Reference UV detector wavelength

[Consumer Products Guy]

I'm doing test method development for a new API in a fragranced consumer product using UV detector. I'm running gradient elution, so I chose a signal reference wavelength and a reference wavelength. Using a reference wavength smooths out the angled baseline, as expected. But using that reference wavelength also makes a matrix component ("maybe" 2:1 S/N) that elutes at similar retention time to my API absolutely disappear. My supervisor and QA think this is cheating for cGMP, I figure it's good science. Supervisor and QA would rather have me dilute out that peak and not use reference wavelength to ensure its peak is under 3:1 S/N, but that would end up making my baseline even more angled when zoomed enough to see.

The question is: is what I'd be doing with reference wavelength legitimate science, or is that cheating? I'd of course do linearity, accuracy, precision, etc. all using that. We have to be within 95% - 105% accuracy with spiked samples, this reference wavelength use won't affect that at all. If I had used reference wavelength since 1st try I'd never even know about that matrix tiny peak.

[HPLCaddict]

My pragmatic view was always that succesful validation is the proof that your method does was it's intended to do. If the method passes validation (and the validation itself is reasonable, of course) I'd declare the method legitimate no matter how "unusual" it might be.

[tom jupille]

In principle, I would agree with LCAddict that if you can validate the method, it's valid (a bit of tautology there!).

However, in this case, I would look long and hard, because that "cancellation" via the reference wavelength has the potential to be a method transfer nightmare:
- not all PDA's support a reference wavelength
- I would imagine that the bandpass would have an effect in there as well.

If it were my problem, my approach would depend on how "similar" the retention times are. If the peaks are close to baseline resolved, then I wouldn't worry too much. If there's more coelution, then the potential issues mentioned above would make me leery of the reference wavelength.