Possible Co-Eluter Issues - DAD

[Gizmo]

I am relatively new to exploring HPLC DAD so my apologies if some of this inquiry come across as rudimentary.

It has been brought to my attention that recent caffeine measurements are varying significantly from what was expected during time studies being conducted by R&D. The method is utilizing HPLC-DAD (one wavelength). The large decreases over time for this compound (in a complex natural product), as indicated by the results, are being questioned.

One suspicion is that a less stable co-eluter is the cause. This has not been confirmed but brings about my question(s) –

Our method, and a number of others that I have bumped into online it seems, make no mention of an attempt to conduct a “peak purity” type of check utilizing multiple wavelengths to compare the peak of interest in the sample to that of a pure standard. I understand that this is not a way to prove with certainty that no co-eluters are present but rather a way to show that they are there if variation exists in the expected pattern. But it still seems like an extremely helpful and important step when dealing with complex matrices where no 0 level control exists.

Am I wrong in feeling that this method needs to be scrapped, or at least revisited to determine if an evident co-eluter exists? And am I wrong in being a bit surprised in not seeing multi wavelength comparisons to pure standards for complex matrix samples in some of the research out there? Some make mention of using SPE or solvent extraction to remove matrix interferences but make no mention of efforts to confirm the extent of the cleanup at the RT of the analyte of interest. At least that was my take away.

I am going to be moving the caffeine assay to GCMS/SPME for now as due to present circumstances that will proceed more rapidly but I will be needing a more reliable HPLC DAD method in the future and plan on making as full an account as possible for co-eluters using multiple wavelengths and comparing each matrix to a pure standard – unless you guys tell me I’m nuts or otherwise deficient in my knowledge or critical thinking in this matter –

Thanks

[James_Ball]

It would help to confirm the identity of your Caffeine peak definitely. Short of LCMSMS, another way to check that method would be to analyze a sample then spike the sample with a known amount of Caffeine and see if you get the proper new concentration of the original value and added standard. With this you can also look more closely at the peak shape and see if there is any change in the profile which would suggest a co-elution.

Does the method require spikes of the matrix to verify efficiency of the extraction and preparation? Do you also analyze a standard after the samples to make sure the sensitivity of the instrument hasn't changed due to contamination of the column or cell?

[mattmullaney]

Good Afternoon Gizmo,

It makes sense to me to try to confirm that there is a co-elution using the DAD within the samples as compared to the standard caffeine. Also, to sharpen the peak(s) to see if there is a possible co-eluter:

1. Inject less on column with same method, examine peak shape.

2. Dilute sample with water to make the sample solvent composition, inject more on the column by the multiplier of the dilution with water with same method, examine peak shape...if possible given the sample prep.

Also...

3. Lower the time constant, otherwise same method of analysis. Should narrow peak, examine for rider peaks.

Or even...

4. Pick a HPLC column and eluent conditions that are different as possible from the incumbent method, see what happens...maybe you'll find a new peak altogether. Time consuming...also if you apply GC-MS/SPE, you're kind of doing this anyway.

[Consumer Products Guy]

Is this an isocratic elution? If so, humor me and extend the run time like 10 minutes and inject some samples, see if there are any late-eluters that just might be eluting near the analyte peak of the next injection.

If gradient assay, experiment with increasing the wash-out time and the reequilibration time.

[mattmullaney]

@ James Ball, both excellent ideas--my thanks!

[Gizmo]

These are all quite good suggestions. I now need to look into how to proceed as confirming co-eluter issues is something that needs to occur.

Just for chuckles – I’d like to describe in a little more detail what my original idea was on this that was based on the “peak purity” concept within some of the Agilent (and presumably other vendor’s) software. And when I say original I mean that it started my thought process – I actually believe someone must have done, or is doing this or something like it and I have just not asked the Google in the proper way that leads me to it.

Run pure caffeine at the assay wavelength in addition to 4 or 5 other wavelengths and determine a fingerprint pattern in terms of absorbance based on % of max (or perhaps % relative to that observed for the assay wavelength).
Compare that pattern throughout the analyte peak in the matrix – a match would be indicative of pure as the chances of an exact match in the absence of other symptoms of co-elution would be marginal (I think??) – otherwise the presence of a co-eluter is indicated.

What other wavelengths? Not sure – possibly arbitrary – but as long as a repeatable pattern emerges I am not sure that this is something that needs considerable time to determine.

I don’t know if this is a useful extension of the “peak purity” function found in the software or if it would involve less work than some of the other ideas mentioned when there are multiple matrices to be dealt with but I’d be interested in hearing thoughts on this whether in support or to shoot it out of the sky.

[mattmullaney]

Hi Gizmo,

Here is a qualitative description of the peak purity algorithm used in DAD/PDA software:

http://www.chromatographyonline.com/pea ... -detectors

This kind of"n-dimensional" vector algebra is pretty complicated stuff--I'm grateful to get my answer without much bothering or time spent.

This all said you're describing in a qualitative way what the software does--usually the wavelength range, I think, is 15 nm at the minimum (more than five or so). The absorbance ratio of two wavelengths across the peak should be constant if bands are not co-eluting...I think you could do this, though I'd say to trust the software and to use pure standards (beyond doubt, spare no expense) to do this type of assessment of peak homogeneity.

Best Wishes!

[lmh]

As an example, Shimadzu's software offers some clever stuff to check ratios across the peak, but also offers a very simple option of showing the peak at a range of different wavelengths, all superimposed. If the peak's shape changes significantly with wavelength, it's an indication of a near-coeluter with not quite the same absorbtion spectrum.
Caffeine is fairly polar, and people sometimes seem to look for it in gradients where it comes out very early. This can also make measurements unreliable.