[mandytan]

In a HPLC chromatogram for a standard, the peak area is 20000.
2000 times dilution of the standard gave a peak area of 8.

The peak area of this diluted standard is used for impurities limit test (i.e. the area of any impurity must be less than the peak area in the diluted standard, which in this case it would be .

Theoretically,a 2000 times dilution of the standard should give an area count of 20.

Is this huge difference in area count (8 versus theoretical 20) due to the large dilution factor?

[Arkcon]

It could be an error in dilution, or it could be caused by a variety of integration errors.  The shape of the peak may change, and fail to integrate the same way.  At very low levels, the baseline really affects the area, if you zoom in repeatedly, you'll eventually see the baseline noise and how much of a small peak can be gained or lost.  The optical system may not be linear over a very wide range -- although that is rare.  You initial large peak may have saturated the detector ... so that it was really bigger than it appeared.  A really small peak may be under represented because there is a time factor ... the detector can only look at the flow cell so many times a second, it may miss some of a very narrow peak.  Remember, absorbance has to be above some threshold greater than baseline in order for the software to say "peak starts here."  A low absorbance means it ignores more of the peak, so it may misrepresent the area.

Suggestion: try running some intermediate dilutions, to see if there's a sudden point of deviation, or general, allover the place weirdness.

[mandytan]

Thanks Arkcon:)

A blank was run before the diluted sample was injected. The baseline of all the chromatograms look pretty good, ie. straight & stable. It is also an isocratic run.
We "zoomed into the chromatogram" to look at the main peak and used the same integration parameters for both the Standard and Diluted Standard solution. As for pipetting, we actually used a calibrated automatic pipette to perform the dilution.
Is it possible that this huge difference in area count (theoretical 20 versus experimental could be due to the sample itself?

[mandytan]

Sorry, there was a typo error. The Standard peak area is 40000 and a 2000 times dilution would give a theoretical peak area of 20.

[Arkcon]

I've used automatic pipettes that I thought were calibrated and turned out not to be.  So that's one thing that could have been a problem.  Also, your intermediate samples homogeneity could be amiss.  Also, some samples viscosity and density can be so far from the calibrant (usu. water) that you can get surprising results.  Or maybe you simply did it wrong.   Try to run some intermediate dilution to see if you can track this down.  Try some dilutions made with non-automatic pipettes and volumetric glassware, and see if you get much better results.  Although, there's always the chance that those are calibrated incorrectly ('tho that is very rare.)

We'll start with this sort of investigation, then we'll consider if the sample is doing weird things.  Again, at very small levels, the detector may fail to see the sample well enough to integrate the baseline the same way as it would for a bigger peak.  Running intermediate samples will help us understand that situation as well.

[Arkcon]

Beyond what I've said, we don't know if the system has some sort of slow leak, or intermittent sample injection problem, so some carefully made, serial dilutions being run would be a good thing to check out the system.

[fledarmus]

Definitely some serial dilutions - a two-point calibration curve across three orders of magnitude is not something I would have a lot of faith in. It would be really nice to have at least three points in the neighborhood of the concentration you expect to be measuring.

It is possible that your integration software is selecting a point too high up the shoulder of your peak to begin quantitating on the low concentration peak. This is especially true if the peak shapes are significantly different - if the low concentration peak is broader in comparison to its height. There may also be a threshold cutoff below which the peak is not quantitated which would be more significant for a smaller peak than a larger one.

[JGK]

In a HPLC chromatogram for a standard, the peak area is 20000.
2000 times dilution of the standard gave a peak area of 8.

The peak area of this diluted standard is used for impurities limit test (i.e. the area of any impurity must be less than the peak area in the diluted standard, which in this case it would be .

Theoretically,a 2000 times dilution of the standard should give an area count of 20.

Is this huge difference in area count (8 versus theoretical 20) due to the large dilution factor?

It could also be due to the linearity of your calibration. have you established even a theoretical limit of quantification (LOQ) or detection (LOD).

What is the concentration of of your diluted STD in relation to the LOQ or LOD of the method?

Have you established linearity (using non zero stds) across the STD - Diluted STD range of concentrations?

[mandytan]

This is a European Pharmacopoeia test method for impurities.
As it is a Limit Test (the area of any impurity must not be greater than the area of the peak in the Diluted Standard),no linearity curve was established. The Standard, Diluted Standard and Sample solutions were analysed and the area of any impurity in the sample solution was compared with the area in the Diluted Standard.

[JGK]

As you have not estalished a linear relationship between area and concentration over the dilution range , the error will likely be as a result of a lack of linearity between area and concentration

[voidSetup]

Is this a compendial method for a drug substance or something?  If it is, then it is probably validated all the way, so linearity should be established.  Can you see any of the impurities in the 2000 fold dilution?

From what I have seen, usually the 2000 fold dilution of the standard is used as a system suitability sample to show that your main peak is detectable with signal:noise greater than 10 for quantification.  The actual quantification is then done against the undiluted standard injections.

[mandytan]

Yes, it is a compendial method for a raw material. The diluted standard contains only the main peak.

[JGK]

Is this a compendial method for a drug substance or something?  If it is, then it is probably validated all the way, so linearity should be established.  Can you see any of the impurities in the 2000 fold dilution?

From what I have seen, usually the 2000 fold dilution of the standard is used as a system suitability sample to show that your main peak is detectable with signal:noise greater than 10 for quantification.  The actual quantification is then done against the undiluted standard injections.

It may well have been validated but not on the equipment Mandytan is using. Unless Mandytan has validated the method in Situ with the present equipment you cannot guarantee linearity. All regulated environmnets require that methods are validated within the lab they are being performed.

[voidSetup]

It may well have been validated but not on the equipment Mandytan is using. Unless Mandytan has validated the method in Situ with the present equipment you cannot guarantee linearity. All regulated environmnets require that methods are validated within the lab they are being performed.

True, good point.  I haven't ever transferred a method, so I'm not sure if you have to go through the whole validation process to verify the method, or just show that the same results are obtained.

[JGK]

It may well have been validated but not on the equipment Mandytan is using. Unless Mandytan has validated the method in Situ with the present equipment you cannot guarantee linearity. All regulated environmnets require that methods are validated within the lab they are being performed.

True, good point.  I haven't ever transferred a method, so I'm not sure if you have to go through the whole validation process to verify the method, or just show that the same results are obtained.

As a minimum you should show that the method satisfies the original validation specs and is reproducible over a number of separate occaisions.