Validation of RPC impurity method in terms of peak %

[miro2009]

Hello,
We need to validate an analytical RPC method that is used for detection of % oxidation of the main analyte in terms of peak area % relative to the main peak. We have a specified limit of 8% for our product which when exceeded the product is considered non conforming. In other words, we need the method to be able to quantify the % (in terms of % peak area) when it is around 8%, but below that, accuracy is not of concern, as long as it didn't reach 8%.
I have tried to check ICH Q2 guidelines but I can't decide under which category the method should fall: whether it is quantitative or limit test for impurities? Both require different validation tests, but in my case I feel I need a combination of both:
1- Specificity (oxidation product baseline separated form the main analyte, anything else to validate here?)
2- LOD, but also I would need to prove accuracy at the 8% level, somehow!
3- Linearity, range, precision: do I really need to perform these in my case?
What also increases the challenge is that we have 3 oxidation products, 1 arises first, followed by the other 2 with further oxidation, so to isolate the impurity "standard" to use for spiking, should I work with a mixture of all 3?? I personally think I should only use the 1 that 1st arises for both simplicity and also to be more realistic as this is the one we will usually see, the other 2 appear only after aggresive oxidation with H2o2.
I would really appreciate your opinions and suggestions on how to approach this validation, and what should be required to be proved from such a method.
I would also appreciate literature suggestions for such specific type of methods.

[tom jupille]

My $0.02 worth:

You're right, it's not a "limit test", which is usually used for something which should not be there at all.

When you get down to essentials, "validation" is defined as "demonstrating that the method does what it purports to do". On that basis what you're doing is closer to a potency analysis:

Specificity: definitely, as you propose.
LOD: Not really relevant so long as you can demonstrate accuracy and precision at the desired level.
Linearity and range: demonstrate over a reasonable interval around your 8% limit (say, 4 - 12% ?). Set up the method to report "<4%" or ">12%" if that's what you find.
Precision: DEFINITELY!! That's what will come into play if you have a "close call".

Now, for the "gotchas". Is your specification actually 8% by peak area or is it 8% by mass? If it's the former, make sure you have your backside covered in terms of how/why it was set at that level (i.e., get someone higher up to sign off on it!). If it's the latter, you should establish that the response factors of the parent and the oxidation products are the same (presumably, you know what the oxidation products are and can obtain or make reference standards). The same thing is true if you decide to use the combined areas of all the oxidation products.