Relative response in different detectors

[Mattias]

I have determined the relative response of a peak at two different wavelengths, using diluted samples where the response at both wavelengths are in the linear region. During method development I did this response factor determination on a Waters PDA detector, getting a stable factor of 10.3.

Now when I validate the method on a Waters dual-wavelength detector, the same relative response is only 8.4. I have never thought of that the relative response could change, but is apparantly so.

Can it be something in the detector design that causes this difference? Will I need to rewrite the method, and perform this response factor determination in every sequence?

[danko]

I’m not sure what you’re intending to use this relative response factor, but if it’s some kind of quantitative determination of one compound at a certain wavelength, relative to another compound at another wavelength, I’m almost convinced that it won’t work.

The reason for that is; various detectors have different signal intensities at different wavelengths (it’s mostly due to design, type of lamp, type of PMT, settings etc.).
So, if I understand the case correctly and you still intend to proceed with your plan, you’ll need to determine this RRF on a daily basis – as a part of the calibration procedure.

I’ve only used this kind of ratio determination in situations where I wanted to get an idea of a compound type - f. ex. proteins typically absorb 10 – 14 times more light at 214 nm, compared to 280 nm.

Best Regards

[Mattias]

The idea is to determine content and impurities in the same injection. At the low wavelength the main peak is too high to be used for quantitation (content) or area% calculations (>3 AU and "clipped off").

For the content determination I want to use the higher wavelength for both standards and samples (peak height about 0.5 AU). For the impurity determination I want to use the standard areas at the higher wavelength for calibration and the peak areas of the impurities at the lower wavelength. But then I need a good relative response factor...

To be on the safe side I will add one injection to determine the relative reponse in every sequence. I guess that this ratio can also be influenced by things like lamp age, bandwidth settings, smoothing functions etc? Very hard to control if the method is transferred to other labs with very different kinds of equipment. Do you think the general approach is strange or even bad practise?

[Mattias]

Maybe I should add that we consider the impurities to have the same response as the main peak. This is probably not true, but I have no better estimation at the moment.

[danko]

I don’t think the approach is too strange, but as you’ve experienced it yourself you can’t dismiss significant variations. As mentioned before, a daily calibration with a standard, containing well defined amounts of the impurities in question will do the trick. But then you’ll have to do some more validation regarding linearity/range LOD, LOQ etc. on representative systems and as you can see the task is getting bigger and bigger and kind of unconventional.
As I understand it, all the compounds have the same absorption wavelength maximum, so some of the compounds will be measured at their maxima, whilst other at a non-optimal wavelength (on a slope) and that in it self can cause significant variations and maybe justification difficulties.

Best Regards

[grzesiek]

it is kinda strange in my opinion, have you considered any other approaches?

[krickos]

Hi

Hmm are you sure this is not a technical related issue due to differences between PDAs (can differ depend on how recent model is) and "normal" UV detectors?
Seems to recall a very informative technical thread about this, but failed to find it.

In any case, in a transfer of analytical methods not long ago a collegue had the opposite case, ie the receiveving lab constantly overestimated a stability indicating impurity by some 30% with their PDA, replacing PDA with their standard UV detector resolved the issue.

[Mattias]

krickos> that is possible - but for now the reason for the effect is not so important. I will need to correct for this anyhow!

grzesiek> The most obvious approach would be to dilute the calibration standards and make one calibration curve per wavelength. But that would require at least six additional injections per sequence. I thought this was a smart shortcut, but I am starting to wonder...

[HW Mueller]

If you don´t KNOW the spectra of all the analytes then any other considerations are for the birds, anyway.
Also, again: HPLC is a technique which requires calibration.

[grzesiek]

"But that would require at least six additional injections per sequence. " - if you make your calibration six point curve

[Mattias]

HW Mueller> That is true. But at this point I am looking at a very complex degradation pattern, and I only know the identity of one impurity. I have spectra for all peaks, but I don't know the response. Identifying all the peaks and to have them synthesised is not in my budget or time allocation. It is a kind of semi-science, like all methods using area% of main peak for impurity calculations (those methods are very common in the Pharma business). Compared to the current EP method, this method at least has a decent total mass balance, I assume because of the lower wavelength.

[HW Mueller]

It is not clear to me for what all this determination of response factors is done. If you do semi-science then you assume the concentrations in a mixture of analytes, use this "assumed" mixture as a standard and calibrate your HPLC. When reporting any figures you would always have to accompany this with a statement of your assumption.
Since one is checking a calibration periodically even when only one apparatus is used it seems obvious that one does a calibration when going to a different setup.
If one has both apparatie calibrated and gets different answes something was done or is wrong.

[Mattias]

Going back to the original question (assuming equal response of impurities and main peak).

I cannot use area% since the main peak has severly overloaded signal (I get about 80% of the expected area response due to non-linearity). I then want to use the same main peak but at a longer wavelength to be in the linear region and then convert the area back to the lower wavelength by a response factor. So it is a relative response factor of the same peak but at different wavelengths. This relative response factor is not the same on different instruments.

It might be a strange approach, but it would save time for me. Regarding the error of the impurity determinations, all impurity peaks will surely be either over or underestimated. But the overall mass balance look fine in all stress conditions (i.e. if i loose 10% of the main peak by hydrolysis or oxidation - I find 10% impurities, give or take a percent).

[lmh]

To go back to the original question, yes, it is possible for two different instruments to give you a different absorbance ratio measured at two wavelengths, because of optical differences in instrument. Two obvious examples:

(1) Bandwidth: obviously if bandwidth differs, so does absorbance. To take the extreme case, if you imagine an absorbance peak that stretches from 300-400nm, if you measure it at 350 and 360nm with a 120nm bandwidth, you'll get no difference whatsoever, but if you measure with a 1nm slit width, you probably will.

(2) Wavelength accuracy: if one instrument is very slightly off wavelength, then the two instruments will agree when looking at the wavelength of maximum absorbance (because here the curve is nearly flat), but disagree when measuring half way down the slope of the absorbance curve (because here there is a big slope).

[HW Mueller]

Practically it is taken care of by a calibration. Theoretically, look at a desrciption of where
the Beer-Lambert Law holds.