998973689 0.03984903
2920.091147 3
4.636944756 0.004763836

lmh wrote:
Yes, titration of some of your analyte onto binding sites in the system is possible. But also, some detectors can be non-linear in weird ways (especially things like electrospray MS). And yes, check integration. If, for example, you're losing a roughly constant lump of the bottom of the peak because the integrator is a bit slow to detect the peak start, this will translate to a negative intercept (I assume throughout that you've got signal on the Y-axis and amount on the X-axis, which is Traditional but Wrong).

Hi Lmh

Thanks for replying

From the chromatogram, it is hard to say the integration is wrong. The peak shape looks good. Anyway,if it is the case, how should i improve?